diff options
Diffstat (limited to 'contrib/ntp/ntpd/ntp_io.c')
-rw-r--r-- | contrib/ntp/ntpd/ntp_io.c | 5359 |
1 files changed, 3054 insertions, 2305 deletions
diff --git a/contrib/ntp/ntpd/ntp_io.c b/contrib/ntp/ntpd/ntp_io.c index 44e7f80..2b5a003 100644 --- a/contrib/ntp/ntpd/ntp_io.c +++ b/contrib/ntp/ntpd/ntp_io.c @@ -7,41 +7,61 @@ # include <config.h> #endif +#include <stdio.h> +#include <signal.h> +#ifdef HAVE_FNMATCH_H +# include <fnmatch.h> +# if !defined(FNM_CASEFOLD) && defined(FNM_IGNORECASE) +# define FNM_CASEFOLD FNM_IGNORECASE +# endif +#endif +#ifdef HAVE_SYS_PARAM_H +# include <sys/param.h> +#endif +#ifdef HAVE_SYS_IOCTL_H +# include <sys/ioctl.h> +#endif +#ifdef HAVE_SYS_SOCKIO_H /* UXPV: SIOC* #defines (Frank Vance <fvance@waii.com>) */ +# include <sys/sockio.h> +#endif +#ifdef HAVE_SYS_UIO_H +# include <sys/uio.h> +#endif + #include "ntp_machine.h" #include "ntpd.h" #include "ntp_io.h" #include "iosignal.h" +#include "ntp_lists.h" #include "ntp_refclock.h" #include "ntp_stdlib.h" +#include "ntp_worker.h" #include "ntp_request.h" -#include "ntp.h" -#include "ntp_unixtime.h" +#include "ntp_assert.h" +#include "timevalops.h" +#include "timespecops.h" +#include "ntpd-opts.h" /* Don't include ISC's version of IPv6 variables and structures */ #define ISC_IPV6_H 1 +#include <isc/mem.h> #include <isc/interfaceiter.h> -#include <isc/list.h> +#include <isc/netaddr.h> #include <isc/result.h> +#include <isc/sockaddr.h> #ifdef SIM #include "ntpsim.h" #endif -#include <stdio.h> -#include <signal.h> -#ifdef HAVE_SYS_PARAM_H -# include <sys/param.h> -#endif /* HAVE_SYS_PARAM_H */ -#ifdef HAVE_SYS_IOCTL_H -# include <sys/ioctl.h> -#endif -#ifdef HAVE_SYS_SOCKIO_H /* UXPV: SIOC* #defines (Frank Vance <fvance@waii.com>) */ -# include <sys/sockio.h> -#endif -#ifdef HAVE_SYS_UIO_H -# include <sys/uio.h> +#ifdef HAS_ROUTING_SOCKET +# include <net/route.h> +# ifdef HAVE_RTNETLINK +# include <linux/rtnetlink.h> +# endif #endif + /* * setsockopt does not always have the same arg declaration * across all platforms. If it's not defined we make it empty @@ -51,63 +71,76 @@ #define SETSOCKOPT_ARG_CAST #endif -/* - * Set up some macros to look for IPv6 and IPv6 multicast - */ - -#if defined(ISC_PLATFORM_HAVEIPV6) && !defined(DISABLE_IPV6) - -#define INCLUDE_IPV6_SUPPORT - -#if defined(INCLUDE_IPV6_SUPPORT) && defined(IPV6_JOIN_GROUP) && defined(IPV6_LEAVE_GROUP) -#define INCLUDE_IPV6_MULTICAST_SUPPORT +extern int listen_to_virtual_ips; -#endif /* IPV6 Multicast Support */ -#endif /* IPv6 Support */ +#ifndef IPTOS_DSCP_EF +#define IPTOS_DSCP_EF 0xb8 +#endif +int qos = IPTOS_DSCP_EF; /* QoS RFC3246 */ -#ifdef INCLUDE_IPV6_SUPPORT -#include <netinet/in.h> -#include <net/if_var.h> -#include <netinet/in_var.h> -#endif /* !INCLUDE_IPV6_SUPPORT */ +#ifdef LEAP_SMEAR +/* TODO burnicki: This should be moved to ntp_timer.c, but if we do so + * we get a linker error. Since we're running out of time before the leap + * second occurs, we let it here where it just works. + */ +int leap_smear_intv; +#endif -extern int listen_to_virtual_ips; -extern const char *specific_interface; +/* + * NIC rule entry + */ +typedef struct nic_rule_tag nic_rule; + +struct nic_rule_tag { + nic_rule * next; + nic_rule_action action; + nic_rule_match match_type; + char * if_name; + sockaddr_u addr; + int prefixlen; +}; -#if defined(SO_TIMESTAMP) && defined(SCM_TIMESTAMP) -#if defined(CMSG_FIRSTHDR) -#define HAVE_TIMESTAMP -#define USE_TIMESTAMP_CMSG -#ifndef TIMESTAMP_CTLMSGBUF_SIZE -#define TIMESTAMP_CTLMSGBUF_SIZE 1536 /* moderate default */ -#endif +/* + * NIC rule listhead. Entries are added at the head so that the first + * match in the list is the last matching rule specified. + */ +nic_rule *nic_rule_list; + + +#if defined(SO_BINTIME) && defined(SCM_BINTIME) && defined(CMSG_FIRSTHDR) +# define HAVE_PACKET_TIMESTAMP +# define HAVE_BINTIME +# ifdef BINTIME_CTLMSGBUF_SIZE +# define CMSG_BUFSIZE BINTIME_CTLMSGBUF_SIZE +# else +# define CMSG_BUFSIZE 1536 /* moderate default */ +# endif +#elif defined(SO_TIMESTAMPNS) && defined(SCM_TIMESTAMPNS) && defined(CMSG_FIRSTHDR) +# define HAVE_PACKET_TIMESTAMP +# define HAVE_TIMESTAMPNS +# ifdef TIMESTAMPNS_CTLMSGBUF_SIZE +# define CMSG_BUFSIZE TIMESTAMPNS_CTLMSGBUF_SIZE +# else +# define CMSG_BUFSIZE 1536 /* moderate default */ +# endif +#elif defined(SO_TIMESTAMP) && defined(SCM_TIMESTAMP) && defined(CMSG_FIRSTHDR) +# define HAVE_PACKET_TIMESTAMP +# define HAVE_TIMESTAMP +# ifdef TIMESTAMP_CTLMSGBUF_SIZE +# define CMSG_BUFSIZE TIMESTAMP_CTLMSGBUF_SIZE +# else +# define CMSG_BUFSIZE 1536 /* moderate default */ +# endif #else /* fill in for old/other timestamp interfaces */ #endif -#endif #if defined(SYS_WINNT) -#include <transmitbuff.h> +#include "win32_io.h" #include <isc/win32os.h> -/* - * Define this macro to control the behavior of connection - * resets on UDP sockets. See Microsoft KnowledgeBase Article Q263823 - * for details. - * NOTE: This requires that Windows 2000 systems install Service Pack 2 - * or later. - */ -#ifndef SIO_UDP_CONNRESET -#define SIO_UDP_CONNRESET _WSAIOW(IOC_VENDOR,12) #endif /* - * Windows C runtime ioctl() can't deal properly with sockets, - * map to ioctlsocket for this source file. - */ -#define ioctl(fd, opt, val) ioctlsocket((fd), (opt), (u_long *)(val)) -#endif /* SYS_WINNT */ - -/* * We do asynchronous input using the SIGIO facility. A number of * recvbuf buffers are