ntp_io.c revision 276992
1/*
2 * ntp_io.c - input/output routines for ntpd.	The socket-opening code
3 *		   was shamelessly stolen from ntpd.
4 */
5
6#ifdef HAVE_CONFIG_H
7# include <config.h>
8#endif
9
10#include "ntp_machine.h"
11#include "ntpd.h"
12#include "ntp_io.h"
13#include "iosignal.h"
14#include "ntp_refclock.h"
15#include "ntp_stdlib.h"
16#include "ntp_request.h"
17#include "ntp.h"
18#include "ntp_unixtime.h"
19
20/* Don't include ISC's version of IPv6 variables and structures */
21#define ISC_IPV6_H 1
22#include <isc/interfaceiter.h>
23#include <isc/list.h>
24#include <isc/result.h>
25
26#ifdef SIM
27#include "ntpsim.h"
28#endif
29
30#include <stdio.h>
31#include <signal.h>
32#ifdef HAVE_SYS_PARAM_H
33# include <sys/param.h>
34#endif /* HAVE_SYS_PARAM_H */
35#ifdef HAVE_SYS_IOCTL_H
36# include <sys/ioctl.h>
37#endif
38#ifdef HAVE_SYS_SOCKIO_H	/* UXPV: SIOC* #defines (Frank Vance <fvance@waii.com>) */
39# include <sys/sockio.h>
40#endif
41#ifdef HAVE_SYS_UIO_H
42# include <sys/uio.h>
43#endif
44
45/*
46 * setsockopt does not always have the same arg declaration
47 * across all platforms. If it's not defined we make it empty
48 */
49
50#ifndef SETSOCKOPT_ARG_CAST
51#define SETSOCKOPT_ARG_CAST
52#endif
53
54/*
55 * Set up some macros to look for IPv6 and IPv6 multicast
56 */
57
58#if defined(ISC_PLATFORM_HAVEIPV6) && !defined(DISABLE_IPV6)
59
60#define INCLUDE_IPV6_SUPPORT
61
62#if defined(INCLUDE_IPV6_SUPPORT) && defined(IPV6_JOIN_GROUP) && defined(IPV6_LEAVE_GROUP)
63#define INCLUDE_IPV6_MULTICAST_SUPPORT
64
65#endif	/* IPV6 Multicast Support */
66#endif  /* IPv6 Support */
67
68#ifdef INCLUDE_IPV6_SUPPORT
69#include <netinet/in.h>
70#include <net/if_var.h>
71#include <netinet/in_var.h>
72#endif /* !INCLUDE_IPV6_SUPPORT */
73
74extern int listen_to_virtual_ips;
75extern const char *specific_interface;
76
77#if defined(SO_TIMESTAMP) && defined(SCM_TIMESTAMP)
78#if defined(CMSG_FIRSTHDR)
79#define HAVE_TIMESTAMP
80#define USE_TIMESTAMP_CMSG
81#ifndef TIMESTAMP_CTLMSGBUF_SIZE
82#define TIMESTAMP_CTLMSGBUF_SIZE 1536 /* moderate default */
83#endif
84#else
85/* fill in for old/other timestamp interfaces */
86#endif
87#endif
88
89#if defined(SYS_WINNT)
90#include <transmitbuff.h>
91#include <isc/win32os.h>
92/*
93 * Define this macro to control the behavior of connection
94 * resets on UDP sockets.  See Microsoft KnowledgeBase Article Q263823
95 * for details.
96 * NOTE: This requires that Windows 2000 systems install Service Pack 2
97 * or later.
98 */
99#ifndef SIO_UDP_CONNRESET
100#define SIO_UDP_CONNRESET _WSAIOW(IOC_VENDOR,12)
101#endif
102
103/*
104 * Windows C runtime ioctl() can't deal properly with sockets,
105 * map to ioctlsocket for this source file.
106 */
107#define ioctl(fd, opt, val)  ioctlsocket((fd), (opt), (u_long *)(val))
108#endif  /* SYS_WINNT */
109
110/*
111 * We do asynchronous input using the SIGIO facility.  A number of
112 * recvbuf buffers are preallocated for input.	In the signal
113 * handler we poll to see which sockets are ready and read the
114 * packets from them into the recvbuf's along with a time stamp and
115 * an indication of the source host and the interface it was received
116 * through.  This allows us to get as accurate receive time stamps
117 * as possible independent of other processing going on.
118 *
119 * We watch the number of recvbufs available to the signal handler
120 * and allocate more when this number drops below the low water
121 * mark.  If the signal handler should run out of buffers in the
122 * interim it will drop incoming frames, the idea being that it is
123 * better to drop a packet than to be inaccurate.
124 */
125
126
127/*
128 * Other statistics of possible interest
129 */
130volatile u_long packets_dropped;	/* total number of packets dropped on reception */
131volatile u_long packets_ignored;	/* packets received on wild card interface */
132volatile u_long packets_received;	/* total number of packets received */
133u_long packets_sent;	/* total number of packets sent */
134u_long packets_notsent; /* total number of packets which couldn't be sent */
135
136volatile u_long handler_calls;	/* number of calls to interrupt handler */
137volatile u_long handler_pkts;	/* number of pkts received by handler */
138u_long io_timereset;		/* time counters were reset */
139
140/*
141 * Interface stuff
142 */
143struct interface *any_interface;	/* default ipv4 interface */
144struct interface *any6_interface;	/* default ipv6 interface */
145struct interface *loopback_interface;	/* loopback ipv4 interface */
146
147int ninterfaces;			/* Total number of interfaces */
148
149volatile int disable_dynamic_updates;   /* when set to != 0 dynamic updates won't happen */
150
151#ifdef REFCLOCK
152/*
153 * Refclock stuff.	We keep a chain of structures with data concerning
154 * the guys we are doing I/O for.
155 */
156static	struct refclockio *refio;
157#endif /* REFCLOCK */
158
159
160/*
161 * Define what the possible "soft" errors can be.  These are non-fatal returns
162 * of various network related functions, like recv() and so on.
163 *
164 * For some reason, BSDI (and perhaps others) will sometimes return <0
165 * from recv() but will have errno==0.  This is broken, but we have to
166 * work around it here.
167 */
168#define SOFT_ERROR(e)	((e) == EAGAIN || \
169			 (e) == EWOULDBLOCK || \
170			 (e) == EINTR || \
171			 (e) == 0)
172
173/*
174 * File descriptor masks etc. for call to select
175 * Not needed for I/O Completion Ports
176 */
177fd_set activefds;
178int maxactivefd;
179/*
180 * bit alternating value to detect verified interfaces during an update cycle
181 */
182static  u_char          sys_interphase = 0;
183
184static  struct interface *new_interface P((struct interface *));
185static  void add_interface P((struct interface *));
186static  int update_interfaces P((u_short, interface_receiver_t, void *));
187static  void remove_interface P((struct interface *));
188static  struct interface *create_interface P((u_short, struct interface *));
189
190static int	move_fd		P((SOCKET));
191
192/*
193 * Multicast functions
194 */
195static	isc_boolean_t	addr_ismulticast	 P((struct sockaddr_storage *));
196/*
197 * Not all platforms support multicast
198 */
199#ifdef MCAST
200static	isc_boolean_t	socket_multicast_enable	 P((struct interface *, int, struct sockaddr_storage *));
201static	isc_boolean_t	socket_multicast_disable P((struct interface *, struct sockaddr_storage *));
202#endif
203
204#ifdef DEBUG
205static void print_interface	P((struct interface *, char *, char *));
206#define DPRINT_INTERFACE(_LVL_, _ARGS_) do { if (debug >= (_LVL_)) { print_interface _ARGS_; } } while (0)
207#else
208#define DPRINT_INTERFACE(_LVL_, _ARGS_) do {} while (0)
209#endif
210
211typedef struct vsock vsock_t;
212enum desc_type { FD_TYPE_SOCKET, FD_TYPE_FILE };
213
214struct vsock {
215	SOCKET				fd;
216	enum desc_type                  type;
217	ISC_LINK(vsock_t)		link;
218};
219
220#if !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET)
221/*
222 * async notification processing (e. g. routing sockets)
223 */
224/*
225 * support for receiving data on fd that is not a refclock or a socket
226 * like e. g. routing sockets
227 */
228struct asyncio_reader {
229	SOCKET fd;		                    /* fd to be read */
230	void  *data;		                    /* possibly local data */
231	void (*receiver)(struct asyncio_reader *);  /* input handler */
232	ISC_LINK(struct asyncio_reader) link;       /* the list this is being kept in */
233};
234
235ISC_LIST(struct asyncio_reader) asyncio_reader_list;
236
237static void delete_asyncio_reader P((struct asyncio_reader *));
238static struct asyncio_reader *new_asyncio_reader P((void));
239static void add_asyncio_reader P((struct asyncio_reader *, enum desc_type));
240static void remove_asyncio_reader P((struct asyncio_reader *));
241
242#endif /* !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET) */
243
244static void init_async_notifications P((void));
245
246static	int create_sockets	P((u_short));
247static	SOCKET	open_socket	P((struct sockaddr_storage *, int, int, struct interface *));
248static	char *	fdbits		P((int, fd_set *));
249static	void	set_reuseaddr	P((int));
250static	isc_boolean_t	socket_broadcast_enable	 P((struct interface *, SOCKET, struct sockaddr_storage *));
251static	isc_boolean_t	socket_broadcast_disable P((struct interface *, struct sockaddr_storage *));
252
253ISC_LIST(vsock_t)	fd_list;
254
255typedef struct remaddr remaddr_t;
256
257struct remaddr {
258      struct sockaddr_storage	 addr;
259      struct interface               *interface;
260      ISC_LINK(remaddr_t)	 link;
261};
262
263ISC_LIST(remaddr_t)       remoteaddr_list;
264
265ISC_LIST(struct interface)     inter_list;
266
267static struct interface *wildipv4 = NULL;
268static struct interface *wildipv6 = NULL;
269
270static void	add_fd_to_list	P((SOCKET, enum desc_type));
271static void	close_and_delete_fd_from_list	P((SOCKET));
272static void	add_addr_to_list	P((struct sockaddr_storage *, struct interface *));
273static void	delete_addr_from_list	P((struct sockaddr_storage *));
274static struct interface *find_addr_in_list	P((struct sockaddr_storage *));
275static struct interface *find_flagged_addr_in_list P((struct sockaddr_storage *, int));
276static void	create_wildcards	P((u_short));
277static isc_boolean_t	address_okay	P((struct interface *));
278static void		convert_isc_if		P((isc_interface_t *, struct interface *, u_short));
279static void	delete_interface_from_list	P((struct interface *));
280static struct interface *getinterface	P((struct sockaddr_storage *, int));
281static struct interface *findlocalinterface	P((struct sockaddr_storage *, int));
282static struct interface *findlocalcastinterface	P((struct sockaddr_storage *, int));
283
284/*
285 * Routines to read the ntp packets
286 */
287#if !defined(HAVE_IO_COMPLETION_PORT)
288static inline int     read_network_packet	P((SOCKET, struct interface *, l_fp));
289static inline int     read_refclock_packet	P((SOCKET, struct refclockio *, l_fp));
290#endif
291
292#ifdef SYS_WINNT
293/*
294 * Windows 2000 systems incorrectly cause UDP sockets using WASRecvFrom
295 * to not work correctly, returning a WSACONNRESET error when a WSASendTo
296 * fails with an "ICMP port unreachable" response and preventing the
297 * socket from using the WSARecvFrom in subsequent operations.
298 * The function below fixes this, but requires that Windows 2000
299 * Service Pack 2 or later be installed on the system.  NT 4.0
300 * systems are not affected by this and work correctly.
301 * See Microsoft Knowledge Base Article Q263823 for details of this.
302 */
303void
304connection_reset_fix(
305	SOCKET fd,
306	struct sockaddr_storage *addr
307	)
308{
309	DWORD dwBytesReturned = 0;
310	BOOL  bNewBehavior = FALSE;
311	DWORD status;
312
313	/*
314	 * disable bad behavior using IOCTL: SIO_UDP_CONNRESET
315	 * NT 4.0 has no problem
316	 */
317	if (isc_win32os_majorversion() >= 5) {
318		status = WSAIoctl(fd, SIO_UDP_CONNRESET, &bNewBehavior,
319				  sizeof(bNewBehavior), NULL, 0,
320				  &dwBytesReturned, NULL, NULL);
321		if (SOCKET_ERROR == status)
322			netsyslog(LOG_ERR, "connection_reset_fix() "
323					   "failed for address %s: %m",
324					   stoa(addr));
325	}
326}
327#endif
328
329/*
330 * on Unix systems the stdio library typically
331 * makes use of file descriptors in the lower
332 * integer range. stdio usually will make use
333 * of the file descriptor in the range of
334 * [0..FOPEN_MAX)
335 * in order to keep this range clean for socket
336 * file descriptors we attempt to move them above
337 * FOPEM_MAX. This is not as easy as it sounds as
338 * FOPEN_MAX changes from implementation to implementation
339 * and may exceed to current file decriptor limits.
340 * We are using following strategy:
341 * - keep a current socket fd boundary initialized with
342 *   max(0, min(getdtablesize() - FD_CHUNK, FOPEN_MAX))
343 * - attempt to move the descriptor to the boundary or
344 *   above.
345 *   - if that fails and boundary > 0 set boundary
346 *     to min(0, socket_fd_boundary - FD_CHUNK)
347 *     -> retry
348 *     if failure and boundary == 0 return old fd
349 *   - on success close old fd return new fd
350 *
351 * effects:
352 *   - fds will be moved above the socket fd boundary
353 *     if at all possible.
354 *   - the socket boundary will be reduced until
355 *     allocation is possible or 0 is reached - at this
356 *     point the algrithm will be disabled
357 */
358static int move_fd(SOCKET fd)
359{
360#if !defined(SYS_WINNT) && defined(F_DUPFD)
361#ifndef FD_CHUNK
362#define FD_CHUNK	10
363#endif
364/*
365 * number of fds we would like to have for
366 * stdio FILE* available.
367 * we can pick a "low" number as our use of
368 * FILE* is limited to log files and temporarily
369 * to data and config files. Except for log files
370 * we don't keep the other FILE* open beyond the
371 * scope of the function that opened it.
372 */
373#ifndef FD_PREFERRED_SOCKBOUNDARY
374#define FD_PREFERRED_SOCKBOUNDARY 48
375#endif
376
377#ifndef HAVE_GETDTABLESIZE
378/*
379 * if we have no idea about the max fd value set up things
380 * so we will start at FOPEN_MAX
381 */
382#define getdtablesize() (FOPEN_MAX+FD_CHUNK)
383#endif
384
385#ifndef FOPEN_MAX
386#define FOPEN_MAX	20	/* assume that for the lack of anything better */
387#endif
388	static SOCKET socket_boundary = -1;
389	SOCKET newfd;
390
391	/*
392	 * check whether boundary has be set up
393	 * already
394	 */
395	if (socket_boundary == -1) {
396		socket_boundary = max(0, min(getdtablesize() - FD_CHUNK,
397					     min(FOPEN_MAX, FD_PREFERRED_SOCKBOUNDARY)));
398#ifdef DEBUG
399		msyslog(LOG_DEBUG, "ntp_io: estimated max descriptors: %d, initial socket boundary: %d",
400			getdtablesize(), socket_boundary);
401#endif
402	}
403
404	/*
405	 * Leave a space for stdio to work in. potentially moving the
406	 * socket_boundary lower until allocation succeeds.
407	 */
408	do {
409		if (fd >= 0 && fd < socket_boundary) {
410			/* inside reserved range: attempt to move fd */
411			newfd = fcntl(fd, F_DUPFD, socket_boundary);
412
413			if (newfd != -1) {
414				/* success: drop the old one - return the new one */
415				(void)close(fd);
416				return (newfd);
417			}
418		} else {
419			/* outside reserved range: no work - return the original one */
420			return (fd);
421		}
422		socket_boundary = max(0, socket_boundary - FD_CHUNK);
423#ifdef DEBUG
424		msyslog(LOG_DEBUG, "ntp_io: selecting new socket boundary: %d",
425			socket_boundary);
426#endif
427	} while (socket_boundary > 0);
428#endif /* !defined(SYS_WINNT) && defined(F_DUPFD) */
429	return (fd);
430}
431
432#ifdef DEBUG_TIMING
433/*
434 * collect timing information for various processing
435 * paths. currently we only pass then on to the file
436 * for later processing. this could also do histogram
437 * based analysis in other to reduce the load (and skew)
438 * dur to the file output
439 */
440void
441collect_timing(struct recvbuf *rb, const char *tag, int count, l_fp *dts)
442{
443	char buf[2048];
444
445	snprintf(buf, sizeof(buf), "%s %d %s %s",
446		 (rb != NULL) ?
447		 ((rb->dstadr) ? stoa(&rb->recv_srcadr) : "-REFCLOCK-") : "-",
448		 count, lfptoa(dts, 9), tag);
449	record_timing_stats(buf);
450}
451#endif
452
453/*
454 * About dynamic interfaces, sockets, reception and more...
455 *
456 * the code solves following tasks:
457 *
458 *   - keep a current list of active interfaces in order
459 *     to bind to to the interface address on NTP_PORT so that
460 *     all wild and specific bindings for NTP_PORT are taken by ntpd
461 *     to avoid other daemons messing with the time or sockets.
462 *   - all interfaces keep a list of peers that are referencing
463 *     the interface in order to quickly re-assign the peers to
464 *     new interface in case an interface is deleted (=> gone from system or
465 *     down)
466 *   - have a preconfigured socket ready with the right local address
467 *     for transmission and reception
468 *   - have an address list for all destination addresses used within ntpd
469 *     to find the "right" preconfigured socket.
470 *   - facilitate updating the internal interface list with respect to
471 *     the current kernel state
472 *
473 * special issues:
474 *
475 *   - mapping of multicast addresses to the interface affected is not always
476 *     one to one - especially on hosts with multiple interfaces
477 *     the code here currently allocates a separate interface entry for those
478 *     multicast addresses
479 *     iff it is able to bind to a *new* socket with the multicast address (flags |= MCASTIF)
480 *     in case of failure the multicast address is bound to an existing interface.
481 *   - on some systems it is perfectly legal to assign the same address to
482 *     multiple interfaces. Therefore this code does not keep a list of interfaces
483 *     but a list of interfaces that represent a unique address as determined by the kernel
484 *     by the procedure in findlocalinterface. Thus it is perfectly legal to see only
485 *     one representative of a group of real interfaces if they share the same address.
486 *
487 * Frank Kardel 20050910
488 */
489
490/*
491 * init_io - initialize I/O data structures and call socket creation routine
492 */
493void
494init_io(void)
495{
496#ifdef SYS_WINNT
497	init_io_completion_port();
498
499	if (!Win32InitSockets())
500	{
501		netsyslog(LOG_ERR, "No useable winsock.dll: %m");
502		exit(1);
503	}
504	init_transmitbuff();
505#endif /* SYS_WINNT */
506
507	/*
508	 * Init buffer free list and stat counters
509	 */
510	init_recvbuff(RECV_INIT);
511
512	packets_dropped = packets_received = 0;
513	packets_ignored = 0;
514	packets_sent = packets_notsent = 0;
515	handler_calls = handler_pkts = 0;
516	io_timereset = 0;
517	loopback_interface = NULL;
518	any_interface = NULL;
519	any6_interface = NULL;
520
521#ifdef REFCLOCK
522	refio = NULL;
523#endif
524
525#if defined(HAVE_SIGNALED_IO)
526	(void) set_signal();
527#endif
528
529	ISC_LIST_INIT(fd_list);
530
531#if !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET)
532	ISC_LIST_INIT(asyncio_reader_list);
533#endif
534
535        ISC_LIST_INIT(remoteaddr_list);
536
537	ISC_LIST_INIT(inter_list);
538
539	/*
540	 * Create the sockets
541	 */
542	BLOCKIO();
543	(void) create_sockets(htons(NTP_PORT));
544	UNBLOCKIO();
545
546	init_async_notifications();
547
548	DPRINTF(3, ("init_io: maxactivefd %d\n", maxactivefd));
549}
550
551#ifdef DEBUG
552/*
553 * function to dump the contents of the interface structure
554 * for debugging use only.
555 */
556void
557interface_dump(struct interface *itf)
558{
559	u_char* cp;
560	int i;
561	/* Limit the size of the sockaddr_storage hex dump */
562	int maxsize = min(32, sizeof(struct sockaddr_storage));
563
564	printf("Dumping interface: %p\n", itf);
565	printf("fd = %d\n", itf->fd);
566	printf("bfd = %d\n", itf->bfd);
567	printf("sin = %s,\n", stoa(&(itf->sin)));
568	cp = (u_char*) &(itf->sin);
569	for(i = 0; i < maxsize; i++)
570	{
571		printf("%02x", *cp++);
572		if((i+1)%4 == 0)
573			printf(" ");
574	}
575	printf("\n");
576	printf("bcast = %s,\n", stoa(&(itf->bcast)));
577	cp = (u_char*) &(itf->bcast);
578	for(i = 0; i < maxsize; i++)
579	{
580		printf("%02x", *cp++);
581		if((i+1)%4 == 0)
582			printf(" ");
583	}
584	printf("\n");
585	printf("mask = %s,\n", stoa(&(itf->mask)));
586	cp = (u_char*) &(itf->mask);
587	for(i = 0; i < maxsize; i++)
588	{
589		printf("%02x", *cp++);
590		if((i+1)%4 == 0)
591			printf(" ");
592	}
593	printf("\n");
594	printf("name = %s\n", itf->name);
595	printf("flags = 0x%08x\n", itf->flags);
596	printf("last_ttl = %d\n", itf->last_ttl);
597	printf("addr_refid = %08x\n", itf->addr_refid);
598	printf("num_mcast = %d\n", itf->num_mcast);
599	printf("received = %ld\n", itf->received);
600	printf("sent = %ld\n", itf->sent);
601	printf("notsent = %ld\n", itf->notsent);
602	printf("ifindex = %u\n", itf->ifindex);
603	printf("scopeid = %u\n", itf->scopeid);
604	printf("peercnt = %u\n", itf->peercnt);
605	printf("phase = %u\n", itf->phase);
606}
607
608/*
609 * print_interface - helper to output debug information
610 */
611static void
612print_interface(struct interface *iface, char *pfx, char *sfx)
613{
614	printf("%sinterface #%d: fd=%d, bfd=%d, name=%s, flags=0x%x, scope=%d, ifindex=%d",
615	       pfx,
616	       iface->ifnum,
617	       iface->fd,
618	       iface->bfd,
619	       iface->name,
620	       iface->flags,
621	       iface->scopeid,
622	       iface->ifindex);
623	/* Leave these as three printf calls. */
624	printf(", sin=%s",
625	       stoa((&iface->sin)));
626	if (iface->flags & INT_BROADCAST)
627		printf(", bcast=%s,",
628		       stoa((&iface->bcast)));
629	if (iface->family == AF_INET)
630	  printf(", mask=%s",
631		 stoa((&iface->mask)));
632	printf(", %s:%s", iface->ignore_packets == ISC_FALSE ? "Enabled" : "Disabled", sfx);
633	if (debug > 4)	/* in-depth debugging only */
634		interface_dump(iface);
635}
636
637#endif
638
639#if !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET)
640/*
641 * create an asyncio_reader structure
642 */
643static struct asyncio_reader *
644new_asyncio_reader()
645{
646	struct asyncio_reader *reader;
647
648	reader = (struct asyncio_reader *)emalloc(sizeof(struct asyncio_reader));
649
650	memset((char *)reader, 0, sizeof(*reader));
651	ISC_LINK_INIT(reader, link);
652	reader->fd = INVALID_SOCKET;
653	return reader;
654}
655
656/*
657 * delete a reader
658 */
659static void
660delete_asyncio_reader(struct asyncio_reader *reader)
661{
662	free(reader);
663}
664
665/*
666 * add asynchio_reader
667 */
668static void
669add_asyncio_reader(struct asyncio_reader *reader, enum desc_type type)
670{
671	ISC_LIST_APPEND(asyncio_reader_list, reader, link);
672	add_fd_to_list(reader->fd, type);
673}
674
675/*
676 * remove asynchio_reader
677 */
678static void
679remove_asyncio_reader(struct asyncio_reader *reader)
680{
681	ISC_LIST_UNLINK_TYPE(asyncio_reader_list, reader, link, struct asyncio_reader);
682
683	if (reader->fd != INVALID_SOCKET)
684		close_and_delete_fd_from_list(reader->fd);
685
686	reader->fd = INVALID_SOCKET;
687}
688#endif /* !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET) */
689
690/*
691 * interface list enumerator - visitor pattern
692 */
693void
694interface_enumerate(interface_receiver_t receiver, void *data)
695{
696	interface_info_t ifi;
697        struct interface *interf;
698
699	ifi.action = IFS_EXISTS;
700
701	for (interf = ISC_LIST_HEAD(inter_list);
702	     interf != NULL;
703	     interf = ISC_LIST_NEXT(interf, link)) {
704		ifi.interface = interf;
705		receiver(data, &ifi);
706	}
707}
708
709/*
710 * do standard initialization of interface structure
711 */
712static void
713init_interface(struct interface *interface)
714{
715	memset((char *)interface, 0, sizeof(struct interface));
716	ISC_LINK_INIT(interface, link);
717	ISC_LIST_INIT(interface->peers);
718	interface->fd = INVALID_SOCKET;
719	interface->bfd = INVALID_SOCKET;
720	interface->num_mcast = 0;
721	interface->received = 0;
722	interface->sent = 0;
723	interface->notsent = 0;
724	interface->peercnt = 0;
725	interface->phase = sys_interphase;
726}
727
728/*
729 * create new interface structure initialize from
730 * template structure or via standard initialization
731 * function
732 */
733static struct interface *
734new_interface(struct interface *interface)
735{
736	static u_int sys_ifnum = 0;
737
738	struct interface *iface = (struct interface *)emalloc(sizeof(struct interface));
739
740	if (interface != NULL)
741	{
742		memcpy((char*)iface, (char*)interface, sizeof(*interface));
743	}
744	else
745	{
746		init_interface(iface);
747	}
748
749	iface->ifnum = sys_ifnum++;  /* count every new instance of an interface in the system */
750	iface->starttime = current_time;
751
752	return iface;
753}
754
755/*
756 * return interface storage into free memory pool
757 */
758static void
759delete_interface(struct interface *interface)
760{
761	free(interface);
762}
763
764/*
765 * link interface into list of known interfaces
766 */
767static void
768add_interface(struct interface *interface)
769{
770	static struct interface *listhead = NULL;
771
772	/*
773	 * For ntpd, the first few interfaces (wildcard, localhost)
774	 * will never be removed.  This means inter_list.head is
775	 * unchanging once initialized.  Take advantage of that to
776	 * watch for changes and catch corruption earlier.  This
777	 * helped track down corruption caused by using FD_SET with
778	 * a descriptor numerically larger than FD_SETSIZE.
779	 */
780	if (NULL == listhead)
781		listhead = inter_list.head;
782
783	if (listhead != inter_list.head) {
784		msyslog(LOG_ERR, "add_interface inter_list.head corrupted: was %p now %p",
785			listhead, inter_list.head);
786		exit(1);
787	}
788	/*
789	 * Calculate the address hash
790	 */
791	interface->addr_refid = addr2refid(&interface->sin);
792
793	ISC_LIST_APPEND(inter_list, interface, link);
794	ninterfaces++;
795}
796
797/*
798 * remove interface from known interface list and clean up
799 * associated resources
800 */
801static void
802remove_interface(struct interface *interface)
803{
804	struct sockaddr_storage resmask;
805
806	ISC_LIST_UNLINK_TYPE(inter_list, interface, link, struct interface);
807
808	delete_interface_from_list(interface);
809
810	if (interface->fd != INVALID_SOCKET)
811	{
812		msyslog(LOG_INFO, "Deleting interface #%d %s, %s#%d, interface stats: received=%ld, sent=%ld, dropped=%ld, active_time=%ld secs",
813			interface->ifnum,
814			interface->name,
815			stoa((&interface->sin)),
816			NTP_PORT,  /* XXX should extract port from sin structure */
817			interface->received,
818			interface->sent,
819			interface->notsent,
820			current_time - interface->starttime);
821
822		close_and_delete_fd_from_list(interface->fd);
823	}
824
825	if (interface->bfd != INVALID_SOCKET)
826	{
827		msyslog(LOG_INFO, "Deleting interface #%d %s, broadcast address %s#%d",
828			interface->ifnum,
829			interface->name,
830			stoa((&interface->bcast)),
831			(u_short) NTP_PORT);  /* XXX extract port from sin structure */
832		close_and_delete_fd_from_list(interface->bfd);
833	}
834
835	ninterfaces--;
836	ntp_monclearinterface(interface);
837
838	/* remove restrict interface entry */
839
840	/*
841	 * Blacklist bound interface address
842	 */
843	SET_HOSTMASK(&resmask, interface->sin.ss_family);
844	hack_restrict(RESTRICT_REMOVEIF, &interface->sin, &resmask,
845		      RESM_NTPONLY|RESM_INTERFACE, RES_IGNORE);
846}
847
848static void
849list_if_listening(struct interface *interface, u_short port)
850{
851	msyslog(LOG_INFO, "Listening on interface #%d %s, %s#%d %s",
852		interface->ifnum,
853		interface->name,
854		stoa((&interface->sin)),
855		ntohs( (u_short) port),
856		(interface->ignore_packets == ISC_FALSE) ?
857		"Enabled": "Disabled");
858}
859
860static void
861create_wildcards(u_short port) {
862	isc_boolean_t okipv4 = ISC_TRUE;
863	/*
864	 * create pseudo-interface with wildcard IPv4 address
865	 */
866#ifdef IPV6_V6ONLY
867	if(isc_net_probeipv4() != ISC_R_SUCCESS)
868		okipv4 = ISC_FALSE;
869#endif
870
871	if(okipv4 == ISC_TRUE) {
872	        struct interface *interface = new_interface(NULL);
873
874		interface->family = AF_INET;
875		interface->sin.ss_family = AF_INET;
876		((struct sockaddr_in*)&interface->sin)->sin_addr.s_addr = htonl(INADDR_ANY);
877		((struct sockaddr_in*)&interface->sin)->sin_port = port;
878		(void) strncpy(interface->name, "wildcard", sizeof(interface->name));
879		interface->mask.ss_family = AF_INET;
880		((struct sockaddr_in*)&interface->mask)->sin_addr.s_addr = htonl(~(u_int32)0);
881		interface->flags = INT_BROADCAST | INT_UP | INT_WILDCARD;
882		interface->ignore_packets = ISC_TRUE;
883#if defined(MCAST)
884		/*
885		 * enable possible multicast reception on the broadcast socket
886		 */
887		interface->bcast.ss_family = AF_INET;
888		((struct sockaddr_in*)&interface->bcast)->sin_port = port;
889		((struct sockaddr_in*)&interface->bcast)->sin_addr.s_addr = htonl(INADDR_ANY);
890#endif /* MCAST */
891		interface->fd = open_socket(&interface->sin,
892				 interface->flags, 1, interface);
893
894		if (interface->fd != INVALID_SOCKET) {
895			wildipv4 = interface;
896			any_interface = interface;
897
898			add_addr_to_list(&interface->sin, interface);
899			add_interface(interface);
900			list_if_listening(interface, port);
901		} else {
902			msyslog(LOG_ERR, "unable to bind to wildcard socket address %s - another process may be running - EXITING",
903				stoa((&interface->sin)));
904			exit(1);
905		}
906	}
907
908#ifdef INCLUDE_IPV6_SUPPORT
909	/*
910	 * create pseudo-interface with wildcard IPv6 address
911	 */
912	if (isc_net_probeipv6() == ISC_R_SUCCESS) {
913	        struct interface *interface = new_interface(NULL);
914
915		interface->family = AF_INET6;
916		interface->sin.ss_family = AF_INET6;
917		((struct sockaddr_in6*)&interface->sin)->sin6_addr = in6addr_any;
918 		((struct sockaddr_in6*)&interface->sin)->sin6_port = port;
919# ifdef ISC_PLATFORM_HAVESCOPEID
920 		((struct sockaddr_in6*)&interface->sin)->sin6_scope_id = 0;
921# endif
922		(void) strncpy(interface->name, "wildcard", sizeof(interface->name));
923		interface->mask.ss_family = AF_INET6;
924		memset(&((struct sockaddr_in6*)&interface->mask)->sin6_addr.s6_addr, 0xff, sizeof(struct in6_addr));
925		interface->flags = INT_UP | INT_WILDCARD;
926		interface->ignore_packets = ISC_TRUE;
927
928		interface->fd = open_socket(&interface->sin,
929				 interface->flags, 1, interface);
930
931		if (interface->fd != INVALID_SOCKET) {
932			wildipv6 = interface;
933			any6_interface = interface;
934			add_addr_to_list(&interface->sin, interface);
935			add_interface(interface);
936			list_if_listening(interface, port);
937		} else {
938			msyslog(LOG_ERR, "unable to bind to wildcard socket address %s - another process may be running - EXITING",
939				stoa((&interface->sin)));
940			exit(1);
941		}
942	}
943#endif
944}
945
946
947static isc_boolean_t
948address_okay(struct interface *iface) {
949
950	DPRINTF(4, ("address_okay: listen Virtual: %d, IF name: %s\n",
951		    listen_to_virtual_ips, iface->name));
952
953	/*
954	 * Always allow the loopback
955	 */
956	if((iface->flags & INT_LOOPBACK) != 0) {
957		DPRINTF(4, ("address_okay: loopback - OK\n"));
958		return (ISC_TRUE);
959	}
960
961	/*
962	 * Check if the interface is specified
963	 */
964	if (specific_interface != NULL) {
965		if (strcasecmp(iface->name, specific_interface) == 0) {
966			DPRINTF(4, ("address_okay: specific interface name matched - OK\n"));
967			return (ISC_TRUE);
968		} else {
969			DPRINTF(4, ("address_okay: specific interface name NOT matched - FAIL\n"));
970			return (ISC_FALSE);
971		}
972	}
973	else {
974		if (listen_to_virtual_ips == 0  &&
975		    (strchr(iface->name, (int)':') != NULL)) {
976			DPRINTF(4, ("address_okay: virtual ip/alias - FAIL\n"));
977			return (ISC_FALSE);
978		}
979	}
980
981	DPRINTF(4, ("address_okay: OK\n"));
982	return (ISC_TRUE);
983}
984
985static void
986convert_isc_if(isc_interface_t *isc_if, struct interface *itf, u_short port)
987{
988	itf->scopeid = 0;
989	itf->family = (short) isc_if->af;
990	strcpy(itf->name, isc_if->name);
991
992	if(isc_if->af == AF_INET) {
993		itf->sin.ss_family = (u_short) isc_if->af;
994		memcpy(&(((struct sockaddr_in*)&itf->sin)->sin_addr),
995		       &(isc_if->address.type.in),
996		       sizeof(struct in_addr));
997		((struct sockaddr_in*)&itf->sin)->sin_port = port;
998
999		if((isc_if->flags & INTERFACE_F_BROADCAST) != 0) {
1000			itf->flags |= INT_BROADCAST;
1001			itf->bcast.ss_family = itf->sin.ss_family;
1002			memcpy(&(((struct sockaddr_in*)&itf->bcast)->sin_addr),
1003			       &(isc_if->broadcast.type.in),
1004				 sizeof(struct in_addr));
1005			((struct sockaddr_in*)&itf->bcast)->sin_port = port;
1006		}
1007
1008		itf->mask.ss_family = itf->sin.ss_family;
1009		memcpy(&(((struct sockaddr_in*)&itf->mask)->sin_addr),
1010		       &(isc_if->netmask.type.in),
1011		       sizeof(struct in_addr));
1012		((struct sockaddr_in*)&itf->mask)->sin_port = port;
1013	}
1014#ifdef INCLUDE_IPV6_SUPPORT
1015	else if (isc_if->af == AF_INET6) {
1016		itf->sin.ss_family = (u_short) isc_if->af;
1017		memcpy(&(((struct sockaddr_in6 *)&itf->sin)->sin6_addr),
1018		       &(isc_if->address.type.in6),
1019		       sizeof(((struct sockaddr_in6 *)&itf->sin)->sin6_addr));
1020		((struct sockaddr_in6 *)&itf->sin)->sin6_port = port;
1021
1022#ifdef ISC_PLATFORM_HAVESCOPEID
1023		((struct sockaddr_in6 *)&itf->sin)->sin6_scope_id = isc_netaddr_getzone(&isc_if->address);
1024		itf->scopeid = isc_netaddr_getzone(&isc_if->address);
1025#endif
1026		itf->mask.ss_family = itf->sin.ss_family;
1027		memcpy(&(((struct sockaddr_in6 *)&itf->mask)->sin6_addr),
1028		       &(isc_if->netmask.type.in6),
1029		       sizeof(struct in6_addr));
1030		((struct sockaddr_in6 *)&itf->mask)->sin6_port = port;
1031		/* Copy the interface index */
1032		itf->ifindex = isc_if->ifindex;
1033	}
1034#endif /* INCLUDE_IPV6_SUPPORT */
1035
1036
1037	/* Process the rest of the flags */
1038
1039	if((isc_if->flags & INTERFACE_F_UP) != 0)
1040		itf->flags |= INT_UP;
1041	if((isc_if->flags & INTERFACE_F_LOOPBACK) != 0)
1042		itf->flags |= INT_LOOPBACK;
1043	if((isc_if->flags & INTERFACE_F_POINTTOPOINT) != 0)
1044		itf->flags |= INT_PPP;
1045	if((isc_if->flags & INTERFACE_F_MULTICAST) != 0)
1046		itf->flags |= INT_MULTICAST;
1047
1048}
1049
1050/*
1051 * refresh_interface
1052 *
1053 * some OSes have been observed to keep
1054 * cached routes even when more specific routes
1055 * become available.