preallocated for input. In the signal * handler we poll to see which sockets are ready and read the @@ -130,8 +163,8 @@ extern const char *specific_interface; volatile u_long packets_dropped; /* total number of packets dropped on reception */ volatile u_long packets_ignored; /* packets received on wild card interface */ volatile u_long packets_received; /* total number of packets received */ -u_long packets_sent; /* total number of packets sent */ -u_long packets_notsent; /* total number of packets which couldn't be sent */ + u_long packets_sent; /* total number of packets sent */ + u_long packets_notsent; /* total number of packets which couldn't be sent */ volatile u_long handler_calls; /* number of calls to interrupt handler */ volatile u_long handler_pkts; /* number of pkts received by handler */ @@ -140,13 +173,15 @@ u_long io_timereset; /* time counters were reset */ /* * Interface stuff */ -struct interface *any_interface; /* default ipv4 interface */ -struct interface *any6_interface; /* default ipv6 interface */ -struct interface *loopback_interface; /* loopback ipv4 interface */ +endpt * any_interface; /* wildcard ipv4 interface */ +endpt * any6_interface; /* wildcard ipv6 interface */ +endpt * loopback_interface; /* loopback ipv4 interface */ +isc_boolean_t broadcast_client_enabled; /* is broadcast client enabled */ +u_int sys_ifnum; /* next .ifnum to assign */ int ninterfaces; /* Total number of interfaces */ -volatile int disable_dynamic_updates; /* when set to != 0 dynamic updates won't happen */ +int disable_dynamic_updates; /* scan interfaces once only */ #ifdef REFCLOCK /* @@ -156,67 +191,62 @@ volatile int disable_dynamic_updates; /* when set to != 0 dynamic updates won' static struct refclockio *refio; #endif /* REFCLOCK */ - -/* - * Define what the possible "soft" errors can be. These are non-fatal returns - * of various network related functions, like recv() and so on. - * - * For some reason, BSDI (and perhaps others) will sometimes return <0 - * from recv() but will have errno==0. This is broken, but we have to - * work around it here. - */ -#define SOFT_ERROR(e) ((e) == EAGAIN || \ - (e) == EWOULDBLOCK || \ - (e) == EINTR || \ - (e) == 0) - /* * File descriptor masks etc. for call to select - * Not needed for I/O Completion Ports + * Not needed for I/O Completion Ports or anything outside this file */ -fd_set activefds; -int maxactivefd; +static fd_set activefds; +static int maxactivefd; + /* * bit alternating value to detect verified interfaces during an update cycle */ -static u_char sys_interphase = 0; +static u_short sys_interphase = 0; -static struct interface *new_interface P((struct interface *)); -static void add_interface P((struct interface *)); -static int update_interfaces P((u_short, interface_receiver_t, void *)); -static void remove_interface P((struct interface *)); -static struct interface *create_interface P((u_short, struct interface *)); +static endpt * new_interface(endpt *); +static void add_interface(endpt *); +static int update_interfaces(u_short, interface_receiver_t, + void *); +static void remove_interface(endpt *); +static endpt * create_interface(u_short, endpt *); -static int move_fd P((SOCKET)); +static int is_wildcard_addr (const sockaddr_u *); /* * Multicast functions */ -static isc_boolean_t addr_ismulticast P((struct sockaddr_storage *)); +static isc_boolean_t addr_ismulticast (sockaddr_u *); +static isc_boolean_t is_not_bindable (sockaddr_u *, + const char *); + /* * Not all platforms support multicast */ #ifdef MCAST -static isc_boolean_t socket_multicast_enable P((struct interface *, int, struct sockaddr_storage *)); -static isc_boolean_t socket_multicast_disable P((struct interface *, struct sockaddr_storage *)); +static isc_boolean_t socket_multicast_enable (endpt *, sockaddr_u *); +static isc_boolean_t socket_multicast_disable(endpt *, sockaddr_u *); #endif #ifdef DEBUG -static void print_interface P((struct interface *, char *, char *)); -#define DPRINT_INTERFACE(_LVL_, _ARGS_) do { if (debug >= (_LVL_)) { print_interface _ARGS_; } } while (0) +static void interface_dump (const endpt *); +static void sockaddr_dump (const sockaddr_u *); +static void print_interface (const endpt *, const char *, const char *); +#define DPRINT_INTERFACE(level, args) do { if (debug >= (level)) { print_interface args; } } while (0) #else -#define DPRINT_INTERFACE(_LVL_, _ARGS_) do {} while (0) +#define DPRINT_INTERFACE(level, args) do {} while (0) #endif typedef struct vsock vsock_t; enum desc_type { FD_TYPE_SOCKET, FD_TYPE_FILE }; struct vsock { - SOCKET fd; - enum desc_type type; - ISC_LINK(vsock_t) link; + vsock_t * link; + SOCKET fd; + enum desc_type type; }; +vsock_t *fd_list; + #if !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET) /* * async notification processing (e. g. routing sockets) @@ -226,213 +256,132 @@ struct vsock { * like e. g. routing sockets */ struct asyncio_reader { - SOCKET fd; /* fd to be read */ - void *data; /* possibly local data */ + struct asyncio_reader *link; /* the list this is being kept in */ + SOCKET fd; /* fd to be read */ + void *data; /* possibly local data */ void (*receiver)(struct asyncio_reader *); /* input handler */ - ISC_LINK(struct asyncio_reader) link; /* the list this is being kept in */ }; -ISC_LIST(struct asyncio_reader) asyncio_reader_list; +struct asyncio_reader *asyncio_reader_list; -static void delete_asyncio_reader P((struct asyncio_reader *)); -static struct asyncio_reader *new_asyncio_reader P((void)); -static void add_asyncio_reader P((struct asyncio_reader *, enum desc_type)); -static void remove_asyncio_reader P((struct asyncio_reader *)); +static void delete_asyncio_reader (struct asyncio_reader *); +static struct asyncio_reader *new_asyncio_reader (void); +static void add_asyncio_reader (struct asyncio_reader *, enum desc_type); +static void remove_asyncio_reader (struct asyncio_reader *); #endif /* !