1056 * this can be mitigated by re-binding
1057 * the socket.
1058 */
1059static int
1060refresh_interface(struct interface * interface)
1061{
1062#ifdef  OS_MISSES_SPECIFIC_ROUTE_UPDATES
1063	if (interface->fd != INVALID_SOCKET)
1064	{
1065		close_and_delete_fd_from_list(interface->fd);
1066		interface->fd = open_socket(&interface->sin,
1067					    interface->flags, 0, interface);
1068		 /*
1069		  * reset TTL indication so TTL is is set again
1070		  * next time around
1071		  */
1072		interface->last_ttl = 0;
1073		return interface->fd != INVALID_SOCKET;
1074	}
1075	else
1076	{
1077		return 0;	/* invalid sockets are not refreshable */
1078	}
1079#else /* !OS_MISSES_SPECIFIC_ROUTE_UPDATES */
1080	return interface->fd != INVALID_SOCKET;
1081#endif /* !OS_MISSES_SPECIFIC_ROUTE_UPDATES */
1082}
1083
1084/*
1085 * interface_update - externally callable update function
1086 */
1087void
1088interface_update(interface_receiver_t receiver, void *data)
1089{
1090	if (!disable_dynamic_updates) {
1091		int new_interface_found;
1092
1093		BLOCKIO();
1094		new_interface_found = update_interfaces(htons(NTP_PORT), receiver, data);
1095		UNBLOCKIO();
1096
1097		if (new_interface_found) {
1098#ifdef DEBUG
1099			msyslog(LOG_DEBUG, "new interface(s) found: waking up resolver");
1100#endif
1101#ifdef SYS_WINNT
1102			/* wake up the resolver thread */
1103			if (ResolverEventHandle != NULL)
1104				SetEvent(ResolverEventHandle);
1105#else
1106			/* write any single byte to the pipe to wake up the resolver process */
1107			write( resolver_pipe_fd[1], &new_interface_found, 1 );
1108#endif
1109		}
1110	}
1111}
1112
1113/*
1114 * find out if a given interface structure contains
1115 * a wildcard address
1116 */
1117static int
1118is_wildcard_addr(struct sockaddr_storage *sas)
1119{
1120	if (sas->ss_family == AF_INET &&
1121	    ((struct sockaddr_in*)sas)->sin_addr.s_addr == htonl(INADDR_ANY))
1122		return 1;
1123
1124#ifdef INCLUDE_IPV6_SUPPORT
1125	if (sas->ss_family == AF_INET6 &&
1126	    memcmp(&((struct sockaddr_in6*)sas)->sin6_addr, &in6addr_any,
1127		   sizeof(in6addr_any)) == 0)
1128		return 1;
1129#endif
1130
1131	return 0;
1132}
1133
1134#ifdef OS_NEEDS_REUSEADDR_FOR_IFADDRBIND
1135/*
1136 * enable/disable re-use of wildcard address socket
1137 */
1138static void
1139set_wildcard_reuse(int family, int on)
1140{
1141	int onvalue = 1;
1142	int offvalue = 0;
1143	int *onoff;
1144	SOCKET fd = INVALID_SOCKET;
1145
1146	onoff = on ? &onvalue : &offvalue;
1147
1148	switch (family) {
1149	case AF_INET:
1150		if (any_interface) {
1151			fd = any_interface->fd;
1152		}
1153		break;
1154
1155#ifdef INCLUDE_IPV6_SUPPORT
1156	case AF_INET6:
1157		if (any6_interface) {
1158			fd = any6_interface->fd;
1159		}
1160		break;
1161#endif /* !INCLUDE_IPV6_SUPPORT */
1162	}
1163
1164	if (fd != INVALID_SOCKET) {
1165		if (setsockopt(fd, SOL_SOCKET,
1166			       SO_REUSEADDR, (char *)onoff,
1167			       sizeof(*onoff))) {
1168			netsyslog(LOG_ERR, "set_wildcard_reuse: setsockopt(SO_REUSEADDR, %s) failed: %m", *onoff ? "on" : "off");
1169		}
1170		DPRINTF(4, ("set SO_REUSEADDR to %s on %s\n", *onoff ? "ON" : "OFF",
1171			    stoa((family == AF_INET) ?
1172				  &any_interface->sin : &any6_interface->sin)));
1173	}
1174}
1175#endif /* OS_NEEDS_REUSEADDR_FOR_IFADDRBIND */
1176
1177#ifdef INCLUDE_IPV6_SUPPORT
1178static isc_boolean_t
1179is_anycast(struct sockaddr *sa, char *name)
1180{
1181#if defined(SIOCGIFAFLAG_IN6) && defined(IN6_IFF_ANYCAST)
1182	struct in6_ifreq ifr6;
1183	int fd;
1184	u_int32_t flags6;
1185
1186	if (sa->sa_family != AF_INET6)
1187		return ISC_FALSE;
1188	if ((fd = socket(AF_INET6, SOCK_DGRAM, 0)) < 0)
1189		return ISC_FALSE;
1190	memset(&ifr6, 0, sizeof(ifr6));
1191	memcpy(&ifr6.ifr_addr, (struct sockaddr_in6 *)sa,
1192	    sizeof(struct sockaddr_in6));
1193	strlcpy(ifr6.ifr_name, name, IF_NAMESIZE);
1194	if (ioctl(fd, SIOCGIFAFLAG_IN6, &ifr6) < 0) {
1195		close(fd);
1196		return ISC_FALSE;
1197	}
1198	close(fd);
1199	flags6 = ifr6.ifr_ifru.ifru_flags6;
1200	if ((flags6 & IN6_IFF_ANYCAST) != 0)
1201		return ISC_TRUE;
1202#endif /* !SIOCGIFAFLAG_IN6 || !IN6_IFF_ANYCAST */
1203	return ISC_FALSE;
1204}
1205#endif /* !INCLUDE_IPV6_SUPPORT */
1206
1207/*
1208 * update_interface strategy
1209 *
1210 * toggle configuration phase
1211 *
1212 * Phase 1:
1213 * forall currently existing interfaces
1214 *   if address is known:
1215 *       drop socket - rebind again
1216 *
1217 *   if address is NOT known:
1218 *     attempt to create a new interface entry
1219 *
1220 * Phase 2:
1221 * forall currently known non MCAST and WILDCARD interfaces
1222 *   if interface does not match configuration phase (not seen in phase 1):
1223 *     remove interface from known interface list
1224 *     forall peers associated with this interface
1225 *       disconnect peer from this interface
1226 *
1227 * Phase 3:
1228 *   attempt to re-assign interfaces to peers
1229 *
1230 */
1231
1232static int
1233update_interfaces(
1234	u_short port,
1235	interface_receiver_t receiver,
1236	void *data
1237	)
1238{
1239	interface_info_t ifi;
1240	isc_mem_t *mctx = NULL;
1241	isc_interfaceiter_t *iter = NULL;
1242	isc_boolean_t scan_ipv4 = ISC_FALSE;
1243	isc_boolean_t scan_ipv6 = ISC_FALSE;
1244	isc_result_t result;
1245	int new_interface_found = 0;
1246
1247	DPRINTF(3, ("update_interfaces(%d)\n", ntohs( (u_short) port)));
1248
1249#ifdef INCLUDE_IPV6_SUPPORT
1250	if (isc_net_probeipv6() == ISC_R_SUCCESS)
1251		scan_ipv6 = ISC_TRUE;
1252#if defined(DEBUG)
1253	else
1254		if (debug)
1255			netsyslog(LOG_ERR, "no IPv6 interfaces found");
1256#endif
1257#endif
1258	if (isc_net_probeipv6() == ISC_R_SUCCESS)
1259		scan_ipv6 = ISC_TRUE;
1260#if defined(ISC_PLATFORM_HAVEIPV6) && defined(DEBUG)
1261	else
1262		if (debug)
1263			netsyslog(LOG_ERR, "no IPv6 interfaces found");
1264#endif
1265
1266	if (isc_net_probeipv4() == ISC_R_SUCCESS)
1267		scan_ipv4 = ISC_TRUE;
1268#ifdef DEBUG
1269	else
1270		if(debug)
1271			netsyslog(LOG_ERR, "no IPv4 interfaces found");
1272#endif
1273	/*
1274	 * phase one - scan interfaces
1275	 * - create those that are not found
1276	 * - update those that are found
1277	 */
1278
1279	result = isc_interfaceiter_create(mctx, &iter);
1280
1281	if (result != ISC_R_SUCCESS)
1282		return 0;
1283
1284	sys_interphase ^= 0x1;	/* toggle system phase for finding untouched (to be deleted) interfaces */
1285
1286	for (result = isc_interfaceiter_first(iter);
1287	     result == ISC_R_SUCCESS;
1288	     result = isc_interfaceiter_next(iter))
1289	{
1290		isc_interface_t isc_if;
1291		unsigned int family;
1292		struct interface interface;
1293		struct interface *iface;
1294
1295		result = isc_interfaceiter_current(iter, &isc_if);
1296
1297		if (result != ISC_R_SUCCESS)
1298			break;
1299
1300		/* See if we have a valid family to use */
1301		family = isc_if.address.family;
1302		if (family != AF_INET && family != AF_INET6)
1303			continue;
1304		if (scan_ipv4 == ISC_FALSE && family == AF_INET)
1305			continue;
1306		if (scan_ipv6 == ISC_FALSE && family == AF_INET6)
1307			continue;
1308
1309		/*
1310		 * create prototype
1311		 */
1312		init_interface(&interface);
1313
1314		convert_isc_if(&isc_if, &interface, port);
1315
1316		/*
1317		 * Check to see if we are going to use the interface
1318		 * If we don't use it we mark it to drop any packet
1319		 * received but we still must create the socket and
1320		 * bind to it. This prevents other apps binding to it
1321		 * and potentially causing problems with more than one
1322		 * process fiddling with the clock
1323		 */
1324		if (address_okay(&interface) == ISC_TRUE) {
1325			interface.ignore_packets = ISC_FALSE;
1326		}
1327		else {
1328			interface.ignore_packets = ISC_TRUE;
1329		}
1330
1331		DPRINT_INTERFACE(4, (&interface, "examining ", "\n"));
1332
1333		if (!(interface.flags & INT_UP))  { /* interfaces must be UP to be usable */
1334			DPRINTF(4, ("skipping interface %s (%s) - DOWN\n", interface.name, stoa(&interface.sin)));
1335			continue;
1336		}
1337
1338		/*
1339		 * skip any interfaces UP and bound to a wildcard
1340		 * address - some dhcp clients produce that in the
1341		 * wild
1342		 */
1343		if (is_wildcard_addr(&interface.sin))
1344			continue;
1345
1346#ifdef INCLUDE_IPV6_SUPPORT
1347		if (is_anycast((struct sockaddr *)&interface.sin, isc_if.name))
1348			continue;
1349#endif /* !INCLUDE_IPV6_SUPPORT */
1350
1351		/*
1352		 * map to local *address* in order
1353		 * to map all duplicate interfaces to an interface structure
1354		 * with the appropriate socket (our name space is
1355		 * (ip-address) - NOT (interface name, ip-address))
1356		 */
1357		iface = getinterface(&interface.sin, INT_WILDCARD);
1358
1359		if (iface && refresh_interface(iface))
1360		{
1361			/*
1362			 * found existing and up to date interface - mark present
1363			 */
1364
1365			iface->phase = sys_interphase;
1366			DPRINT_INTERFACE(4, (iface, "updating ", " present\n"));
1367			ifi.action = IFS_EXISTS;
1368			ifi.interface = iface;
1369			if (receiver)
1370				receiver(data, &ifi);
1371		}
1372		else
1373		{
1374			/*
1375			 * this is new or refreshing failed - add to our interface list
1376			 * if refreshing failed we will delete the interface structure in
1377			 * phase 2 as the interface was not marked current. We can bind to
1378			 * the address as the refresh code already closed the offending socket
1379			 */
1380
1381			iface = create_interface(port, &interface);
1382
1383			if (iface)
1384			{
1385				ifi.action = IFS_CREATED;
1386				ifi.interface = iface;
1387				if (receiver)
1388					receiver(data, &ifi);
1389
1390				new_interface_found = 1;
1391
1392				DPRINT_INTERFACE(3, (iface, "updating ", " new - created\n"));
1393			}
1394			else
1395			{
1396				DPRINT_INTERFACE(3, (&interface, "updating ", " new - creation FAILED"));
1397
1398				msyslog(LOG_INFO, "failed to initialize interface for address %s", stoa(&interface.sin));
1399				continue;
1400			}
1401		}
1402	}
1403
1404	isc_interfaceiter_destroy(&iter);
1405
1406	/*
1407	 * phase 2 - delete gone interfaces - reassigning peers to other interfaces
1408	 */
1409	{
1410		struct interface *interf = ISC_LIST_HEAD(inter_list);
1411
1412		while (interf != NULL)
1413		{
1414			struct interface *next = ISC_LIST_NEXT(interf, link);
1415
1416			if (!(interf->flags & (INT_WILDCARD|INT_MCASTIF))) {
1417				/*
1418				 * if phase does not match sys_phase this interface was not
1419				 * enumerated during interface scan - so it is gone and
1420				 * will be deleted here unless it is solely an MCAST/WILDCARD interface
1421				 */
1422				if (interf->phase != sys_interphase) {
1423					struct peer *peer;
1424					DPRINT_INTERFACE(3, (interf, "updating ", "GONE - deleting\n"));
1425					remove_interface(interf);
1426
1427					ifi.action = IFS_DELETED;
1428					ifi.interface = interf;
1429					if (receiver)
1430						receiver(data, &ifi);
1431
1432					peer = ISC_LIST_HEAD(interf->peers);
1433					/*
1434					 * disconnect peer from deleted interface
1435					 */
1436					while (peer != NULL) {
1437						struct peer *npeer = ISC_LIST_NEXT(peer, ilink);
1438
1439						/*
1440						 * this one just lost it's interface
1441						 */
1442						set_peerdstadr(peer, NULL);
1443
1444						peer = npeer;
1445					}
1446
1447					/*
1448					 * update globals in case we lose
1449					 * a loopback interface
1450					 */
1451					if (interf == loopback_interface)
1452						loopback_interface = NULL;
1453
1454					delete_interface(interf);
1455				}
1456			}
1457			interf = next;
1458		}
1459	}
1460
1461	/*
1462	 * phase 3 - re-configure as the world has changed if necessary
1463	 */
1464	refresh_all_peerinterfaces();
1465	return new_interface_found;
1466}
1467
1468
1469/*
1470 * create_sockets - create a socket for each interface plus a default
1471 *			socket for when we don't know where to send
1472 */
1473static int
1474create_sockets(
1475	u_short port
1476	)
1477{
1478#ifndef HAVE_IO_COMPLETION_PORT
1479	/*
1480	 * I/O Completion Ports don't care about the select and FD_SET
1481	 */
1482	maxactivefd = 0;
1483	FD_ZERO(&activefds);
1484#endif
1485
1486	DPRINTF(2, ("create_sockets(%d)\n", ntohs( (u_short) port)));
1487
1488	create_wildcards(port);
1489
1490	update_interfaces(port, NULL, NULL);
1491
1492	/*
1493	 * Now that we have opened all the sockets, turn off the reuse
1494	 * flag for security.
1495	 */
1496	set_reuseaddr(0);
1497
1498	DPRINTF(2, ("create_sockets: Total interfaces = %d\n", ninterfaces));
1499
1500	return ninterfaces;
1501}
1502
1503/*
1504 * create_interface - create a new interface for a given prototype
1505 *		      binding the socket.
1506 */
1507static struct interface *
1508create_interface(
1509		 u_short port,
1510		 struct interface *iface
1511		 )
1512{
1513	struct sockaddr_storage resmask;
1514	struct interface *interface;
1515
1516	DPRINTF(2, ("create_interface(%s#%d)\n", stoa(&iface->sin), ntohs( (u_short) port)));
1517
1518	/* build an interface */
1519	interface = new_interface(iface);
1520
1521	/*
1522	 * create socket
1523	 */
1524	interface->fd = open_socket(&interface->sin,
1525				 interface->flags, 0, interface);
1526
1527	if (interface->fd != INVALID_SOCKET)
1528		list_if_listening(interface, port);
1529
1530	if ((interface->flags & INT_BROADCAST) &&
1531	    interface->bfd != INVALID_SOCKET)
1532	  msyslog(LOG_INFO, "Listening on broadcast address %s#%d",
1533		  stoa((&interface->bcast)),
1534		  ntohs( (u_short) port));
1535
1536	if (interface->fd == INVALID_SOCKET &&
1537	    interface->bfd == INVALID_SOCKET) {
1538		msyslog(LOG_ERR, "unable to create socket on %s (%d) for %s#%d",
1539			interface->name,
1540			interface->ifnum,
1541			stoa((&interface->sin)),
1542			ntohs( (u_short) port));
1543		delete_interface(interface);
1544		return NULL;
1545	}
1546
1547        /*
1548	 * Blacklist bound interface address
1549	 */
1550
1551	SET_HOSTMASK(&resmask, interface->sin.ss_family);
1552	hack_restrict(RESTRICT_FLAGS, &interface->sin, &resmask,
1553		      RESM_NTPONLY|RESM_INTERFACE, RES_IGNORE);
1554
1555	/*
1556	 * set globals with the first found
1557	 * loopback interface of the appropriate class
1558	 */
1559	if ((loopback_interface == NULL) &&
1560	    (interface->family == AF_INET) &&
1561	    ((interface->flags & INT_LOOPBACK) != 0))
1562	{
1563		loopback_interface = interface;
1564	}
1565
1566	/*
1567	 * put into our interface list
1568	 */
1569	add_addr_to_list(&interface->sin, interface);
1570	add_interface(interface);
1571
1572	DPRINT_INTERFACE(2, (interface, "created ", "\n"));
1573	return interface;
1574}
1575
1576
1577#ifdef SO_EXCLUSIVEADDRUSE
1578static void
1579set_excladdruse(int fd)
1580{
1581	int one = 1;
1582	int failed;
1583
1584	failed = setsockopt(fd, SOL_SOCKET, SO_EXCLUSIVEADDRUSE,
1585			    (char *)&one, sizeof(one));
1586
1587	if (failed)
1588		netsyslog(LOG_ERR,
1589			  "setsockopt(%d, SO_EXCLUSIVEADDRUSE, on): %m", fd);
1590}
1591#endif  /* SO_EXCLUSIVEADDRUSE */
1592
1593
1594/*
1595 * set_reuseaddr() - set/clear REUSEADDR on all sockets
1596 *			NB possible hole - should we be doing this on broadcast
1597 *			fd's also?
1598 */
1599static void
1600set_reuseaddr(int flag) {
1601	struct interface *interf;
1602
1603#ifndef SO_EXCLUSIVEADDRUSE
1604
1605	for (interf = ISC_LIST_HEAD(inter_list);
1606	     interf != NULL;
1607	     interf = ISC_LIST_NEXT(interf, link)) {
1608
1609		if (interf->flags & INT_WILDCARD)
1610			continue;
1611
1612		/*
1613		 * if interf->fd  is INVALID_SOCKET, we might have a adapter
1614		 * configured but not present
1615		 */
1616		DPRINTF(4, ("setting SO_REUSEADDR on %.16s@%s to %s\n", interf->name, stoa(&interf->sin), flag ? "on" : "off"));
1617
1618		if (interf->fd != INVALID_SOCKET) {
1619			if (setsockopt(interf->fd, SOL_SOCKET,
1620					SO_REUSEADDR, (char *)&flag,
1621					sizeof(flag))) {
1622				netsyslog(LOG_ERR, "set_reuseaddr: setsockopt(SO_REUSEADDR, %s) failed: %m", flag ? "on" : "off");
1623			}
1624		}
1625	}
1626#endif /* ! SO_EXCLUSIVEADDRUSE */
1627}
1628
1629/*
1630 * This is just a wrapper around an internal function so we can
1631 * make other changes as necessary later on
1632 */
1633void
1634enable_broadcast(struct interface *iface, struct sockaddr_storage *baddr)
1635{
1636#ifdef SO_BROADCAST
1637	socket_broadcast_enable(iface, iface->fd, baddr);
1638#endif
1639}
1640
1641#ifdef OPEN_BCAST_SOCKET
1642/*
1643 * Enable a broadcast address to a given socket
1644 * The socket is in the inter_list all we need to do is enable
1645 * broadcasting. It is not this function's job to select the socket
1646 */
1647static isc_boolean_t
1648socket_broadcast_enable(struct interface *iface, SOCKET fd, struct sockaddr_storage *maddr)
1649{
1650#ifdef SO_BROADCAST
1651	int on = 1;
1652
1653	if (maddr->ss_family == AF_INET)
1654	{
1655		/* if this interface can support broadcast, set SO_BROADCAST */
1656		if (setsockopt(fd, SOL_SOCKET, SO_BROADCAST,
1657			       (char *)&on, sizeof(on)))
1658		{
1659			netsyslog(LOG_ERR, "setsockopt(SO_BROADCAST) enable failure on address %s: %m",
1660				stoa(maddr));
1661		}
1662#ifdef DEBUG
1663		else if (debug > 1) {
1664			printf("Broadcast enabled on socket %d for address %s\n",
1665				fd, stoa(maddr));
1666		}
1667#endif
1668	}
1669	iface->flags |= INT_BCASTOPEN;
1670	return ISC_TRUE;
1671#else
1672	return ISC_FALSE;
1673#endif /* SO_BROADCAST */
1674}
1675
1676/*
1677 * Remove a broadcast address from a given socket
1678 * The socket is in the inter_list all we need to do is disable
1679 * broadcasting. It is not this function's job to select the socket
1680 */
1681static isc_boolean_t
1682socket_broadcast_disable(struct interface *iface, struct sockaddr_storage *maddr)
1683{
1684#ifdef SO_BROADCAST
1685	int off = 0;	/* This seems to be OK as an int */
1686
1687	if (maddr->ss_family == AF_INET)
1688	{
1689		if (setsockopt(iface->fd, SOL_SOCKET, SO_BROADCAST,
1690			       (char *)&off, sizeof(off)))
1691		{
1692			netsyslog(LOG_ERR, "setsockopt(SO_BROADCAST) disable failure on address %s: %m",
1693				stoa(maddr));
1694		}
1695	}
1696	iface->flags &= ~INT_BCASTOPEN;
1697	return ISC_TRUE;
1698#else
1699	return ISC_FALSE;
1700#endif /* SO_BROADCAST */
1701}
1702
1703#endif /* OPEN_BCAST_SOCKET */
1704/*
1705 * Check to see if the address is a multicast address
1706 */
1707static isc_boolean_t
1708addr_ismulticast(struct sockaddr_storage *maddr)
1709{
1710	switch (maddr->ss_family)
1711	{
1712	case AF_INET :
1713		if (!IN_CLASSD(ntohl(((struct sockaddr_in*)maddr)->sin_addr.s_addr))) {
1714			DPRINTF(4, ("multicast address %s not class D\n", stoa(maddr)));
1715			return (ISC_FALSE);
1716		}
1717		else
1718		{
1719			return (ISC_TRUE);
1720		}
1721
1722	case AF_INET6 :
1723#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
1724		if (!IN6_IS_ADDR_MULTICAST(&((struct sockaddr_in6*)maddr)->sin6_addr)) {
1725			DPRINTF(4, ("address %s not IPv6 multicast address\n", stoa(maddr)));
1726			return (ISC_FALSE);
1727		}
1728		else
1729		{
1730			return (ISC_TRUE);
1731		}
1732
1733/*
1734 * If we don't have IPV6 support any IPV6 address is not multicast
1735 */
1736#else
1737		return (ISC_FALSE);
1738#endif
1739	/*
1740	 * Never valid
1741	 */
1742	default:
1743		return (ISC_FALSE);
1744	}
1745}
1746
1747/*
1748 * Multicast servers need to set the appropriate Multicast interface
1749 * socket option in order for it to know which interface to use for
1750 * send the multicast packet.
1751 */
1752void
1753enable_multicast_if(struct interface *iface, struct sockaddr_storage *maddr)
1754{
1755#ifdef MCAST
1756#ifdef IP_MULTICAST_LOOP
1757	/*u_char*/ TYPEOF_IP_MULTICAST_LOOP off = 0;
1758#endif
1759#ifdef IPV6_MULTICAST_LOOP
1760	u_int off6 = 0;		/* RFC 3493, 5.2. defines type unsigned int */
1761#endif
1762
1763	switch (maddr->ss_family)
1764	{
1765	case AF_INET:
1766		if (setsockopt(iface->fd, IPPROTO_IP, IP_MULTICAST_IF,
1767		   (char *)&(((struct sockaddr_in*)&iface->sin)->sin_addr.s_addr),
1768		    sizeof(struct in_addr)) == -1) {
1769			netsyslog(LOG_ERR,
1770			"setsockopt IP_MULTICAST_IF failure: %m on socket %d, addr %s for multicast address %s",
1771			iface->fd, stoa(&iface->sin), stoa(maddr));
1772			return;
1773		}
1774#ifdef IP_MULTICAST_LOOP
1775		/*
1776		 * Don't send back to itself, but allow it to fail to set it
1777		 */
1778		if (setsockopt(iface->fd, IPPROTO_IP, IP_MULTICAST_LOOP,
1779		       SETSOCKOPT_ARG_CAST &off, sizeof(off)) == -1) {
1780			netsyslog(LOG_ERR,
1781			"setsockopt IP_MULTICAST_LOOP failure: %m on socket %d, addr %s for multicast address %s",
1782			iface->fd, stoa(&iface->sin), stoa(maddr));
1783		}
1784#endif
1785	DPRINTF(4, ("Added IPv4 multicast interface on socket %d, addr %s for multicast address %s\n",
1786			    iface->fd, stoa(&iface->sin),
1787			    stoa(maddr)));
1788		break;
1789
1790	case AF_INET6:
1791#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
1792		if (setsockopt(iface->fd, IPPROTO_IPV6, IPV6_MULTICAST_IF,
1793		    (char *) &iface->scopeid, sizeof(iface->scopeid)) == -1) {
1794			netsyslog(LOG_ERR,
1795			"setsockopt IPV6_MULTICAST_IF failure: %m on socket %d, addr %s, scope %d for multicast address %s",
1796			iface->fd, stoa(&iface->sin), iface->scopeid,
1797			stoa(maddr));
1798			return;
1799		}
1800#ifdef IPV6_MULTICAST_LOOP
1801		/*
1802		 * Don't send back to itself, but allow it to fail to set it
1803		 */
1804		if (setsockopt(iface->fd, IPPROTO_IPV6, IPV6_MULTICAST_LOOP,
1805		       (char *) &off6, sizeof(off6)) == -1) {
1806			netsyslog(LOG_ERR,
1807			"setsockopt IPV6_MULTICAST_LOOP failure: %m on socket %d, addr %s for multicast address %s",
1808			iface->fd, stoa(&iface->sin), stoa(maddr));
1809		}
1810#endif
1811		DPRINTF(4, ("Added IPv6 multicast interface on socket %d, addr %s, scope %d for multicast address %s\n",
1812			    iface->fd,  stoa(&iface->sin), iface->scopeid,
1813			    stoa(maddr)));
1814		break;
1815#else
1816		return;
1817#endif	/* INCLUDE_IPV6_MULTICAST_SUPPORT */
1818	}
1819	return;
1820#endif
1821}
1822
1823/*
1824 * Add a multicast address to a given socket
1825 * The socket is in the inter_list all we need to do is enable
1826 * multicasting. It is not this function's job to select the socket
1827 */
1828static isc_boolean_t
1829socket_multicast_enable(struct interface *iface, int lscope, struct sockaddr_storage *maddr)
1830{
1831#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
1832	struct ipv6_mreq mreq6;
1833	struct in6_addr iaddr6;
1834#endif /* INCLUDE_IPV6_MULTICAST_SUPPORT */
1835
1836	struct ip_mreq mreq;
1837
1838	if (find_addr_in_list(maddr)) {
1839		DPRINTF(4, ("socket_multicast_enable(%s): already enabled\n", stoa(maddr)));
1840		return ISC_TRUE;
1841	}
1842
1843	switch (maddr->ss_family)
1844	{
1845	case AF_INET:
1846		memset((char *)&mreq, 0, sizeof(mreq));
1847		mreq.imr_multiaddr = (((struct sockaddr_in*)maddr)->sin_addr);
1848		mreq.imr_interface.s_addr = htonl(INADDR_ANY);
1849		if (setsockopt(iface->fd, IPPROTO_IP, IP_ADD_MEMBERSHIP,
1850			(char *)&mreq, sizeof(mreq)) == -1) {
1851			netsyslog(LOG_ERR,
1852			"setsockopt IP_ADD_MEMBERSHIP failure: %m on socket %d, addr %s for %x / %x (%s)",
1853			iface->fd, stoa(&iface->sin),
1854			mreq.imr_multiaddr.s_addr,
1855			mreq.imr_interface.s_addr, stoa(maddr));
1856			return ISC_FALSE;
1857		}
1858		DPRINTF(4, ("Added IPv4 multicast membership on socket %d, addr %s for %x / %x (%s)\n",
1859			    iface->fd, stoa(&iface->sin),
1860			    mreq.imr_multiaddr.s_addr,
1861			    mreq.imr_interface.s_addr, stoa(maddr)));
1862		break;
1863
1864	case AF_INET6:
1865#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
1866		/*
1867		 * Enable reception of multicast packets
1868		 * If the address is link-local we can get the interface index
1869		 * from the scope id. Don't do this for other types of multicast
1870		 * addresses. For now let the kernel figure it out.
1871		 */
1872		memset((char *)&mreq6, 0, sizeof(mreq6));
1873		iaddr6 = ((struct sockaddr_in6*)maddr)->sin6_addr;
1874		mreq6.ipv6mr_multiaddr = iaddr6;
1875		mreq6.ipv6mr_interface = lscope;
1876
1877		if (setsockopt(iface->fd, IPPROTO_IPV6, IPV6_JOIN_GROUP,
1878			(char *)&mreq6, sizeof(mreq6)) == -1) {
1879			netsyslog(LOG_ERR,
1880			 "setsockopt IPV6_JOIN_GROUP failure: %m on socket %d, addr %s for interface %d(%s)",
1881			iface->fd, stoa(&iface->sin),
1882			mreq6.ipv6mr_interface, stoa(maddr));
1883			return ISC_FALSE;
1884		}
1885		DPRINTF(4, ("Added IPv6 multicast group on socket %d, addr %s for interface %d(%s)\n",
1886			    iface->fd, stoa(&iface->sin),
1887			    mreq6.ipv6mr_interface, stoa(maddr)));
1888		break;
1889#else
1890		return ISC_FALSE;
1891#endif	/* INCLUDE_IPV6_MULTICAST_SUPPORT */
1892	}
1893	iface->flags |= INT_MCASTOPEN;
1894	iface->num_mcast++;
1895	add_addr_to_list(maddr, iface);
1896	return ISC_TRUE;
1897}
1898
1899/*
1900 * Remove a multicast address from a given socket
1901 * The socket is in the inter_list all we need to do is disable
1902 * multicasting. It is not this function's job to select the socket
1903 */
1904static isc_boolean_t
1905socket_multicast_disable(struct interface *iface, struct sockaddr_storage *maddr)
1906{
1907#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
1908	struct ipv6_mreq mreq6;
1909	struct in6_addr iaddr6;
1910#endif /* INCLUDE_IPV6_MULTICAST_SUPPORT */
1911
1912	struct ip_mreq mreq;
1913	memset((char *)&mreq, 0, sizeof(mreq));
1914
1915	if (find_addr_in_list(maddr) == NULL) {
1916		DPRINTF(4, ("socket_multicast_disable(%s): not enabled\n", stoa(maddr)));
1917		return ISC_TRUE;
1918	}
1919
1920	switch (maddr->ss_family)
1921	{
1922	case AF_INET:
1923		mreq.imr_multiaddr = (((struct sockaddr_in*)&maddr)->sin_addr);
1924		mreq.imr_interface.s_addr = ((struct sockaddr_in*)&iface->sin)->sin_addr.s_addr;
1925		if (setsockopt(iface->fd, IPPROTO_IP, IP_DROP_MEMBERSHIP,
1926			(char *)&mreq, sizeof(mreq)) == -1) {
1927			netsyslog(LOG_ERR,
1928			"setsockopt IP_DROP_MEMBERSHIP failure: %m on socket %d, addr %s for %x / %x (%s)",
1929			iface->fd, stoa(&iface->sin),
1930			mreq.imr_multiaddr.s_addr,
1931			mreq.imr_interface.s_addr, stoa(maddr));
1932			return ISC_FALSE;
1933		}
1934		break;
1935	case AF_INET6:
1936#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
1937		/*
1938		 * Disable reception of multicast packets
1939		 * If the address is link-local we can get the interface index
1940		 * from the scope id. Don't do this for other types of multicast
1941		 * addresses. For now let the kernel figure it out.