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET) */ -static void init_async_notifications P((void)); - -static int create_sockets P((u_short)); -static SOCKET open_socket P((struct sockaddr_storage *, int, int, struct interface *)); -static char * fdbits P((int, fd_set *)); -static void set_reuseaddr P((int)); -static isc_boolean_t socket_broadcast_enable P((struct interface *, SOCKET, struct sockaddr_storage *)); -static isc_boolean_t socket_broadcast_disable P((struct interface *, struct sockaddr_storage *)); - -ISC_LIST(vsock_t) fd_list; +static void init_async_notifications (void); + +static int addr_eqprefix (const sockaddr_u *, const sockaddr_u *, + int); +static int addr_samesubnet (const sockaddr_u *, const sockaddr_u *, + const sockaddr_u *, const sockaddr_u *); +static int create_sockets (u_short); +static SOCKET open_socket (sockaddr_u *, int, int, endpt *); +static char * fdbits (int, fd_set *); +static void set_reuseaddr (int); +static isc_boolean_t socket_broadcast_enable (struct interface *, SOCKET, sockaddr_u *); +#ifdef OS_MISSES_SPECIFIC_ROUTE_UPDATES +static isc_boolean_t socket_broadcast_disable (struct interface *, sockaddr_u *); +#endif typedef struct remaddr remaddr_t; struct remaddr { - struct sockaddr_storage addr; - struct interface *interface; - ISC_LINK(remaddr_t) link; + remaddr_t * link; + sockaddr_u addr; + endpt * ep; }; -ISC_LIST(remaddr_t) remoteaddr_list; +remaddr_t * remoteaddr_list; +endpt * ep_list; /* complete endpt list */ +endpt * mc4_list; /* IPv4 mcast-capable unicast endpts */ +endpt * mc6_list; /* IPv6 mcast-capable unicast endpts */ -ISC_LIST(struct interface) inter_list; +static endpt * wildipv4; +static endpt * wildipv6; -static struct interface *wildipv4 = NULL; -static struct interface *wildipv6 = NULL; +#ifdef SYS_WINNT +int accept_wildcard_if_for_winnt; +#else +const int accept_wildcard_if_for_winnt = FALSE; +#endif -static void add_fd_to_list P((SOCKET, enum desc_type)); -static void close_and_delete_fd_from_list P((SOCKET)); -static void add_addr_to_list P((struct sockaddr_storage *, struct interface *)); -static void delete_addr_from_list P((struct sockaddr_storage *)); -static struct interface *find_addr_in_list P((struct sockaddr_storage *)); -static struct interface *find_flagged_addr_in_list P((struct sockaddr_storage *, int)); -static void create_wildcards P((u_short)); -static isc_boolean_t address_okay P((struct interface *)); -static void convert_isc_if P((isc_interface_t *, struct interface *, u_short)); -static void delete_interface_from_list P((struct interface *)); -static struct interface *getinterface P((struct sockaddr_storage *, int)); -static struct interface *findlocalinterface P((struct sockaddr_storage *, int)); -static struct interface *findlocalcastinterface P((struct sockaddr_storage *, int)); +static void add_fd_to_list (SOCKET, enum desc_type); +static endpt * find_addr_in_list (sockaddr_u *); +static endpt * find_flagged_addr_in_list(sockaddr_u *, u_int32); +static void delete_addr_from_list (sockaddr_u *); +static void delete_interface_from_list(endpt *); +static void close_and_delete_fd_from_list(SOCKET); +static void add_addr_to_list (sockaddr_u *, endpt *); +static void create_wildcards (u_short); +static endpt * findlocalinterface (sockaddr_u *, int, int); +static endpt * findclosestinterface (sockaddr_u *, int); +#ifdef DEBUG +static const char * action_text (nic_rule_action); +#endif +static nic_rule_action interface_action(char *, sockaddr_u *, u_int32); +static void convert_isc_if (isc_interface_t *, + endpt *, u_short); +static void calc_addr_distance(sockaddr_u *, + const sockaddr_u *, + const sockaddr_u *); +static int cmp_addr_distance(const sockaddr_u *, + const sockaddr_u *); /* * Routines to read the ntp packets */ #if !defined(HAVE_IO_COMPLETION_PORT) -static inline int read_network_packet P((SOCKET, struct interface *, l_fp)); -static inline int read_refclock_packet P((SOCKET, struct refclockio *, l_fp)); -#endif - -#ifdef SYS_WINNT -/* - * Windows 2000 systems incorrectly cause UDP sockets using WASRecvFrom - * to not work correctly, returning a WSACONNRESET error when a WSASendTo - * fails with an "ICMP port unreachable" response and preventing the - * socket from using the WSARecvFrom in subsequent operations. - * The function below fixes this, but requires that Windows 2000 - * Service Pack 2 or later be installed on the system. NT 4.0 - * systems are not affected by this and work correctly. - * See Microsoft Knowledge Base Article Q263823 for details of this. - */ -void -connection_reset_fix( - SOCKET fd, - struct sockaddr_storage *addr - ) -{ - DWORD dwBytesReturned = 0; - BOOL bNewBehavior = FALSE; - DWORD status; - - /* - * disable bad behavior using IOCTL: SIO_UDP_CONNRESET - * NT 4.0 has no problem - */ - if (isc_win32os_majorversion() >= 5) { - status = WSAIoctl(fd, SIO_UDP_CONNRESET, &bNewBehavior, - sizeof(bNewBehavior), NULL, 0, - &dwBytesReturned, NULL, NULL); - if (SOCKET_ERROR == status) - netsyslog(LOG_ERR, "connection_reset_fix() " - "failed for address %s: %m", - stoa(addr)); - } -} -#endif - -/* - * on Unix systems the stdio library typically - * makes use of file descriptors in the lower - * integer range. stdio usually will make use - * of the file descriptor in the range of - * [0..FOPEN_MAX) - * in order to keep this range clean for socket - * file descriptors we attempt to move them above - * FOPEM_MAX. This is not as easy as it sounds as - * FOPEN_MAX changes from implementation to implementation - * and may exceed to current file decriptor limits. - * We are using following strategy: - * - keep a current socket fd boundary initialized with - * max(0, min(getdtablesize() - FD_CHUNK, FOPEN_MAX)) - * - attempt to move the descriptor to the boundary or - * above. - * - if that fails and boundary > 0 set boundary - * to min(0, socket_fd_boundary - FD_CHUNK) - * -> retry - * if failure and boundary == 0 return old fd - * - on success close old fd return new fd - * - * effects: - * - fds will be moved above the socket fd boundary - * if at all possible. - * - the socket boundary will be reduced until - * allocation is possible or 0 is reached - at this - * point the algrithm will be disabled - */ -static int move_fd(SOCKET fd) -{ -#if !defined(SYS_WINNT) && defined(F_DUPFD) -#ifndef FD_CHUNK -#define FD_CHUNK 10 +static inline int read_network_packet (SOCKET, struct interface *, l_fp); +static void ntpd_addremove_io_fd (int, int, int); +static input_handler_t input_handler; +#ifdef REFCLOCK +static inline int read_refclock_packet (SOCKET, struct refclockio *, l_fp); #endif -/* - * number of fds we would like to have for - * stdio FILE* available. - * we can pick a "low" number as our use of - * FILE* is limited to log files and temporarily - * to data and config files. Except for log files - * we don't keep the other FILE* open beyond the - * scope of the function that opened it. - */ -#ifndef FD_PREFERRED_SOCKBOUNDARY -#define FD_PREFERRED_SOCKBOUNDARY 48 #endif -#ifndef HAVE_GETDTABLESIZE -/* - * if we have no idea about the max fd value set up things - * so we will start at FOPEN_MAX - */ -#define getdtablesize() (FOPEN_MAX+FD_CHUNK) -#endif -#ifndef FOPEN_MAX -#define FOPEN_MAX 20 /* assume that for the lack of anything better */ -#endif - static SOCKET socket_boundary = -1; - SOCKET newfd; - /* - * check whether boundary has be set up - * already - */ - if (socket_boundary == -1) { - socket_boundary = max(0, min(getdtablesize() - FD_CHUNK, - min(FOPEN_MAX, FD_PREFERRED_SOCKBOUNDARY))); -#ifdef DEBUG - msyslog(LOG_DEBUG, "ntp_io: estimated max descriptors: %d, initial socket boundary: %d", - getdtablesize(), socket_boundary); -#endif +#ifndef HAVE_IO_COMPLETION_PORT +void +maintain_activefds( + int fd, + int closing + ) +{ + int i; + + if (fd < 0 || fd >= FD_SETSIZE) { + msyslog(LOG_ERR, + "Too many sockets in use, FD_SETSIZE %d exceeded by fd %d", + FD_SETSIZE, fd); + exit(1); } - /* - * Leave a space for stdio to work in. potentially moving the - * socket_boundary lower until allocation succeeds. - */ - do { - if (fd >= 0 && fd < socket_boundary) { - /* inside reserved range: attempt to move fd */ - newfd = fcntl(fd, F_DUPFD, socket_boundary); - - if (newfd != -1) { - /* success: drop the old one - return the new one */ - (void)close(fd); - return (newfd); - } - } else { - /* outside reserved range: no work - return the original one */ - return (fd); + if (!closing) { + FD_SET(fd, &activefds); + maxactivefd = max(fd, maxactivefd); + } else { + FD_CLR(fd, &activefds); + if (maxactivefd && fd == maxactivefd) { + for (i = maxactivefd - 1; i >= 0; i--) + if (FD_ISSET(i, &activefds)) { + maxactivefd = i; + break; + } + NTP_INSIST(fd != maxactivefd); } - socket_boundary = max(0, socket_boundary - FD_CHUNK); -#ifdef DEBUG - msyslog(LOG_DEBUG, "ntp_io: selecting new socket boundary: %d", - socket_boundary); -#endif - } while (socket_boundary > 0); -#endif /* !defined(SYS_WINNT) && defined(F_DUPFD) */ - return (fd); + } } +#endif /* !HAVE_IO_COMPLETION_PORT */ + #ifdef DEBUG_TIMING /* * collect timing information for various processing - * paths. currently we only pass then on to the file + * paths. currently we only pass them on to the file * for later processing. this could also do histogram * based analysis in other to reduce the load (and skew) * dur to the file output @@ -440,16 +389,19 @@ static int move_fd(SOCKET fd) void collect_timing(struct recvbuf *rb, const char *tag, int count, l_fp *dts) { - char buf[2048]; - - snprintf(buf, sizeof(buf), "%s %d %s %s", - (rb != NULL) ? - ((rb->dstadr) ? stoa(&rb->recv_srcadr) : "-REFCLOCK-") : "-", + char buf[256]; + + snprintf(buf, sizeof(buf), "%s %d %s %s", + (rb != NULL) + ? ((rb->dstadr != NULL) + ? stoa(&rb->recv_srcadr) + : "-REFCLOCK-") + : "-", count, lfptoa(dts, 9), tag); record_timing_stats(buf); } #endif - + /* * About dynamic interfaces, sockets, reception and more... * @@ -459,7 +411,7 @@ collect_timing(struct recvbuf *rb, const char *tag, int count, l_fp *dts) * to bind to to the interface address on NTP_PORT so that * all wild and specific bindings for NTP_PORT are taken by ntpd * to avoid other daemons messing with the time or sockets. - * - all interfaces keep a list of peers that are referencing + * - all interfaces keep a list of peers that are referencing * the interface in order to quickly re-assign the peers to * new interface in case an interface is deleted (=> gone from system or * down) @@ -483,114 +435,95 @@ collect_timing(struct recvbuf *rb, const char *tag, int count, l_fp *dts) * but a list of interfaces that represent a unique address as determined by the kernel * by the procedure in findlocalinterface. Thus it is perfectly legal to see only * one representative of a group of real interfaces if they share the same address. - * + * * Frank Kardel 20050910 */ /* - * init_io - initialize I/O data structures and call socket creation routine + * init_io - initialize I/O module. */ void init_io(void) { -#ifdef SYS_WINNT - init_io_completion_port(); - - if (!Win32InitSockets()) - { - netsyslog(LOG_ERR, "No useable winsock.dll: %m"); - exit(1); - } - init_transmitbuff(); -#endif /* SYS_WINNT */ - - /* - * Init buffer free list and stat counters - */ + /* Init buffer free list and stat counters */ init_recvbuff(RECV_INIT); + /* update interface every 5 minutes as default */ + interface_interval = 300; - packets_dropped = packets_received = 0; - packets_ignored = 0; - packets_sent = packets_notsent = 0; - handler_calls = handler_pkts = 0; - io_timereset = 0; - loopback_interface = NULL; - any_interface = NULL; - any6_interface = NULL; +#ifdef WORK_PIPE + addremove_io_fd = &ntpd_addremove_io_fd; +#endif -#ifdef REFCLOCK - refio = NULL; +#ifdef SYS_WINNT + init_io_completion_port(); #endif #if defined(HAVE_SIGNALED_IO) - (void) set_signal(); + (void) set_signal(input_handler); #endif +} - ISC_LIST_INIT(fd_list); -#if !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET) - ISC_LIST_INIT(asyncio_reader_list); -#endif +static void +ntpd_addremove_io_fd( + int fd, + int is_pipe, + int remove_it + ) +{ + UNUSED_ARG(is_pipe); + +#ifdef HAVE_SIGNALED_IO + init_socket_sig(fd); +#endif /* not HAVE_SIGNALED_IO */ + + maintain_activefds(fd, remove_it); +} + + +/* + * io_open_sockets - call socket creation routine + */ +void +io_open_sockets(void) +{ + static int already_opened; - ISC_LIST_INIT(remoteaddr_list); + if (already_opened || HAVE_OPT( SAVECONFIGQUIT )) + return; - ISC_LIST_INIT(inter_list); + already_opened = 1; /* * Create the sockets */ BLOCKIO(); - (void) create_sockets(htons(NTP_PORT)); + create_sockets(NTP_PORT); UNBLOCKIO(); init_async_notifications(); - DPRINTF(3, ("init_io: maxactivefd %d\n", maxactivefd)); + DPRINTF(3, ("io_open_sockets: maxactivefd %d\n", maxactivefd)); } + #ifdef DEBUG /* * function to dump the contents of the interface structure * for debugging use only. */ void -interface_dump(struct interface *itf) +interface_dump(const endpt *itf) { - u_char* cp; - int i; - /* Limit the size of the sockaddr_storage hex dump */ - int maxsize = min(32, sizeof(struct sockaddr_storage)); - printf("Dumping interface: %p\n", itf); printf("fd = %d\n", itf->fd); printf("bfd = %d\n", itf->bfd); - printf("sin = %s,\n", stoa(&(itf->sin))); - cp = (u_char*) &(itf->sin); - for(i = 0; i < maxsize; i++) - { - printf("%02x", *cp++); - if((i+1)%4 == 0) - printf(" "); - } - printf("\n"); - printf("bcast = %s,\n", stoa(&(itf->bcast))); - cp = (u_char*) &(itf->bcast); - for(i = 0; i < maxsize; i++) - { - printf("%02x", *cp++); - if((i+1)%4 == 0) - printf(" "); - } - printf("\n"); - printf("mask = %s,\n", stoa(&(itf->mask))); - cp = (u_char*) &(itf->mask); - for(i = 0; i < maxsize; i++) - { - printf("%02x", *cp++); - if((i+1)%4 == 0) - printf(" "); - } - printf("\n"); + printf("sin = %s,\n", stoa(&itf->sin)); + sockaddr_dump(&itf->sin); + printf("bcast = %s,\n", stoa(&itf->bcast)); + sockaddr_dump(&itf->bcast); + printf("mask = %s,\n", stoa(&itf->mask)); + sockaddr_dump(&itf->mask); printf("name = %s\n", itf->name); printf("flags = 0x%08x\n", itf->flags); printf("last_ttl = %d\n", itf->last_ttl); @@ -600,40 +533,60 @@ interface_dump(struct interface *itf) printf("sent = %ld\n", itf->sent); printf("notsent = %ld\n", itf->notsent); printf("ifindex = %u\n", itf->ifindex); - printf("scopeid = %u\n", itf->scopeid); printf("peercnt = %u\n", itf->peercnt); printf("phase = %u\n", itf->phase); } /* + * sockaddr_dump - hex dump the start of a sockaddr_u + */ +static void +sockaddr_dump(const sockaddr_u *psau) +{ + /* Limit the size of the sockaddr_in6 hex dump */ + const int maxsize = min(32, sizeof(psau->sa6)); + const u_char * cp; + int i; + + /* XXX: Should we limit maxsize based on psau->saX.sin_family? */ + cp = (const void *)&psau->sa6; + + for(i = 0; i < maxsize; i++) { + printf("%02x", *cp++); + if (!((i + 1) % 4)) + printf(" "); + } + printf("\n"); +} + +/* * print_interface - helper to output debug information */ static void -print_interface(struct interface *iface, char *pfx, char *sfx) +print_interface(const endpt *iface, const char *pfx, const char *sfx) { - printf("%sinterface #%d: fd=%d, bfd=%d, name=%s, flags=0x%x, scope=%d, ifindex=%d", + printf("%sinterface #%d: fd=%d, bfd=%d, name=%s, flags=0x%x, ifindex=%u, sin=%s", pfx, iface->ifnum, iface->fd, iface->bfd, iface->name, iface->flags, - iface->scopeid, - iface->ifindex); - /* Leave these as three printf calls. */ - printf(", sin=%s", - stoa((&iface->sin))); - if (iface->flags & INT_BROADCAST) - printf(", bcast=%s,", - stoa((&iface->bcast))); - if (iface->family == AF_INET) - printf(", mask=%s", - stoa((&iface->mask))); - printf(", %s:%s", iface->ignore_packets == ISC_FALSE ? "Enabled" : "Disabled", sfx); + iface->ifindex, + stoa(&iface->sin)); + if (AF_INET == iface->family) { + if (iface->flags & INT_BROADCAST) + printf(", bcast=%s", stoa(&iface->bcast)); + printf(", mask=%s", stoa(&iface->mask)); + } + printf(", %s:%s", + (iface->ignore_packets) + ? "Disabled" + : "Enabled", + sfx); if (debug > 4) /* in-depth debugging only */ interface_dump(iface); } - #endif #if !