1942		 */
1943		iaddr6 = ((struct sockaddr_in6*)&maddr)->sin6_addr;
1944		mreq6.ipv6mr_multiaddr = iaddr6;
1945		mreq6.ipv6mr_interface = iface->scopeid;
1946
1947		if (setsockopt(iface->fd, IPPROTO_IPV6, IPV6_LEAVE_GROUP,
1948			(char *)&mreq6, sizeof(mreq6)) == -1) {
1949			netsyslog(LOG_ERR,
1950			"setsockopt IPV6_LEAVE_GROUP failure: %m on socket %d, addr %s for %d(%s)",
1951			iface->fd, stoa(&iface->sin),
1952			mreq6.ipv6mr_interface, stoa(maddr));
1953			return ISC_FALSE;
1954		}
1955		break;
1956#else
1957		return ISC_FALSE;
1958#endif	/* INCLUDE_IPV6_MULTICAST_SUPPORT */
1959
1960	}
1961	iface->num_mcast--;
1962	if (iface->num_mcast <= 0) {
1963                iface->num_mcast = 0;
1964		iface->flags &= ~INT_MCASTOPEN;
1965	}
1966	return ISC_TRUE;
1967}
1968
1969/*
1970 * io_setbclient - open the broadcast client sockets
1971 */
1972void
1973io_setbclient(void)
1974{
1975#ifdef OPEN_BCAST_SOCKET
1976        struct interface *interf;
1977	int nif = 0;
1978	isc_boolean_t jstatus;
1979	SOCKET fd;
1980
1981	set_reuseaddr(1);
1982
1983	for (interf = ISC_LIST_HEAD(inter_list);
1984	     interf != NULL;
1985	     interf = ISC_LIST_NEXT(interf, link)) {
1986	        if (interf->flags & INT_WILDCARD)
1987		        continue;
1988
1989		/* use only allowed addresses */
1990		if (interf->ignore_packets == ISC_TRUE)
1991			continue;
1992		/* Only IPv4 addresses are valid for broadcast */
1993		if (interf->sin.ss_family != AF_INET)
1994			continue;
1995
1996		/* Is this a broadcast address? */
1997		if (!(interf->flags & INT_BROADCAST))
1998			continue;
1999
2000		/* Skip the loopback addresses */
2001		if (interf->flags & INT_LOOPBACK)
2002			continue;
2003
2004		/* Do we already have the broadcast address open? */
2005		if (interf->flags & INT_BCASTOPEN) {
2006		/* account for already open interfaces to aviod misleading warning below */
2007			nif++;
2008			continue;
2009		}
2010
2011		/*
2012		 * Try to open the broadcast address
2013		 */
2014		interf->family = AF_INET;
2015		interf->bfd = open_socket(&interf->bcast,
2016				    INT_BROADCAST, 0, interf);
2017
2018		 /*
2019		 * If we succeeded then we use it otherwise
2020		 * enable the underlying address
2021		 */
2022		if (interf->bfd == INVALID_SOCKET) {
2023			fd = interf->fd;
2024		}
2025		else {
2026			fd = interf->bfd;
2027		}
2028
2029		/* Enable Broadcast on socket */
2030		jstatus = socket_broadcast_enable(interf, fd, &interf->sin);
2031		if (jstatus == ISC_TRUE)
2032		{
2033			nif++;
2034			netsyslog(LOG_INFO,"io_setbclient: Opened broadcast client on interface #%d %s, socket: %d",
2035				  interf->ifnum, interf->name, fd);
2036			interf->addr_refid = addr2refid(&interf->sin);
2037		}
2038	}
2039	set_reuseaddr(0);
2040#ifdef DEBUG
2041	if (debug)
2042		if (nif > 0)
2043			printf("io_setbclient: Opened broadcast clients\n");
2044#endif
2045	if (nif == 0)
2046		netsyslog(LOG_ERR, "Unable to listen for broadcasts, no broadcast interfaces available");
2047#else
2048	netsyslog(LOG_ERR, "io_setbclient: Broadcast Client disabled by build");
2049#endif
2050}
2051
2052/*
2053 * io_unsetbclient - close the broadcast client sockets
2054 */
2055void
2056io_unsetbclient(void)
2057{
2058        struct interface *interf;
2059	isc_boolean_t lstatus;
2060
2061	for (interf = ISC_LIST_HEAD(inter_list);
2062	     interf != NULL;
2063	     interf = ISC_LIST_NEXT(interf, link))
2064	{
2065	        if (interf->flags & INT_WILDCARD)
2066		    continue;
2067
2068		if (!(interf->flags & INT_BCASTOPEN))
2069		    continue;
2070		lstatus = socket_broadcast_disable(interf, &interf->sin);
2071	}
2072}
2073
2074/*
2075 * io_multicast_add() - add multicast group address
2076 */
2077void
2078io_multicast_add(
2079	struct sockaddr_storage addr
2080	)
2081{
2082#ifdef MCAST
2083	struct interface *interface;
2084#ifndef MULTICAST_NONEWSOCKET
2085	struct interface *iface;
2086#endif
2087	int lscope = 0;
2088
2089	/*
2090	 * Check to see if this is a multicast address
2091	 */
2092	if (addr_ismulticast(&addr) == ISC_FALSE)
2093		return;
2094
2095	/* If we already have it we can just return */
2096	if (find_flagged_addr_in_list(&addr, INT_MCASTOPEN|INT_MCASTIF) != NULL)
2097	{
2098		netsyslog(LOG_INFO, "Duplicate request found for multicast address %s",
2099			stoa(&addr));
2100		return;
2101	}
2102
2103#ifndef MULTICAST_NONEWSOCKET
2104	interface = new_interface(NULL);
2105
2106	/*
2107	 * Open a new socket for the multicast address
2108	 */
2109	interface->sin.ss_family = addr.ss_family;
2110	interface->family = addr.ss_family;
2111
2112	switch(addr.ss_family) {
2113	case AF_INET:
2114		memcpy(&(((struct sockaddr_in *)&interface->sin)->sin_addr),
2115		       &(((struct sockaddr_in*)&addr)->sin_addr),
2116		       sizeof(struct in_addr));
2117		((struct sockaddr_in*)&interface->sin)->sin_port = htons(NTP_PORT);
2118		memset(&((struct sockaddr_in*)&interface->mask)->sin_addr.s_addr, 0xff, sizeof(struct in_addr));
2119		break;
2120	case AF_INET6:
2121#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
2122		memcpy(&(((struct sockaddr_in6 *)&interface->sin)->sin6_addr),
2123		       &((struct sockaddr_in6*)&addr)->sin6_addr,
2124		       sizeof(struct in6_addr));
2125		((struct sockaddr_in6*)&interface->sin)->sin6_port = htons(NTP_PORT);
2126#ifdef ISC_PLATFORM_HAVESCOPEID
2127		((struct sockaddr_in6*)&interface->sin)->sin6_scope_id = ((struct sockaddr_in6*)&addr)->sin6_scope_id;
2128#endif
2129		memset(&((struct sockaddr_in6*)&interface->mask)->sin6_addr.s6_addr, 0xff, sizeof(struct in6_addr));
2130#endif
2131		iface = findlocalcastinterface(&addr, INT_MULTICAST);
2132		if (iface) {
2133# ifdef ISC_PLATFORM_HAVESCOPEID
2134			lscope = ((struct sockaddr_in6*)&iface->sin)->sin6_scope_id;
2135# endif
2136			DPRINTF(4, ("Found interface #%d %s, scope: %d for address %s\n", iface->ifnum, iface->name, lscope, stoa(&addr)));
2137		}
2138		break;
2139	}
2140
2141	set_reuseaddr(1);
2142	interface->bfd = INVALID_SOCKET;
2143	interface->fd = open_socket(&interface->sin,
2144			    INT_MULTICAST, 0, interface);
2145
2146	if (interface->fd != INVALID_SOCKET)
2147	{
2148		interface->bfd = INVALID_SOCKET;
2149		interface->ignore_packets = ISC_FALSE;
2150		interface->flags |= INT_MCASTIF;
2151
2152		(void) strncpy(interface->name, "multicast",
2153			sizeof(interface->name));
2154		((struct sockaddr_in*)&interface->mask)->sin_addr.s_addr =
2155						htonl(~(u_int32)0);
2156		DPRINT_INTERFACE(2, (interface, "multicast add ", "\n"));
2157		/* socket_multicast_enable() will add this address to the addresslist */
2158		add_interface(interface);
2159		list_if_listening(interface, htons(NTP_PORT));
2160	}
2161	else
2162	{
2163		delete_interface(interface);  /* re-use existing interface */
2164		interface = NULL;
2165		if (addr.ss_family == AF_INET)
2166			interface = wildipv4;
2167		else if (addr.ss_family == AF_INET6)
2168			interface = wildipv6;
2169
2170		if (interface != NULL) {
2171			/* HACK ! -- stuff in an address */
2172			interface->bcast = addr;
2173			netsyslog(LOG_ERR,
2174			 "...multicast address %s using wildcard interface #%d %s",
2175				  stoa(&addr), interface->ifnum, interface->name);
2176		} else {
2177			netsyslog(LOG_ERR,
2178			"No multicast socket available to use for address %s",
2179			stoa(&addr));
2180			return;
2181		}
2182	}
2183#else
2184	/*
2185	 * For the case where we can't use a separate socket
2186	 */
2187	interface = findlocalcastinterface(&addr, INT_MULTICAST);
2188	/*
2189	 * If we don't have a valid socket, just return
2190	 */
2191	if (!interface)
2192	{
2193		netsyslog(LOG_ERR,
2194		"Cannot add multicast address %s: Cannot find slot",
2195		stoa(&addr));
2196		return;
2197	}
2198
2199#endif
2200	{
2201		isc_boolean_t jstatus;
2202		jstatus = socket_multicast_enable(interface, lscope, &addr);
2203
2204		if (jstatus == ISC_TRUE)
2205			netsyslog(LOG_INFO, "Added Multicast Listener %s on interface #%d %s\n", stoa(&addr), interface->ifnum, interface->name);
2206		else
2207			netsyslog(LOG_ERR, "Failed to add Multicast Listener %s\n", stoa(&addr));
2208	}
2209#else /* MCAST */
2210	netsyslog(LOG_ERR,
2211		  "Cannot add multicast address %s: no Multicast support",
2212		  stoa(&addr));
2213#endif /* MCAST */
2214	return;
2215}
2216
2217/*
2218 * io_multicast_del() - delete multicast group address
2219 */
2220void
2221io_multicast_del(
2222	struct sockaddr_storage addr
2223	)
2224{
2225#ifdef MCAST
2226        struct interface *interface;
2227	isc_boolean_t lstatus;
2228
2229	/*
2230	 * Check to see if this is a multicast address
2231	 */
2232	if (addr_ismulticast(&addr) == ISC_FALSE)
2233	{
2234		netsyslog(LOG_ERR,
2235			 "invalid multicast address %s", stoa(&addr));
2236		return;
2237	}
2238
2239	switch (addr.ss_family)
2240	{
2241	case AF_INET :
2242		/*
2243		 * Disable reception of multicast packets
2244		 */
2245		interface = find_flagged_addr_in_list(&addr, INT_MCASTOPEN);
2246		while ( interface != NULL) {
2247			lstatus = socket_multicast_disable(interface, &addr);
2248			interface = find_flagged_addr_in_list(&addr, INT_MCASTOPEN);
2249		}
2250		break;
2251
2252#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
2253	case AF_INET6 :
2254		/*
2255		 * Disable reception of multicast packets
2256		 */
2257		for (interface = ISC_LIST_HEAD(inter_list);
2258		     interface != NULL;
2259		     interface = ISC_LIST_NEXT(interface, link))
2260		{
2261                        if (interface->flags & INT_WILDCARD)
2262			        continue;
2263
2264			/* Be sure it's the correct family */
2265			if (interface->sin.ss_family != AF_INET6)
2266				continue;
2267			if (!(interface->flags & INT_MCASTOPEN))
2268				continue;
2269			if (!(interface->fd < 0))
2270				continue;
2271			if (!SOCKCMP(&addr, &interface->sin))
2272				continue;
2273			lstatus = socket_multicast_disable(interface, &addr);
2274		}
2275		break;
2276#endif /* INCLUDE_IPV6_MULTICAST_SUPPORT */
2277
2278	}/* switch */
2279
2280        delete_addr_from_list(&addr);
2281
2282#else /* not MCAST */
2283	netsyslog(LOG_ERR, "this function requires multicast kernel");
2284#endif /* not MCAST */
2285}
2286
2287/*
2288 * init_nonblocking_io() - set up descriptor to be non blocking
2289 */
2290static void init_nonblocking_io(SOCKET fd)
2291{
2292	/*
2293	 * set non-blocking,
2294	 */
2295
2296#ifdef USE_FIONBIO
2297	/* in vxWorks we use FIONBIO, but the others are defined for old systems, so
2298	 * all hell breaks loose if we leave them defined
2299	 */
2300#undef O_NONBLOCK
2301#undef FNDELAY
2302#undef O_NDELAY
2303#endif
2304
2305#if defined(O_NONBLOCK) /* POSIX */
2306	if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0)
2307	{
2308		netsyslog(LOG_ERR, "fcntl(O_NONBLOCK) fails on fd #%d: %m",
2309			fd);
2310		exit(1);
2311		/*NOTREACHED*/
2312	}
2313#elif defined(FNDELAY)
2314	if (fcntl(fd, F_SETFL, FNDELAY) < 0)
2315	{
2316		netsyslog(LOG_ERR, "fcntl(FNDELAY) fails on fd #%d: %m",
2317			fd);
2318		exit(1);
2319		/*NOTREACHED*/
2320	}
2321#elif defined(O_NDELAY) /* generally the same as FNDELAY */
2322	if (fcntl(fd, F_SETFL, O_NDELAY) < 0)
2323	{
2324		netsyslog(LOG_ERR, "fcntl(O_NDELAY) fails on fd #%d: %m",
2325			fd);
2326		exit(1);
2327		/*NOTREACHED*/
2328	}
2329#elif defined(FIONBIO)
2330	{
2331		int on = 1;
2332		if (ioctl(fd,FIONBIO,&on) < 0)
2333		{
2334			netsyslog(LOG_ERR, "ioctl(FIONBIO) fails on fd #%d: %m",
2335				fd);
2336			exit(1);
2337			/*NOTREACHED*/
2338		}
2339	}
2340#elif defined(FIOSNBIO)
2341	if (ioctl(fd,FIOSNBIO,&on) < 0)
2342	{
2343		netsyslog(LOG_ERR, "ioctl(FIOSNBIO) fails on fd #%d: %m",
2344			fd);
2345		exit(1);
2346		/*NOTREACHED*/
2347	}
2348#else
2349# include "Bletch: Need non-blocking I/O!"
2350#endif
2351}
2352
2353/*
2354 * open_socket - open a socket, returning the file descriptor
2355 */
2356
2357static SOCKET
2358open_socket(
2359	struct sockaddr_storage *addr,
2360	int flags,
2361	int turn_off_reuse,
2362	struct interface *interf
2363	)
2364{
2365	int errval;
2366	SOCKET fd;
2367	/*
2368	 * int is OK for REUSEADR per
2369	 * http://www.kohala.com/start/mcast.api.txt
2370	 */
2371	int on = 1;
2372	int off = 0;
2373
2374#if defined(IPTOS_LOWDELAY) && defined(IPPROTO_IP) && defined(IP_TOS)
2375	int tos;
2376#endif /* IPTOS_LOWDELAY && IPPROTO_IP && IP_TOS */
2377
2378	if ((addr->ss_family == AF_INET6) && (isc_net_probeipv6() != ISC_R_SUCCESS))
2379		return (INVALID_SOCKET);
2380
2381	/* create a datagram (UDP) socket */
2382	fd = socket(addr->ss_family, SOCK_DGRAM, 0);
2383	if (INVALID_SOCKET == fd) {
2384#ifndef SYS_WINNT
2385		errval = errno;
2386#else
2387		errval = WSAGetLastError();
2388#endif
2389		netsyslog(LOG_ERR,
2390			  "socket(AF_INET%s, SOCK_DGRAM, 0) failed on address %s: %m",
2391			  (addr->ss_family == AF_INET6) ? "6" : "",
2392			  stoa(addr));
2393
2394		if (errval == EPROTONOSUPPORT ||
2395		    errval == EAFNOSUPPORT ||
2396		    errval == EPFNOSUPPORT)
2397			return (INVALID_SOCKET);
2398		msyslog(LOG_ERR, "unexpected error code %d (not PROTONOSUPPORT|AFNOSUPPORT|FPNOSUPPORT) - exiting", errval);
2399		exit(1);
2400		/*NOTREACHED*/
2401	}
2402
2403#ifdef SYS_WINNT
2404	connection_reset_fix(fd, addr);
2405#endif
2406	/*
2407	 * Fixup the file descriptor for some systems
2408	 * See bug #530 for details of the issue.
2409	 */
2410	fd = move_fd(fd);
2411
2412	/*
2413	 * set SO_REUSEADDR since we will be binding the same port
2414	 * number on each interface according to turn_off_reuse.
2415	 * This is undesirable on Windows versions starting with
2416	 * Windows XP (numeric version 5.1).
2417	 */
2418#ifdef SYS_WINNT
2419	if (isc_win32os_versioncheck(5, 1, 0, 0) < 0)  /* before 5.1 */
2420#endif
2421		if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR,
2422			       (char *)(turn_off_reuse
2423					? &off
2424					: &on),
2425			       sizeof(on))) {
2426
2427			netsyslog(LOG_ERR, "setsockopt SO_REUSEADDR %s"
2428					   " fails for address %s: %m",
2429					   turn_off_reuse
2430						? "off"
2431						: "on",
2432					   stoa(addr));
2433			closesocket(fd);
2434			return INVALID_SOCKET;
2435		}
2436#ifdef SO_EXCLUSIVEADDRUSE
2437	/*
2438	 * setting SO_EXCLUSIVEADDRUSE on the wildcard we open
2439	 * first will cause more specific binds to fail.