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET) @@ -641,15 +594,13 @@ print_interface(struct interface *iface, char *pfx, char *sfx) * create an asyncio_reader structure */ static struct asyncio_reader * -new_asyncio_reader() +new_asyncio_reader(void) { struct asyncio_reader *reader; - reader = (struct asyncio_reader *)emalloc(sizeof(struct asyncio_reader)); - - memset((char *)reader, 0, sizeof(*reader)); - ISC_LINK_INIT(reader, link); + reader = emalloc_zero(sizeof(*reader)); reader->fd = INVALID_SOCKET; + return reader; } @@ -657,7 +608,9 @@ new_asyncio_reader() * delete a reader */ static void -delete_asyncio_reader(struct asyncio_reader *reader) +delete_asyncio_reader( + struct asyncio_reader *reader + ) { free(reader); } @@ -666,19 +619,26 @@ delete_asyncio_reader(struct asyncio_reader *reader) * add asynchio_reader */ static void -add_asyncio_reader(struct asyncio_reader *reader, enum desc_type type) +add_asyncio_reader( + struct asyncio_reader * reader, + enum desc_type type) { - ISC_LIST_APPEND(asyncio_reader_list, reader, link); + LINK_SLIST(asyncio_reader_list, reader, link); add_fd_to_list(reader->fd, type); } - + /* * remove asynchio_reader */ static void -remove_asyncio_reader(struct asyncio_reader *reader) +remove_asyncio_reader( + struct asyncio_reader *reader + ) { - ISC_LIST_UNLINK_TYPE(asyncio_reader_list, reader, link, struct asyncio_reader); + struct asyncio_reader *unlinked; + + UNLINK_SLIST(unlinked, asyncio_reader_list, reader, link, + struct asyncio_reader); if (reader->fd != INVALID_SOCKET) close_and_delete_fd_from_list(reader->fd); @@ -687,366 +647,874 @@ remove_asyncio_reader(struct asyncio_reader *reader) } #endif /* !defined(HAVE_IO_COMPLETION_PORT) && defined(HAS_ROUTING_SOCKET) */ + +/* compare two sockaddr prefixes */ +static int +addr_eqprefix( + const sockaddr_u * a, + const sockaddr_u * b, + int prefixlen + ) +{ + isc_netaddr_t isc_a; + isc_netaddr_t isc_b; + isc_sockaddr_t isc_sa; + + ZERO(isc_sa); + memcpy(&isc_sa.type, a, min(sizeof(isc_sa.type), sizeof(*a))); + isc_netaddr_fromsockaddr(&isc_a, &isc_sa); + + ZERO(isc_sa); + memcpy(&isc_sa.type, b, min(sizeof(isc_sa.type), sizeof(*b))); + isc_netaddr_fromsockaddr(&isc_b, &isc_sa); + + return (int)isc_netaddr_eqprefix(&isc_a, &isc_b, + (u_int)prefixlen); +} + + +static int +addr_samesubnet( + const sockaddr_u * a, + const sockaddr_u * a_mask, + const sockaddr_u * b, + const sockaddr_u * b_mask + ) +{ + const u_int32 * pa; + const u_int32 * pa_limit; + const u_int32 * pb; + const u_int32 * pm; + size_t loops; + + NTP_REQUIRE(AF(a) == AF(a_mask)); + NTP_REQUIRE(AF(b) == AF(b_mask)); + /* + * With address and mask families verified to match, comparing + * the masks also validates the address's families match. + */ + if (!SOCK_EQ(a_mask, b_mask)) + return FALSE; + + if (IS_IPV6(a)) { + loops = sizeof(NSRCADR6(a)) / sizeof(*pa); + pa = (const void *)&NSRCADR6(a); + pb = (const void *)&NSRCADR6(b); + pm = (const void *)&NSRCADR6(a_mask); + } else { + loops = sizeof(NSRCADR(a)) / sizeof(*pa); + pa = (const void *)&NSRCADR(a); + pb = (const void *)&NSRCADR(b); + pm = (const void *)&NSRCADR(a_mask); + } + for (pa_limit = pa + loops; pa < pa_limit; pa++, pb++, pm++) + if ((*pa & *pm) != (*pb & *pm)) + return FALSE; + + return TRUE; +} + + +/* + * Code to tell if we have an IP address + * If we have then return the sockaddr structure + * and set the return value + * see the bind9/getaddresses.c for details + */ +int +is_ip_address( + const char * host, + u_short af, + sockaddr_u * addr + ) +{ + struct in_addr in4; + struct addrinfo hints; + struct addrinfo *result; + struct sockaddr_in6 *resaddr6; + char tmpbuf[128]; + char *pch; + + NTP_REQUIRE(host != NULL); + NTP_REQUIRE(addr != NULL); + + ZERO_SOCK(addr); + + /* + * Try IPv4, then IPv6. In order to handle the extended format + * for IPv6 scoped addresses (address%scope_ID), we'll use a local + * working buffer of 128 bytes. The length is an ad-hoc value, but + * should be enough for this purpose; the buffer can contain a string + * of at least 80 bytes for scope_ID in addition to any IPv6 numeric + * addresses (up to 46 bytes), the delimiter character and the + * terminating NULL character. + */ + if (AF_UNSPEC == af || AF_INET == af) + if (inet_pton(AF_INET, host, &in4) == 1) { + AF(addr) = AF_INET; + SET_ADDR4N(addr, in4.s_addr); + + return TRUE; + } + + if (AF_UNSPEC == af || AF_INET6 == af) + if (sizeof(tmpbuf) > strlen(host)) { + if ('[' == host[0]) { + strlcpy(tmpbuf, &host[1], sizeof(tmpbuf)); + pch = strchr(tmpbuf, ']'); + if (pch != NULL) + *pch = '\0'; + } else { + strlcpy(tmpbuf, host, sizeof(tmpbuf)); + } + ZERO(hints); + hints.ai_family = AF_INET6; + hints.ai_flags |= AI_NUMERICHOST; + if (getaddrinfo(tmpbuf, NULL, &hints, &result) == 0) { + AF(addr) = AF_INET6; + resaddr6 = (struct sockaddr_in6 *)result->ai_addr; + SET_ADDR6N(addr, resaddr6->sin6_addr); + SET_SCOPE(addr, resaddr6->sin6_scope_id); + + freeaddrinfo(result); + return TRUE; + } + } + /* + * If