2440	 */
2441	if (!(interf->flags & INT_WILDCARD))
2442		set_excladdruse(fd);
2443#endif
2444
2445	/*
2446	 * IPv4 specific options go here
2447	 */
2448	if (addr->ss_family == AF_INET) {
2449#if defined(IPTOS_LOWDELAY) && defined(IPPROTO_IP) && defined(IP_TOS)
2450	/* set IP_TOS to minimize packet delay */
2451		tos = IPTOS_LOWDELAY;
2452		if (setsockopt(fd, IPPROTO_IP, IP_TOS, (char *) &tos, sizeof(tos)) < 0)
2453		{
2454			netsyslog(LOG_ERR, "setsockopt IPTOS_LOWDELAY on fails on address %s: %m",
2455				stoa(addr));
2456		}
2457#endif /* IPTOS_LOWDELAY && IPPROTO_IP && IP_TOS */
2458	}
2459
2460	/*
2461	 * IPv6 specific options go here
2462	 */
2463        if (addr->ss_family == AF_INET6) {
2464#if defined(IPV6_V6ONLY)
2465                if (setsockopt(fd, IPPROTO_IPV6, IPV6_V6ONLY,
2466                	(char*)&on, sizeof(on)))
2467                {
2468                	netsyslog(LOG_ERR, "setsockopt IPV6_V6ONLY on fails on address %s: %m",
2469				stoa(addr));
2470		}
2471#endif /* IPV6_V6ONLY */
2472#if defined(IPV6_BINDV6ONLY)
2473                if (setsockopt(fd, IPPROTO_IPV6, IPV6_BINDV6ONLY,
2474                	(char*)&on, sizeof(on)))
2475                {
2476                	netsyslog(LOG_ERR,
2477			    "setsockopt IPV6_BINDV6ONLY on fails on address %s: %m",
2478			    stoa(addr));
2479		}
2480#endif /* IPV6_BINDV6ONLY */
2481	}
2482
2483#ifdef OS_NEEDS_REUSEADDR_FOR_IFADDRBIND
2484	/*
2485	 * some OSes don't allow binding to more specific
2486	 * addresses if a wildcard address already bound
2487	 * to the port and SO_REUSEADDR is not set
2488	 */
2489	if (!is_wildcard_addr(addr)) {
2490		set_wildcard_reuse(addr->ss_family, 1);
2491	}
2492#endif
2493
2494	/*
2495	 * bind the local address.
2496	 */
2497	errval = bind(fd, (struct sockaddr *)addr, SOCKLEN(addr));
2498
2499#ifdef OS_NEEDS_REUSEADDR_FOR_IFADDRBIND
2500	/*
2501	 * some OSes don't allow binding to more specific
2502	 * addresses if a wildcard address already bound
2503	 * to the port and REUSE_ADDR is not set
2504	 */
2505	if (!is_wildcard_addr(addr)) {
2506		set_wildcard_reuse(addr->ss_family, 0);
2507	}
2508#endif
2509
2510	if (errval < 0) {
2511		/*
2512		 * Don't log this under all conditions
2513		 */
2514		if (turn_off_reuse == 0
2515#ifdef DEBUG
2516		    || debug > 1
2517#endif
2518			) {
2519			if (addr->ss_family == AF_INET)
2520				netsyslog(LOG_ERR,
2521					  "bind() fd %d, family AF_INET, port %d, addr %s, in_classd=%d flags=0x%x fails: %m",
2522					  fd, (int)ntohs(((struct sockaddr_in*)addr)->sin_port),
2523					  stoa(addr),
2524					  IN_CLASSD(ntohl(((struct sockaddr_in*)addr)->sin_addr.s_addr)),
2525					  flags);
2526#ifdef INCLUDE_IPV6_SUPPORT
2527			else if (addr->ss_family == AF_INET6)
2528				netsyslog(LOG_ERR,
2529					  "bind() fd %d, family AF_INET6, port %d, scope %d, addr %s, mcast=%d flags=0x%x fails: %m",
2530					  fd, (int)ntohs(((struct sockaddr_in6*)addr)->sin6_port),
2531# ifdef ISC_PLATFORM_HAVESCOPEID
2532					  ((struct sockaddr_in6*)addr)->sin6_scope_id
2533# else
2534					  -1
2535# endif
2536					  , stoa(addr),
2537					  IN6_IS_ADDR_MULTICAST(&((struct sockaddr_in6*)addr)->sin6_addr),
2538					  flags);
2539#endif
2540		}
2541
2542		closesocket(fd);
2543
2544		return INVALID_SOCKET;
2545	}
2546
2547#ifdef HAVE_TIMESTAMP
2548	{
2549		if (setsockopt(fd, SOL_SOCKET, SO_TIMESTAMP,
2550			       (char*)&on, sizeof(on)))
2551		{
2552			netsyslog(LOG_DEBUG,
2553				  "setsockopt SO_TIMESTAMP on fails on address %s: %m",
2554				  stoa(addr));
2555		}
2556#ifdef DEBUG
2557		else
2558		{
2559			DPRINTF(4, ("setsockopt SO_TIMESTAMP enabled on fd %d address %s\n", fd, stoa(addr)));
2560		}
2561#endif
2562	}
2563#endif
2564	DPRINTF(4, ("bind() fd %d, family %d, port %d, addr %s, flags=0x%x\n",
2565		   fd,
2566		   addr->ss_family,
2567		   (int)ntohs(((struct sockaddr_in*)addr)->sin_port),
2568		   stoa(addr),
2569		   interf->flags));
2570
2571	init_nonblocking_io(fd);
2572
2573#ifdef HAVE_SIGNALED_IO
2574	init_socket_sig(fd);
2575#endif /* not HAVE_SIGNALED_IO */
2576
2577	add_fd_to_list(fd, FD_TYPE_SOCKET);
2578
2579#if !defined(SYS_WINNT) && !defined(VMS)
2580	DPRINTF(4, ("flags for fd %d: 0x%x\n", fd,
2581		    fcntl(fd, F_GETFL, 0)));
2582#endif /* SYS_WINNT || VMS */
2583
2584#if defined (HAVE_IO_COMPLETION_PORT)
2585/*
2586 * Add the socket to the completion port
2587 */
2588	if (io_completion_port_add_socket(fd, interf))
2589	{
2590		msyslog(LOG_ERR, "unable to set up io completion port - EXITING");
2591		exit(1);
2592	}
2593#endif
2594	return fd;
2595}
2596
2597/* XXX ELIMINATE sendpkt similar in ntpq.c, ntpdc.c, ntp_io.c, ntptrace.c */
2598/*
2599 * sendpkt - send a packet to the specified destination. Maintain a
2600 * send error cache so that only the first consecutive error for a
2601 * destination is logged.
2602 */
2603void
2604sendpkt(
2605	struct sockaddr_storage *dest,
2606	struct interface *inter,
2607	int ttl,
2608	struct pkt *pkt,
2609	int len
2610	)
2611{
2612	int cc, slot;
2613
2614	/*
2615	 * Send error caches. Empty slots have port == 0
2616	 * Set ERRORCACHESIZE to 0 to disable
2617	 */
2618	struct cache {
2619		u_short port;
2620		struct	in_addr addr;
2621	};
2622
2623#ifdef INCLUDE_IPV6_SUPPORT
2624	struct cache6 {
2625		u_short port;
2626		struct in6_addr addr;
2627	};
2628#endif /* INCLUDE_IPV6_SUPPORT */
2629
2630
2631#ifndef ERRORCACHESIZE
2632#define ERRORCACHESIZE 8
2633#endif
2634#if ERRORCACHESIZE > 0
2635	static struct cache badaddrs[ERRORCACHESIZE];
2636#ifdef INCLUDE_IPV6_SUPPORT
2637	static struct cache6 badaddrs6[ERRORCACHESIZE];
2638#endif /* INCLUDE_IPV6_SUPPORT */
2639#else
2640#define badaddrs ((struct cache *)0)		/* Only used in empty loops! */
2641#ifdef INCLUDE_IPV6_SUPPORT
2642#define badaddrs6 ((struct cache6 *)0)		/* Only used in empty loops! */
2643#endif /* INCLUDE_IPV6_SUPPORT */
2644#endif
2645#ifdef DEBUG
2646	if (debug > 1)
2647	  {
2648	    if (inter != NULL)
2649	      {
2650		printf("%ssendpkt(fd=%d dst=%s, src=%s, ttl=%d, len=%d)\n",
2651		       (ttl > 0) ? "\tMCAST\t***** " : "",
2652		       inter->fd, stoa(dest),
2653		       stoa(&inter->sin), ttl, len);
2654	      }
2655	    else
2656	      {
2657		printf("%ssendpkt(dst=%s, ttl=%d, len=%d): no interface - IGNORED\n",
2658		       (ttl > 0) ? "\tMCAST\t***** " : "",
2659		       stoa(dest),
2660		       ttl, len);
2661	      }
2662	  }
2663#endif
2664
2665	if (inter == NULL)	/* unbound peer - drop request and wait for better network conditions */
2666	  return;
2667
2668#ifdef MCAST
2669
2670	/*
2671	 * for the moment we use the bcast option to set multicast ttl
2672	 */
2673	if (ttl > 0 && ttl != inter->last_ttl) {
2674
2675		/*
2676		 * set the multicast ttl for outgoing packets
2677		 */
2678		int rtc;
2679
2680		switch (inter->sin.ss_family) {
2681
2682		case AF_INET :
2683		{
2684			u_char mttl = (u_char) ttl;
2685
2686			rtc = setsockopt(inter->fd, IPPROTO_IP, IP_MULTICAST_TTL,
2687					 (const void *) &mttl, sizeof(mttl));
2688			break;
2689		}
2690
2691#ifdef INCLUDE_IPV6_SUPPORT
2692		case AF_INET6 :
2693		{
2694			u_int ittl = (u_char) ttl;
2695
2696			rtc = setsockopt(inter->fd, IPPROTO_IPV6, IPV6_MULTICAST_HOPS,
2697					 (const void *) &ittl, sizeof(ittl));
2698			break;
2699		}
2700
2701#endif /* INCLUDE_IPV6_SUPPORT */
2702		default:	/* just NOP if not supported */
2703			rtc = 0;
2704			break;
2705		}
2706
2707		if (rtc != 0) {
2708			netsyslog(LOG_ERR, "setsockopt IP_MULTICAST_TTL/IPV6_MULTICAST_HOPS fails on address %s: %m",
2709				  stoa(&inter->sin));
2710		}
2711		else
2712			inter->last_ttl = ttl;
2713	}
2714
2715#endif /* MCAST */
2716
2717	for (slot = ERRORCACHESIZE; --slot >= 0; )
2718		if(dest->ss_family == AF_INET) {
2719			if (badaddrs[slot].port == SRCPORT(dest) &&
2720				badaddrs[slot].addr.s_addr == ((struct sockaddr_in*)dest)->sin_addr.s_addr)
2721			break;
2722		}
2723#ifdef INCLUDE_IPV6_SUPPORT
2724		else if (dest->ss_family == AF_INET6) {
2725			if (badaddrs6[slot].port == SRCPORT(dest) &&
2726				!memcmp(&badaddrs6[slot].addr, &((struct sockaddr_in6*)dest)->sin6_addr, sizeof(struct in6_addr)))
2727			break;
2728		}
2729#endif /* INCLUDE_IPV6_SUPPORT */
2730
2731#if defined(HAVE_IO_COMPLETION_PORT)
2732        cc = io_completion_port_sendto(inter, pkt, len, dest);
2733	if (cc != ERROR_SUCCESS)
2734#else
2735#ifdef SIM
2736        cc = srvr_rply(&ntp_node,  dest, inter, pkt);
2737#else /* SIM */
2738	cc = sendto(inter->fd, (char *)pkt, (unsigned int)len, 0, (struct sockaddr *)dest,
2739		    SOCKLEN(dest));
2740#endif /* SIM */
2741	if (cc == -1)
2742#endif
2743	{
2744		inter->notsent++;
2745		packets_notsent++;
2746
2747#if defined(HAVE_IO_COMPLETION_PORT)
2748		if (cc != WSAEWOULDBLOCK && cc != WSAENOBUFS && slot < 0)
2749#else
2750		if (errno != EWOULDBLOCK && errno != ENOBUFS && slot < 0)
2751#endif
2752		{
2753			/*
2754			 * Remember this, if there's an empty slot
2755			 */
2756			switch (dest->ss_family) {
2757
2758			case AF_INET :
2759
2760				for (slot = ERRORCACHESIZE; --slot >= 0; )
2761					if (badaddrs[slot].port == 0)
2762					{
2763						badaddrs[slot].port = SRCPORT(dest);
2764						badaddrs[slot].addr = ((struct sockaddr_in*)dest)->sin_addr;
2765						break;
2766					}
2767				break;
2768
2769#ifdef INCLUDE_IPV6_SUPPORT
2770			case AF_INET6 :
2771
2772				for (slot = ERRORCACHESIZE; --slot >= 0; )
2773					if (badaddrs6[slot].port == 0)
2774					{
2775						badaddrs6[slot].port = SRCPORT(dest);
2776						badaddrs6[slot].addr = ((struct sockaddr_in6*)dest)->sin6_addr;
2777						break;
2778					}
2779				break;
2780#endif /* INCLUDE_IPV6_SUPPORT */
2781			default:  /* don't care if not supported */
2782				break;
2783			}
2784
2785			netsyslog(LOG_ERR, "sendto(%s) (fd=%d): %m",
2786				  stoa(dest), inter->fd);
2787		}
2788	}
2789	else
2790	{
2791		inter->sent++;
2792		packets_sent++;
2793		/*
2794		 * He's not bad any more
2795		 */
2796		if (slot >= 0)
2797		{
2798			netsyslog(LOG_INFO, "Connection re-established to %s", stoa(dest));
2799			switch (dest->ss_family) {
2800			case AF_INET :
2801				badaddrs[slot].port = 0;
2802				break;
2803#ifdef INCLUDE_IPV6_SUPPORT
2804			case AF_INET6 :
2805				badaddrs6[slot].port = 0;
2806				break;
2807#endif /* INCLUDE_IPV6_SUPPORT */
2808			default:  /* don't care if not supported */
2809				break;
2810			}
2811		}
2812	}
2813}
2814
2815#if !defined(HAVE_IO_COMPLETION_PORT)
2816/*
2817 * fdbits - generate ascii representation of fd_set (FAU debug support)
2818 * HFDF format - highest fd first.
2819 */
2820static char *
2821fdbits(
2822	int count,
2823	fd_set *set
2824	)
2825{
2826	static char buffer[256];
2827	char * buf = buffer;
2828
2829	count = (count < 256) ? count : 255;
2830
2831	while (count >= 0)
2832	{
2833		*buf++ = FD_ISSET(count, set) ? '#' : '-';
2834		count--;
2835	}
2836	*buf = '\0';
2837
2838	return buffer;
2839}
2840
2841/*
2842 * Routine to read the refclock packets for a specific interface
2843 * Return the number of bytes read. That way we know if we should
2844 * read it again or go on to the next one if no bytes returned
2845 */
2846static inline int
2847read_refclock_packet(SOCKET fd, struct refclockio *rp, l_fp ts)
2848{
2849	int i;
2850	int buflen;
2851	register struct recvbuf *rb;
2852
2853	rb = get_free_recv_buffer();
2854
2855	if (rb == NULL)
2856	{
2857		/*
2858		 * No buffer space available - just drop the packet
2859		 */
2860		char buf[RX_BUFF_SIZE];
2861
2862		buflen = read(fd, buf, sizeof buf);
2863		packets_dropped++;
2864		return (buflen);
2865	}
2866
2867	i = (rp->datalen == 0
2868	    || rp->datalen > sizeof(rb->recv_space))
2869	    ? sizeof(rb->recv_space) : rp->datalen;
2870	buflen = read(fd, (char *)&rb->recv_space, (unsigned)i);
2871
2872	if (buflen < 0)
2873	{
2874		if (errno != EINTR && errno != EAGAIN) {
2875			netsyslog(LOG_ERR, "clock read fd %d: %m", fd);
2876		}
2877		freerecvbuf(rb);
2878		return (buflen);
2879	}
2880
2881	/*
2882	 * Got one. Mark how and when it got here,
2883	 * put it on the full list and do bookkeeping.
2884	 */
2885	rb->recv_length = buflen;
2886	rb->recv_srcclock = rp->srcclock;
2887	rb->dstadr = 0;
2888	rb->fd = fd;
2889	rb->recv_time = ts;
2890	rb->receiver = rp->clock_recv;
2891
2892	if (rp->io_input)
2893	{
2894		/*
2895		 * have direct input routine for refclocks
2896		 */
2897		if (rp->io_input(rb) == 0)
2898		{
2899			/*
2900			 * data was consumed - nothing to pass up
2901			 * into block input machine
2902			 */
2903			freerecvbuf(rb);
2904			return (buflen);
2905		}
2906	}
2907
2908	add_full_recv_buffer(rb);
2909
2910	rp->recvcount++;
2911	packets_received++;
2912	return (buflen);
2913}
2914
2915#ifdef HAVE_TIMESTAMP
2916/*
2917 * extract timestamps from control message buffer
2918 */
2919static l_fp
2920	fetch_timestamp(struct recvbuf *rb, struct msghdr *msghdr, l_fp ts)
2921{
2922#ifdef USE_TIMESTAMP_CMSG
2923	struct cmsghdr *cmsghdr;
2924
2925	cmsghdr = CMSG_FIRSTHDR(msghdr);
2926	while (cmsghdr != NULL) {
2927		switch (cmsghdr->cmsg_type)
2928		{
2929		case SCM_TIMESTAMP:
2930		{
2931			struct timeval *tvp = (struct timeval *)CMSG_DATA(cmsghdr);
2932			double dtemp;
2933			l_fp nts;
2934			DPRINTF(4, ("fetch_timestamp: system network time stamp: %ld.%06ld\n", tvp->tv_sec, tvp->tv_usec));
2935			nts.l_i = tvp->tv_sec + JAN_1970;
2936			dtemp = tvp->tv_usec / 1e6;
2937
2938 			/* fuzz lower bits not covered by precision */
2939 			if (sys_precision != 0)
2940 				dtemp += (ntp_random() / FRAC - .5) / (1 <<
2941 								       -sys_precision);
2942
2943			nts.l_uf = (u_int32)(dtemp*FRAC);
2944#ifdef DEBUG_TIMING
2945			{
2946				l_fp dts = ts;
2947				L_SUB(&dts, &nts);
2948				collect_timing(rb, "input processing delay", 1, &dts);
2949				DPRINTF(4, ("fetch_timestamp: timestamp delta: %s (incl. prec fuzz)\n", lfptoa(&dts, 9)));
2950			}
2951#endif
2952			ts = nts;  /* network time stamp */
2953			break;
2954		}
2955		default:
2956			DPRINTF(4, ("fetch_timestamp: skipping control message 0x%x\n", cmsghdr->cmsg_type));
2957			break;
2958		}
2959		cmsghdr = CMSG_NXTHDR(msghdr, cmsghdr);
2960	}
2961#endif
2962	return ts;
2963}
2964#endif
2965
2966/*
2967 * Routine to read the network NTP packets for a specific interface
2968 * Return the number of bytes read. That way we know if we should
2969 * read it again or go on to the next one if no bytes returned
2970 */
2971static inline int
2972read_network_packet(SOCKET fd, struct interface *itf, l_fp ts)
2973{
2974	GETSOCKNAME_SOCKLEN_TYPE fromlen;
2975	int buflen;
2976	register struct recvbuf *rb;
2977#ifdef HAVE_TIMESTAMP
2978	struct msghdr msghdr;
2979	struct iovec iovec;
2980	char control[TIMESTAMP_CTLMSGBUF_SIZE];	/* pick up control messages */
2981#endif
2982
2983	/*
2984	 * Get a buffer and read the frame.  If we
2985	 * haven't got a buffer, or this is received
2986	 * on a disallowed socket, just dump the
2987	 * packet.
2988	 */
2989
2990	rb = get_free_recv_buffer();
2991
2992	if (rb == NULL || itf->ignore_packets == ISC_TRUE)
2993	{
2994		char buf[RX_BUFF_SIZE];
2995		struct sockaddr_storage from;
2996		if (rb != NULL)
2997			freerecvbuf(rb);
2998
2999		fromlen = sizeof(from);
3000		buflen = recvfrom(fd, buf, sizeof(buf), 0,
3001				(struct sockaddr*)&from, &fromlen);
3002		DPRINTF(4, ("%s on (%lu) fd=%d from %s\n",
3003			(itf->ignore_packets == ISC_TRUE) ? "ignore" : "drop",
3004			free_recvbuffs(), fd,
3005			stoa(&from)));
3006		if (itf->ignore_packets == ISC_TRUE)
3007			packets_ignored++;
3008		else
3009			packets_dropped++;
3010		return (buflen);
3011	}
3012
3013	fromlen = sizeof(struct sockaddr_storage);
3014
3015#ifndef HAVE_TIMESTAMP
3016	rb->recv_length = recvfrom(fd,
3017			  (char *)&rb->recv_space,
3018			   sizeof(rb->recv_space), 0,
3019			   (struct sockaddr *)&rb->recv_srcadr,
3020			   &fromlen);
3021#else
3022	iovec.iov_base        = (void *)&rb->recv_space;
3023	iovec.iov_len         = sizeof(rb->recv_space);
3024	msghdr.msg_name       = (void *)&rb->recv_srcadr;
3025	msghdr.msg_namelen    = sizeof(rb->recv_srcadr);
3026	msghdr.msg_iov        = &iovec;
3027	msghdr.msg_iovlen     = 1;
3028	msghdr.msg_control    = (void *)&control;
3029	msghdr.msg_controllen = sizeof(control);
3030	msghdr.msg_flags      = 0;
3031	rb->recv_length       = recvmsg(fd, &msghdr, 0);
3032#endif
3033
3034	buflen = rb->recv_length;
3035
3036	if (buflen == 0 || (buflen == -1 &&
3037	    (errno==EWOULDBLOCK
3038#ifdef EAGAIN
3039	   || errno==EAGAIN
3040#endif
3041	 ))) {
3042		freerecvbuf(rb);
3043		return (buflen);
3044	}
3045	else if (buflen < 0)
3046	{
3047		netsyslog(LOG_ERR, "recvfrom(%s) fd=%d: %m",
3048		stoa(&rb->recv_srcadr), fd);
3049		DPRINTF(5, ("read_network_packet: fd=%d dropped (bad recvfrom)\n", fd));
3050		freerecvbuf(rb);
3051		return (buflen);
3052	}
3053
3054#ifdef DEBUG
3055	if (debug > 2) {
3056		if(rb->recv_srcadr.ss_family == AF_INET)
3057			printf("read_network_packet: fd=%d length %d from %08lx %s\n",
3058				fd, buflen,
3059				(u_long)ntohl(((struct sockaddr_in*)&rb->recv_srcadr)->sin_addr.s_addr) &
3060				0x00000000ffffffff,
3061				stoa(&rb->recv_srcadr));
3062		else
3063			printf("read_network_packet: fd=%d length %d from %s\n",
3064				fd, buflen,
3065				stoa(&rb->recv_srcadr));
3066	}
3067#endif
3068
3069	/*
3070	 * Got one.  Mark how and when it got here,
3071	 * put it on the full list and do bookkeeping.
3072	 */
3073	rb->dstadr = itf;
3074	rb->fd = fd;
3075#ifdef HAVE_TIMESTAMP
3076	ts = fetch_timestamp(rb, &msghdr, ts);  /* pick up a network time stamp if possible */
3077#endif
3078	rb->recv_time = ts;
3079	rb->receiver = receive;
3080
3081	add_full_recv_buffer(rb);
3082
3083	itf->received++;
3084	packets_received++;
3085	return (buflen);
3086}
3087
3088/*
3089 * input_handler - receive packets asynchronously
3090 */
3091void
3092input_handler(
3093	l_fp *cts
3094	)
3095{
3096
3097	int buflen;
3098	int n;
3099	int doing;
3100	SOCKET fd;
3101	struct timeval tvzero;
3102	l_fp ts;			/* Timestamp at BOselect() gob */
3103#ifdef DEBUG_TIMING
3104	l_fp ts_e;			/* Timestamp at EOselect() gob */
3105#endif
3106	fd_set fds;
3107	int select_count = 0;
3108	struct interface *interface;
3109#if defined(HAS_ROUTING_SOCKET)
3110	struct asyncio_reader *asyncio_reader;
3111#endif
3112
3113	handler_calls++;
3114
3115	/*
3116	 * If we have something to do, freeze a timestamp.
3117	 * See below for the other cases (nothing (left) to do or error)
3118	 */
3119	ts = *cts;
3120
3121	/*
3122	 * Do a poll to see who has data
3123	 */
3124
3125	fds = activefds;
3126	tvzero.tv_sec = tvzero.tv_usec = 0;
3127
3128	n = select(maxactivefd+1, &fds, (fd_set *)0, (fd_set *)0, &tvzero);
3129
3130	/*
3131	 * If there are no packets waiting just return
3132	 */
3133	if (n < 0)
3134	{
3135		int err = errno;
3136		/*
3137		 * extended FAU debugging output
3138		 */
3139		if (err != EINTR)
3140		    netsyslog(LOG_ERR,
3141			      "select(%d, %s, 0L, 0L, &0.0) error: %m",
3142			      maxactivefd+1,
3143			      fdbits(maxactivefd, &activefds));
3144		if (err == EBADF) {
3145			int j, b;
3146			fds = activefds;
3147			for (j = 0; j <= maxactivefd; j++)
3148			    if ((FD_ISSET(j, &fds) && (read(j, &b, 0) == -1)))
3149				netsyslog(LOG_ERR, "Bad file descriptor %d", j);
3150		}
3151		return;
3152	}
3153	else if (n == 0)
3154		return;
3155
3156	++handler_pkts;
3157
3158#ifdef REFCLOCK
3159	/*
3160	 * Check out the reference clocks first, if any
3161	 */
3162
3163	if (refio != NULL)
3164	{
3165		register struct refclockio *rp;
3166
3167		for (rp = refio; rp != NULL; rp = rp->next)
3168		{
3169			fd = rp->fd;
3170
3171			if (FD_ISSET(fd, &fds))
3172			{
3173				do {
3174					++select_count;
3175					buflen = read_refclock_packet(fd, rp, ts);
3176				} while (buflen > 0);
3177
3178			} /* End if (FD_ISSET(fd, &fds)) */
3179		} /* End for (rp = refio; rp != 0 && n > 0; rp = rp->next) */
3180	} /* End if (refio != 0) */
3181
3182#endif /* REFCLOCK */
3183
3184	/*
3185	 * Loop through the interfaces looking for data to read.
3186	 */
3187	for (interface = ISC_LIST_TAIL(inter_list);
3188	     interface != NULL;
3189	     interface = ISC_LIST_PREV(interface, link))
3190	{
3191		for (doing = 0; (doing < 2); doing++)
3192		{
3193			if (doing == 0)
3194			{
3195				fd = interface->fd;
3196			}
3197			else
3198			{
3199				if (!(interface->flags & INT_BCASTOPEN))
3200				    break;
3201				fd = interface->bfd;
3202			}
3203			if (fd < 0) continue;
3204			if (FD_ISSET(fd, &fds))
3205			{
3206				do {
3207					++select_count;
3208					buflen = read_network_packet(fd, interface, ts);
3209				} while (buflen > 0);
3210			}
3211		/* Check more interfaces */
3212		}
3213	}
3214
3215#ifdef HAS_ROUTING_SOCKET
3216	/*
3217	 * scan list of asyncio readers - currently only used for routing sockets
3218	 */
3219	asyncio_reader = ISC_LIST_TAIL(asyncio_reader_list);
3220
3221	while (asyncio_reader != NULL)
3222	{
3223	        struct asyncio_reader *next = ISC_LIST_PREV(asyncio_reader, link);
3224		if (FD_ISSET(asyncio_reader->fd, &fds)) {
3225			++select_count;
3226			asyncio_reader->receiver(asyncio_reader);
3227		}
3228		asyncio_reader = next;
3229	}
3230#endif /* HAS_ROUTING_SOCKET */
3231
3232	/*
3233	 * Done everything from that select.
3234	 */
3235
3236	/*
3237	 * If nothing to do, just return.
3238	 * If an error occurred, complain and return.
3239	 */
3240	if (select_count == 0) /* We really had nothing to do */
3241	{
3242#ifdef DEBUG
3243		if (debug)
3244		    netsyslog(LOG_DEBUG, "input_handler: select() returned 0");
3245#endif
3246		return;
3247	}
3248		/* We've done our work */
3249#ifdef DEBUG_TIMING
3250	get_systime(&ts_e);
3251	/*
3252	 * (ts_e - ts) is the amount of time we spent
3253	 * processing this gob of file descriptors.  Log
3254	 * it.
3255	 */
3256	L_SUB(&ts_e, &ts);
3257	collect_timing(NULL, "input handler", 1, &ts_e);
3258	if (debug > 3)
3259	    netsyslog(LOG_INFO, "input_handler: Processed a gob of fd's in %s msec", lfptoms(&ts_e, 6));
3260#endif
3261	/* just bail. */
3262	return;
3263}
3264
3265#endif
3266
3267/*
3268 * findinterface - find local interface corresponding to address
3269 */
3270struct interface *
3271findinterface(
3272	struct sockaddr_storage *addr
3273	)
3274{
3275	struct interface *interface;
3276
3277	interface = findlocalinterface(addr, INT_WILDCARD);
3278
3279	if (interface == NULL)
3280	{
3281		DPRINTF(4, ("Found no interface for address %s - returning wildcard\n",
3282			    stoa(addr)));
3283
3284		return (ANY_INTERFACE_CHOOSE(addr));
3285	}
3286	else
3287	{
3288		DPRINTF(4, ("Found interface #%d %s for address %s\n",
3289			    interface->ifnum, interface->name, stoa(addr)));
3290
3291		return (interface);
3292	}
3293}
3294
3295/*
3296 * findlocalinterface - find local interface index corresponding to address
3297 *
3298 * This code attempts to find the local sending address for an outgoing
3299 * address by connecting a new socket to destinationaddress:NTP_PORT
3300 * and reading the sockname of the resulting connect.
3301 * the complicated sequence simulates the routing table lookup
3302 * for to first hop without duplicating any of the routing logic into
3303 * ntpd. preferably we would have used an API call - but its not there -
3304 * so this is the best we can do here short of duplicating to entire routing
3305 * logic in ntpd which would be a silly and really unportable thing to do.
3306 *
3307 */
3308static struct interface *
3309findlocalinterface(
3310	struct sockaddr_storage *addr,
3311	int flags
3312	)
3313{
3314	SOCKET s;
3315	int rtn;
3316	struct sockaddr_storage saddr;
3317	GETSOCKNAME_SOCKLEN_TYPE saddrlen = SOCKLEN(addr);
3318	struct interface *iface;
3319
3320	DPRINTF(4, ("Finding interface for addr %s in list of addresses\n",
3321		    stoa(addr)));
3322
3323	memset(&saddr, 0, sizeof(saddr));
3324	saddr.ss_family = addr->ss_family;
3325	if(addr->ss_family == AF_INET) {
3326		memcpy(&((struct sockaddr_in*)&saddr)->sin_addr, &((struct sockaddr_in*)addr)->sin_addr, sizeof(struct in_addr));
3327		((struct sockaddr_in*)&saddr)->sin_port = htons(NTP_PORT);
3328	}
3329#ifdef INCLUDE_IPV6_SUPPORT
3330	else if(addr->ss_family == AF_INET6) {
3331 		memcpy(&((struct sockaddr_in6*)&saddr)->sin6_addr, &((struct sockaddr_in6*)addr)->sin6_addr, sizeof(struct in6_addr));
3332		((struct sockaddr_in6*)&saddr)->sin6_port = htons(NTP_PORT);
3333# ifdef ISC_PLATFORM_HAVESCOPEID
3334		((struct sockaddr_in6*)&saddr)->sin6_scope_id = ((struct sockaddr_in6*)addr)->sin6_scope_id;
3335# endif
3336	}
3337#endif
3338
3339	s = socket(addr->ss_family, SOCK_DGRAM, 0);
3340	if (s == INVALID_SOCKET)
3341		return NULL;
3342
3343	rtn = connect(s, (struct sockaddr *)&saddr, SOCKLEN(&saddr));
3344#ifndef SYS_WINNT
3345	if (rtn < 0)
3346#else
3347	if (rtn == SOCKET_ERROR)
3348#endif
3349	{
3350		closesocket(s);
3351		return NULL;
3352	}
3353
3354	rtn = getsockname(s, (struct sockaddr *)&saddr, &saddrlen);
3355	closesocket(s);
3356#ifndef SYS_WINNT
3357	if (rtn < 0)
3358#else
3359	if (rtn == SOCKET_ERROR)
3360#endif
3361		return NULL;
3362
3363	DPRINTF(4, ("findlocalinterface: kernel maps %s to %s\n", stoa(addr), stoa(&saddr)));
3364
3365	iface = getinterface(&saddr, flags);
3366
3367	/* Don't both with ignore interfaces */
3368	if (iface != NULL && iface->ignore_packets == ISC_TRUE)
3369	{
3370		return NULL;
3371	}
3372	else
3373	{
3374		return iface;
3375	}
3376}
3377
3378/*
3379 * fetch an interface structure the matches the
3380 * address is has the given flags not set
3381 */
3382static struct interface *
3383getinterface(struct sockaddr_storage *addr, int flags)
3384{
3385	struct interface *interface = find_addr_in_list(addr);
3386
3387	if (interface != NULL && interface->flags & flags)
3388	{
3389		return NULL;
3390	}
3391	else
3392	{
3393		return interface;
3394	}
3395}
3396
3397/*
3398 * findlocalcastinterface - find local *cast interface index corresponding to address
3399 * depending on the flags passed
3400 */
3401static struct interface *
3402findlocalcastinterface(
3403	struct sockaddr_storage *addr, int flags
3404	)
3405{
3406	struct interface *interface;
3407	struct interface *nif = NULL;
3408#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
3409	isc_boolean_t want_linklocal;
3410#endif
3411
3412	/*
3413	 * see how kernel maps the mcast address
3414	 */
3415        nif = findlocalinterface(addr, 0);
3416
3417	if (nif) {
3418		DPRINTF(2, ("findlocalcastinterface: kernel recommends interface #%d %s\n", nif->ifnum, nif->name));
3419		return nif;
3420	}
3421
3422#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
3423	want_linklocal = ISC_FALSE;
3424	if (addr_ismulticast(addr) && flags == INT_MULTICAST)
3425	{
3426		if (IN6_IS_ADDR_MC_LINKLOCAL(&((struct sockaddr_in6*)addr)->sin6_addr))
3427		{
3428			want_linklocal = ISC_TRUE;
3429		}
3430		else if (IN6_IS_ADDR_MC_SITELOCAL(&((struct sockaddr_in6*)addr)->sin6_addr))
3431		{
3432			want_linklocal = ISC_TRUE;
3433		}
3434	}
3435#endif
3436
3437	for (interface = ISC_LIST_HEAD(inter_list);
3438	     interface != NULL;
3439	     interface = ISC_LIST_NEXT(interface, link))
3440	  {
3441		/* use only allowed addresses */
3442		if (interface->ignore_packets == ISC_TRUE)
3443			continue;
3444
3445		/* Skip the loopback and wildcard addresses */
3446		if (interface->flags & (INT_LOOPBACK|INT_WILDCARD))
3447			continue;
3448
3449		/* Skip if different family */
3450		if(interface->sin.ss_family != addr->ss_family)
3451			continue;
3452
3453		/* Is this it one of these based on flags? */
3454		if (!(interface->flags & flags))
3455			continue;
3456
3457		/* for IPv6 multicast check the address for linklocal */
3458#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT
3459		if (flags == INT_MULTICAST && interface->sin.ss_family == AF_INET6 &&
3460		   (IN6_IS_ADDR_LINKLOCAL(&((struct sockaddr_in6*)&interface->sin)->sin6_addr))
3461		   && want_linklocal == ISC_TRUE)
3462		{
3463			nif = interface;
3464			break;
3465		}
3466		/* If we want a linklocal address and this isn't it, skip */\
3467		if (want_linklocal == ISC_TRUE)
3468			continue;
3469#endif
3470		/* Otherwise just look for the flag */
3471		if((interface->flags & flags))
3472		{
3473			nif = interface;
3474			break;
3475		}
3476	}
3477#ifdef DEBUG
3478	if (debug > 2)
3479	{
3480		if (nif)
3481			printf("findlocalcastinterface: found interface #%d %s\n", nif->ifnum, nif->name);
3482		else
3483			printf("findlocalcastinterface: no interface found for %s flags 0x%x\n", stoa(addr), flags);
3484	}
3485#endif
3486	return (nif);
3487}
3488
3489/*
3490 * findbcastinter - find broadcast interface corresponding to address
3491 */
3492struct interface *
3493findbcastinter(
3494	struct sockaddr_storage *addr
3495	)
3496{
3497#if !defined(MPE) && (defined(SIOCGIFCONF) || defined(SYS_WINNT))
3498        struct interface *interface;
3499
3500
3501	DPRINTF(4, ("Finding broadcast/multicast interface for addr %s in list of addresses\n",
3502		    stoa(addr)));
3503
3504	interface = findlocalinterface(addr, INT_LOOPBACK|INT_WILDCARD);
3505
3506	if (interface != NULL)
3507	{
3508		DPRINTF(4, ("Found bcast-/mcast- interface index #%d %s\n", interface->ifnum, interface->name));
3509		return interface;
3510	}
3511
3512	/* plan B - try to find something reasonable in our lists in case kernel lookup doesn't help */
3513
3514	for (interface = ISC_LIST_HEAD(inter_list);
3515	     interface != NULL;
3516	     interface = ISC_LIST_NEXT(interface, link))
3517	{
3518	        if (interface->flags & INT_WILDCARD)
3519		        continue;
3520
3521		/* Don't bother with ignored interfaces */
3522		if (interface->ignore_packets == ISC_TRUE)
3523			continue;
3524
3525		/*
3526		 * First look if this is the correct family
3527		 */
3528		if(interface->sin.ss_family != addr->ss_family)
3529	  		continue;
3530
3531		/* Skip the loopback addresses */
3532		if (interface->flags & INT_LOOPBACK)
3533			continue;
3534
3535		/*
3536		 * If we are looking to match a multicast address grab it.