we got here it was not an IP address + */ + return FALSE; +} + + /* * interface list enumerator - visitor pattern */ void -interface_enumerate(interface_receiver_t receiver, void *data) +interface_enumerate( + interface_receiver_t receiver, + void * data + ) { interface_info_t ifi; - struct interface *interf; ifi.action = IFS_EXISTS; - - for (interf = ISC_LIST_HEAD(inter_list); - interf != NULL; - interf = ISC_LIST_NEXT(interf, link)) { - ifi.interface = interf; - receiver(data, &ifi); - } + for (ifi.ep = ep_list; ifi.ep != NULL; ifi.ep = ifi.ep->elink) + (*receiver)(data, &ifi); } /* * do standard initialization of interface structure */ static void -init_interface(struct interface *interface) +init_interface( + endpt *ep + ) { - memset((char *)interface, 0, sizeof(struct interface)); - ISC_LINK_INIT(interface, link); - ISC_LIST_INIT(interface->peers); - interface->fd = INVALID_SOCKET; - interface->bfd = INVALID_SOCKET; - interface->num_mcast = 0; - interface->received = 0; - interface->sent = 0; - interface->notsent = 0; - interface->peercnt = 0; - interface->phase = sys_interphase; + ZERO(*ep); + ep->fd = INVALID_SOCKET; + ep->bfd = INVALID_SOCKET; + ep->phase = sys_interphase; } + /* * create new interface structure initialize from * template structure or via standard initialization * function */ static struct interface * -new_interface(struct interface *interface) +new_interface( + struct interface *interface + ) { - static u_int sys_ifnum = 0; + struct interface * iface; - struct interface *iface = (struct interface *)emalloc(sizeof(struct interface)); + iface = emalloc(sizeof(*iface)); - if (interface != NULL) - { - memcpy((char*)iface, (char*)interface, sizeof(*interface)); - } - else - { + if (NULL == interface) init_interface(iface); - } + else /* use the template */ + memcpy(iface, interface, sizeof(*iface)); - iface->ifnum = sys_ifnum++; /* count every new instance of an interface in the system */ + /* count every new instance of an interface in the system */ + iface->ifnum = sys_ifnum++; iface->starttime = current_time; return iface; } + /* * return interface storage into free memory pool */ -static void -delete_interface(struct interface *interface) +static inline void +delete_interface( + endpt *ep + ) { - free(interface); + free(ep); } + /* * link interface into list of known interfaces */ static void -add_interface(struct interface *interface) +add_interface( + endpt * ep + ) { - static struct interface *listhead = NULL; - + endpt ** pmclisthead; + endpt * scan; + endpt * scan_next; + endpt * unlinked; + sockaddr_u * addr; + int ep_local; + int scan_local; + int same_subnet; + int ep_univ_iid; /* iface ID from MAC address */ + int scan_univ_iid; /* see RFC 4291 */ + int ep_privacy; /* random local iface ID */ + int scan_privacy; /* see RFC 4941 */ + int rc; + + /* Calculate the refid */ + ep->addr_refid = addr2refid(&ep->sin); + /* link at tail so ntpdc -c ifstats index increases each row */ + LINK_TAIL_SLIST(ep_list, ep, elink, endpt); + ninterfaces++; +#ifdef MCAST + /* the rest is for enabled multicast-capable addresses only */ + if (ep->ignore_packets || !(INT_MULTICAST & ep->flags) || + INT_LOOPBACK & ep->flags) + return; +# ifndef INCLUDE_IPV6_MULTICAST_SUPPORT + if (AF_INET6 == ep->family) + return; +# endif + pmclisthead = (AF_INET == ep->family) + ? &mc4_list + : &mc6_list; + + if (AF_INET6 == ep->family) { + ep_local = + IN6_IS_ADDR_LINKLOCAL(PSOCK_ADDR6(&ep->sin)) || + IN6_IS_ADDR_SITELOCAL(PSOCK_ADDR6(&ep->sin)); + ep_univ_iid = IS_IID_UNIV(&ep->sin); + ep_privacy = !!(INT_PRIVACY & ep->flags); + } else { + ep_local = FALSE; + ep_univ_iid = FALSE; + ep_privacy = FALSE; + } + DPRINTF(4, ("add_interface mcast-capable %s%s%s%s\n", + stoa(&ep->sin), + (ep_local) ? " link/scope-local" : "", + (ep_univ_iid) ? " univ-IID" : "", + (ep_privacy) ? " privacy" : "")); /* - * For ntpd, the first few interfaces (wildcard, localhost) - * will never be removed. This means inter_list.head is - * unchanging once initialized. Take advantage of that to - * watch for changes and catch corruption earlier. This - * helped track down corruption caused by using FD_SET with - * a descriptor numerically larger than FD_SETSIZE. + * If we have multiple local addresses on the same network + * interface, and some are link- or site-local, do not multicast + * out from the link-/site-local addresses by default, to avoid + * duplicate manycastclient associations between v6 peers using + * link-local and global addresses. link-local can still be + * chosen using "nic ignore myv6globalprefix::/64". + * Similarly, if we have multiple global addresses from the same + * prefix on the same network interface, multicast from one, + * preferring EUI-64, then static, then least RFC 4941 privacy + * addresses. */ - if (NULL == listhead) - listhead = inter_list.head; - - if (listhead != inter_list.head) { - msyslog(LOG_ERR, "add_interface inter_list.head corrupted: was %p now %p", - listhead, inter_list.head); - exit(1); + for (scan = *pmclisthead; scan != NULL; scan = scan_next) { + scan_next = scan->mclink; + if (ep->family != scan->family) + continue; + if (strcmp(ep->name, scan->name)) + continue; + same_subnet = addr_samesubnet(&ep->sin, &ep->mask, + &scan->sin, &scan->mask); + if (AF_INET6 == ep->family) { + addr = &scan->sin; + scan_local = + IN6_IS_ADDR_LINKLOCAL(PSOCK_ADDR6(addr)) || + IN6_IS_ADDR_SITELOCAL(PSOCK_ADDR6(addr)); + scan_univ_iid = IS_IID_UNIV(addr); + scan_privacy = !!