3537		 */
3538		if (addr_ismulticast(addr) == ISC_TRUE && interface->flags & INT_MULTICAST)
3539		{
3540#ifdef INCLUDE_IPV6_SUPPORT
3541			if(addr->ss_family == AF_INET6) {
3542				/* Only use link-local address for link-scope mcast */
3543				if(IN6_IS_ADDR_MC_LINKLOCAL(&((struct sockaddr_in6*)addr)->sin6_addr) &&
3544				  !IN6_IS_ADDR_LINKLOCAL(&((struct sockaddr_in6*)&interface->sin)->sin6_addr)) {
3545					continue;
3546				}
3547			}
3548#endif
3549			break;
3550		}
3551
3552		/*
3553		 * We match only those interfaces marked as
3554		 * broadcastable and either the explicit broadcast
3555		 * address or the network portion of the IP address.
3556		 * Sloppy.
3557		 */
3558		if(addr->ss_family == AF_INET) {
3559			if (SOCKCMP(&interface->bcast, addr)) {
3560				break;
3561			}
3562			if ((NSRCADR(&interface->sin) &
3563			     NSRCADR(&interface->mask)) == (NSRCADR(addr) &
3564							    NSRCADR(&interface->mask)))
3565				break;
3566		}
3567#ifdef INCLUDE_IPV6_SUPPORT
3568		else if(addr->ss_family == AF_INET6) {
3569			if (SOCKCMP(&interface->bcast, addr)) {
3570				break;
3571			}
3572			if (SOCKCMP(netof(&interface->sin), netof(addr))) {
3573				break;
3574			}
3575		}
3576#endif
3577	}
3578#endif /* SIOCGIFCONF */
3579	if (interface == NULL) {
3580		DPRINTF(4, ("No bcast interface found for %s\n", stoa(addr)));
3581		return ANY_INTERFACE_CHOOSE(addr);
3582	} else {
3583		DPRINTF(4, ("Found bcast-/mcast- interface index #%d %s\n", interface->ifnum, interface->name));
3584		return interface;
3585	}
3586}
3587
3588
3589/*
3590 * io_clr_stats - clear I/O module statistics
3591 */
3592void
3593io_clr_stats(void)
3594{
3595	packets_dropped = 0;
3596	packets_ignored = 0;
3597	packets_received = 0;
3598	packets_sent = 0;
3599	packets_notsent = 0;
3600
3601	handler_calls = 0;
3602	handler_pkts = 0;
3603	io_timereset = current_time;
3604}
3605
3606
3607#ifdef REFCLOCK
3608/*
3609 * io_addclock - add a reference clock to the list and arrange that we
3610 *				 get SIGIO interrupts from it.
3611 */
3612int
3613io_addclock(
3614	struct refclockio *rio
3615	)
3616{
3617	BLOCKIO();
3618	/*
3619	 * Stuff the I/O structure in the list and mark the descriptor
3620	 * in use.	There is a harmless (I hope) race condition here.
3621	 */
3622	rio->next = refio;
3623
3624# ifdef HAVE_SIGNALED_IO
3625	if (init_clock_sig(rio))
3626	{
3627		UNBLOCKIO();
3628		return 0;
3629	}
3630# elif defined(HAVE_IO_COMPLETION_PORT)
3631	if (io_completion_port_add_clock_io(rio))
3632	{
3633		UNBLOCKIO();
3634		return 0;
3635	}
3636# endif
3637
3638	/*
3639	 * enqueue
3640	 */
3641	refio = rio;
3642
3643        /*
3644	 * register fd
3645	 */
3646	add_fd_to_list(rio->fd, FD_TYPE_FILE);
3647
3648	UNBLOCKIO();
3649	return 1;
3650}
3651
3652/*
3653 * io_closeclock - close the clock in the I/O structure given
3654 */
3655void
3656io_closeclock(
3657	struct refclockio *rio
3658	)
3659{
3660	BLOCKIO();
3661	/*
3662	 * Remove structure from the list
3663	 */
3664	if (refio == rio)
3665	{
3666		refio = rio->next;
3667	}
3668	else
3669	{
3670		register struct refclockio *rp;
3671
3672		for (rp = refio; rp != NULL; rp = rp->next)
3673		    if (rp->next == rio)
3674		    {
3675			    rp->next = rio->next;
3676			    break;
3677		    }
3678
3679		if (rp == NULL) {
3680			UNBLOCKIO();
3681			return;
3682		}
3683	}
3684
3685	/*
3686	 * Close the descriptor.
3687	 */
3688	close_and_delete_fd_from_list(rio->fd);
3689	UNBLOCKIO();
3690}
3691#endif	/* REFCLOCK */
3692
3693/*
3694 * On NT a SOCKET is an unsigned int so we cannot possibly keep it in
3695 * an array. So we use one of the ISC_LIST functions to hold the
3696 * socket value and use that when we want to enumerate it.
3697 */
3698void
3699kill_asyncio(int startfd)
3700{
3701	vsock_t *lsock;
3702	vsock_t *next;
3703
3704	BLOCKIO();
3705
3706	lsock = ISC_LIST_HEAD(fd_list);
3707	while (lsock != NULL) {
3708		/*
3709		 * careful here - list is being dismantled while
3710		 * we scan it - setting next here insures that
3711		 * we are able to correctly scan the list
3712		 */
3713		next = ISC_LIST_NEXT(lsock, link);
3714		/*
3715		 * will remove socket from list
3716		 */
3717		close_and_delete_fd_from_list(lsock->fd);
3718		lsock = next;
3719	}
3720
3721	UNBLOCKIO();
3722}
3723
3724/*
3725 * Add and delete functions for the list of open sockets
3726 */
3727static void
3728add_fd_to_list(SOCKET fd, enum desc_type type) {
3729	vsock_t *lsock = (vsock_t *)emalloc(sizeof(vsock_t));
3730	lsock->fd = fd;
3731	lsock->type = type;
3732
3733	ISC_LIST_APPEND(fd_list, lsock, link);
3734	/*
3735	 * I/O Completion Ports don't care about the select and FD_SET
3736	 */
3737#ifndef HAVE_IO_COMPLETION_PORT
3738	if (fd < 0 || fd >= FD_SETSIZE) {
3739		msyslog(LOG_ERR, "Too many sockets in use, FD_SETSIZE %d exceeded",
3740			FD_SETSIZE);
3741		exit(1);
3742	}
3743	/*
3744	 * keep activefds in sync
3745	 */
3746	if (fd > maxactivefd)
3747	    maxactivefd = fd;
3748	FD_SET( (u_int)fd, &activefds);
3749#endif
3750}
3751
3752static void
3753close_and_delete_fd_from_list(SOCKET fd) {
3754
3755	vsock_t *next;
3756	vsock_t *lsock = ISC_LIST_HEAD(fd_list);
3757
3758	while(lsock != NULL) {
3759		next = ISC_LIST_NEXT(lsock, link);
3760		if(lsock->fd == fd) {
3761			ISC_LIST_DEQUEUE_TYPE(fd_list, lsock, link, vsock_t);
3762
3763			switch (lsock->type) {
3764			case FD_TYPE_SOCKET:
3765#ifdef SYS_WINNT
3766				closesocket(lsock->fd);
3767				break;
3768#endif
3769			case FD_TYPE_FILE:
3770				(void) close(lsock->fd);
3771				break;
3772			default:
3773				msyslog(LOG_ERR, "internal error - illegal descriptor type %d - EXITING", (int)lsock->type);
3774				exit(1);
3775			}
3776
3777			free(lsock);
3778			/*
3779			 * I/O Completion Ports don't care about select and fd_set
3780			 */
3781#ifndef HAVE_IO_COMPLETION_PORT
3782			/*
3783			 * remove from activefds
3784			 */
3785			FD_CLR( (u_int) fd, &activefds);
3786
3787			if (fd == maxactivefd) {
3788				int i, newmax = 0;
3789				for (i = 0; i < maxactivefd; i++)
3790					if (FD_ISSET(i, &activefds))
3791						newmax = i;
3792				maxactivefd = newmax;
3793			}
3794#endif
3795			break;
3796		}
3797		lsock = next;
3798	}
3799}
3800
3801static void
3802add_addr_to_list(struct sockaddr_storage *addr, struct interface *interface){
3803#ifdef DEBUG
3804	if (find_addr_in_list(addr) == NULL) {
3805#endif
3806		/* not there yet - add to list */
3807		remaddr_t *laddr = (remaddr_t *)emalloc(sizeof(remaddr_t));
3808		memcpy(&laddr->addr, addr, sizeof(struct sockaddr_storage));
3809		laddr->interface = interface;
3810
3811		ISC_LIST_APPEND(remoteaddr_list, laddr, link);
3812
3813		DPRINTF(4, ("Added addr %s to list of addresses\n",
3814			    stoa(addr)));
3815#ifdef DEBUG
3816	} else {
3817		DPRINTF(4, ("WARNING: Attempt to add duplicate addr %s to address list\n",
3818			    stoa(addr)));
3819	}
3820#endif
3821}
3822
3823static void
3824delete_addr_from_list(struct sockaddr_storage *addr) {
3825
3826	remaddr_t *next;
3827	remaddr_t *laddr = ISC_LIST_HEAD(remoteaddr_list);
3828
3829	while(laddr != NULL) {
3830		next = ISC_LIST_NEXT(laddr, link);
3831		if(SOCKCMP(&laddr->addr, addr)) {
3832			ISC_LIST_DEQUEUE_TYPE(remoteaddr_list, laddr, link, remaddr_t);
3833			DPRINTF(4, ("Deleted addr %s from list of addresses\n",
3834				    stoa(addr)));
3835			free(laddr);
3836			break;
3837		}
3838		laddr = next;
3839	}
3840}
3841
3842static void
3843delete_interface_from_list(struct interface *iface) {
3844	remaddr_t *next;
3845	remaddr_t *laddr = ISC_LIST_HEAD(remoteaddr_list);
3846
3847	while(laddr != NULL) {
3848		next = ISC_LIST_NEXT(laddr, link);
3849		if (laddr->interface == iface) {
3850			ISC_LIST_DEQUEUE_TYPE(remoteaddr_list, laddr, link, remaddr_t);
3851			DPRINTF(4, ("Deleted addr %s for interface #%d %s from list of addresses\n",
3852				    stoa(&laddr->addr), iface->ifnum, iface->name));
3853			free(laddr);
3854		}
3855		laddr = next;
3856	}
3857}
3858
3859static struct interface *
3860find_addr_in_list(struct sockaddr_storage *addr) {
3861
3862	remaddr_t *next;
3863	remaddr_t *laddr = ISC_LIST_HEAD(remoteaddr_list);
3864	DPRINTF(4, ("Searching for addr %s in list of addresses - ",
3865		    stoa(addr)));
3866
3867	while(laddr != NULL) {
3868		next = ISC_LIST_NEXT(laddr, link);
3869		if(SOCKCMP(&laddr->addr, addr)) {
3870			DPRINTF(4, ("FOUND\n"));
3871			return laddr->interface;
3872		}
3873		else
3874			laddr = next;
3875	}
3876	DPRINTF(4, ("NOT FOUND\n"));
3877	return NULL; /* Not found */
3878}
3879
3880/*
3881 * Find the given address with the associated flag in the list
3882 */
3883static struct interface *
3884find_flagged_addr_in_list(struct sockaddr_storage *addr, int flag) {
3885
3886	remaddr_t *next;
3887	remaddr_t *laddr = ISC_LIST_HEAD(remoteaddr_list);
3888	DPRINTF(4, ("Finding addr %s in list of addresses\n",
3889		    stoa(addr)));
3890
3891	while(laddr != NULL) {
3892		next = ISC_LIST_NEXT(laddr, link);
3893		if(SOCKCMP(&laddr->addr, addr) && (laddr->interface->flags & flag)) {
3894			return laddr->interface;
3895			break;
3896		}
3897		else
3898			laddr = next;
3899	}
3900	return NULL; /* Not found */
3901}
3902
3903#ifdef HAS_ROUTING_SOCKET
3904#include <net/route.h>
3905
3906#ifndef UPDATE_GRACE
3907#define UPDATE_GRACE	2	/* wait UPDATE_GRACE seconds before scanning */
3908#endif
3909
3910static void
3911process_routing_msgs(struct asyncio_reader *reader)
3912{
3913	char buffer[5120];
3914	char *p = buffer;
3915
3916	int cnt;
3917
3918	if (disable_dynamic_updates) {
3919		/*
3920		 * discard ourselves if we are not need any more
3921		 * usually happens when running unprivileged
3922		 */
3923		remove_asyncio_reader(reader);
3924		delete_asyncio_reader(reader);
3925		return;
3926	}
3927
3928	cnt = read(reader->fd, buffer, sizeof(buffer));
3929
3930	if (cnt < 0) {
3931		msyslog(LOG_ERR, "i/o error on routing socket %m - disabling");
3932		remove_asyncio_reader(reader);
3933		delete_asyncio_reader(reader);
3934		return;
3935	}
3936
3937	/*
3938	 * process routing message
3939	 */
3940	while ((p + sizeof(struct rt_msghdr)) <= (buffer + cnt))
3941	{
3942		struct rt_msghdr *rtm;
3943
3944		rtm = (struct rt_msghdr *)p;
3945		if (rtm->rtm_version != RTM_VERSION) {
3946			msyslog(LOG_ERR, "version mismatch on routing socket %m - disabling");
3947			remove_asyncio_reader(reader);
3948			delete_asyncio_reader(reader);
3949			return;
3950		}
3951
3952		switch (rtm->rtm_type) {
3953#ifdef RTM_NEWADDR
3954		case RTM_NEWADDR:
3955#endif
3956#ifdef RTM_DELADDR
3957		case RTM_DELADDR:
3958#endif
3959#ifdef RTM_ADD
3960		case RTM_ADD:
3961#endif
3962#ifdef RTM_DELETE
3963		case RTM_DELETE:
3964#endif
3965#ifdef RTM_REDIRECT
3966		case RTM_REDIRECT:
3967#endif
3968#ifdef RTM_CHANGE
3969		case RTM_CHANGE:
3970#endif
3971#ifdef RTM_LOSING
3972		case RTM_LOSING:
3973#endif
3974#ifdef RTM_IFINFO
3975		case RTM_IFINFO:
3976#endif
3977#ifdef RTM_IFANNOUNCE
3978		case RTM_IFANNOUNCE:
3979#endif
3980			/*
3981			 * we are keen on new and deleted addresses and if an interface goes up and down or routing changes
3982			 */
3983			DPRINTF(3, ("routing message op = %d: scheduling interface update\n", rtm->rtm_type));
3984			timer_interfacetimeout(current_time + UPDATE_GRACE);
3985			break;
3986		default:
3987			/*
3988			 * the rest doesn't bother us.
3989			 */
3990			DPRINTF(4, ("routing message op = %d: ignored\n", rtm->rtm_type));
3991			break;
3992		}
3993		p += rtm->rtm_msglen;
3994	}
3995}
3996
3997/*
3998 * set up routing notifications
3999 */
4000static void
4001init_async_notifications()
4002{
4003	struct asyncio_reader *reader;
4004	int fd = socket(PF_ROUTE, SOCK_RAW, 0);
4005
4006	if (fd >= 0) {
4007		fd = move_fd(fd);
4008		init_nonblocking_io(fd);
4009#if defined(HAVE_SIGNALED_IO)
4010		init_socket_sig(fd);
4011#endif /* HAVE_SIGNALED_IO */
4012
4013		reader = new_asyncio_reader();
4014
4015		reader->fd = fd;
4016		reader->receiver = process_routing_msgs;
4017
4018		add_asyncio_reader(reader, FD_TYPE_SOCKET);
4019		msyslog(LOG_INFO, "Listening on routing socket on fd #%d for interface updates", fd);
4020	} else {
4021		msyslog(LOG_ERR, "unable to open routing socket (%m) - using polled interface update");
4022	}
4023}
4024#else
4025static void
4026init_async_notifications()
4027{
4028}
4029#endif
4030