(INT_PRIVACY & scan->flags); + } else { + scan_local = FALSE; + scan_univ_iid = FALSE; + scan_privacy = FALSE; + } + DPRINTF(4, ("add_interface mcast-capable scan %s%s%s%s\n", + stoa(&scan->sin), + (scan_local) ? " link/scope-local" : "", + (scan_univ_iid) ? " univ-IID" : "", + (scan_privacy) ? " privacy" : "")); + if ((ep_local && !scan_local) || (same_subnet && + ((ep_privacy && !scan_privacy) || + (!ep_univ_iid && scan_univ_iid)))) { + DPRINTF(4, ("did not add %s to %s of IPv6 multicast-capable list which already has %s\n", + stoa(&ep->sin), + (ep_local) + ? "tail" + : "head", + stoa(&scan->sin))); + return; + } + if ((scan_local && !ep_local) || (same_subnet && + ((scan_privacy && !ep_privacy) || + (!scan_univ_iid && ep_univ_iid)))) { + UNLINK_SLIST(unlinked, *pmclisthead, + scan, mclink, endpt); + DPRINTF(4, ("%s %s from IPv6 multicast-capable list to add %s\n", + (unlinked != scan) + ? "Failed to remove" + : "removed", + stoa(&scan->sin), stoa(&ep->sin))); + } } /* - * Calculate the address hash + * Add link/site local at the tail of the multicast- + * capable unicast interfaces list, so that ntpd will + * send from global addresses before link-/site-local + * ones. */ - interface->addr_refid = addr2refid(&interface->sin); - - ISC_LIST_APPEND(inter_list, interface, link); - ninterfaces++; + if (ep_local) + LINK_TAIL_SLIST(*pmclisthead, ep, mclink, endpt); + else + LINK_SLIST(*pmclisthead, ep, mclink); + DPRINTF(4, ("added %s to %s of IPv%s multicast-capable unicast local address list\n", + stoa(&ep->sin), + (ep_local) + ? "tail" + : "head", + (AF_INET == ep->family) + ? "4" + : "6")); + + if (INVALID_SOCKET == ep->fd) + return; + + /* + * select the local address from which to send to multicast. + */ + switch (AF(&ep->sin)) { + + case AF_INET : + rc = setsockopt(ep->fd, IPPROTO_IP, + IP_MULTICAST_IF, + (void *)&NSRCADR(&ep->sin), + sizeof(NSRCADR(&ep->sin))); + if (rc) + msyslog(LOG_ERR, + "setsockopt IP_MULTICAST_IF %s fails: %m", + stoa(&ep->sin)); + break; + +# ifdef INCLUDE_IPV6_MULTICAST_SUPPORT + case AF_INET6 : + rc = setsockopt(ep->fd, IPPROTO_IPV6, + IPV6_MULTICAST_IF, + (void *)&ep->ifindex, + sizeof(ep->ifindex)); + /* do not complain if bound addr scope is ifindex */ + if (rc && ep->ifindex != SCOPE(&ep->sin)) + msyslog(LOG_ERR, + "setsockopt IPV6_MULTICAST_IF %u for %s fails: %m", + ep->ifindex, stoa(&ep->sin)); + break; +# endif + } +#endif /* MCAST */ } + /* * remove interface from known interface list and clean up * associated resources */ static void -remove_interface(struct interface *interface) +remove_interface( + endpt * ep + ) { - struct sockaddr_storage resmask; - - ISC_LIST_UNLINK_TYPE(inter_list, interface, link, struct interface); - - delete_interface_from_list(interface); - - if (interface->fd != INVALID_SOCKET) - { - msyslog(LOG_INFO, "Deleting interface #%d %s, %s#%d, interface stats: received=%ld, sent=%ld, dropped=%ld, active_time=%ld secs", - interface->ifnum, - interface->name, - stoa((&interface->sin)), - NTP_PORT, /* XXX should extract port from sin structure */ - interface->received, - interface->sent, - interface->notsent, - current_time - interface->starttime); - - close_and_delete_fd_from_list(interface->fd); + endpt * unlinked; + endpt ** pmclisthead; + sockaddr_u resmask; + + UNLINK_SLIST(unlinked, ep_list, ep, elink, endpt); + if (!ep->ignore_packets && INT_MULTICAST & ep->flags) { + pmclisthead = (AF_INET == ep->family) + ? &mc4_list + : &mc6_list; + UNLINK_SLIST(unlinked, *pmclisthead, ep, mclink, endpt); + DPRINTF(4, ("%s %s IPv%s multicast-capable unicast local address list\n", + stoa(&ep->sin), + (unlinked != NULL) + ? "removed from" + : "not found on", + (AF_INET == ep->family) + ? "4" + : "6")); } - - if (interface->bfd != INVALID_SOCKET) - { - msyslog(LOG_INFO, "Deleting interface #%d %s, broadcast address %s#%d", - interface->ifnum, - interface->name, - stoa((&interface->bcast)), - (u_short) NTP_PORT); /* XXX extract port from sin structure */ - close_and_delete_fd_from_list(interface->bfd); + delete_interface_from_list(ep); + + if (ep->fd != INVALID_SOCKET) { + msyslog(LOG_INFO, + "Deleting interface #%d %s, %s#%d, interface stats: received=%ld, sent=%ld, dropped=%ld, active_time=%ld secs", + ep->ifnum, + ep->name, + stoa(&ep->sin), + SRCPORT(&ep->sin), + ep->received, + ep->sent, + ep->notsent, + current_time - ep->starttime); + close_and_delete_fd_from_list(ep->fd); + ep->fd = INVALID_SOCKET; + } + + if (ep->bfd != INVALID_SOCKET) { + msyslog(LOG_INFO, + "stop listening for broadcasts to %s on interface #%d %s", + stoa(&ep->bcast), ep->ifnum, ep->name); + close_and_delete_fd_from_list(ep->bfd); + ep->bfd = INVALID_SOCKET; + ep->flags &= ~INT_BCASTOPEN; } ninterfaces--; - ntp_monclearinterface(interface); + mon_clearinterface(ep); /* remove restrict interface entry */ - - /* - * Blacklist bound interface address - */ - SET_HOSTMASK(&resmask, interface->sin.ss_family); - hack_restrict(RESTRICT_REMOVEIF, &interface->sin, &resmask, - RESM_NTPONLY|RESM_INTERFACE, RES_IGNORE); + SET_HOSTMASK(&resmask, AF(&ep->sin)); + hack_restrict(RESTRICT_REMOVEIF, &ep->sin, &resmask, + RESM_NTPONLY | RESM_INTERFACE, RES_IGNORE, 0); } + static void -list_if_listening(struct interface *interface, u_short port) +log_listen_address( + endpt * ep + ) { - msyslog(LOG_INFO, "Listening on interface #%d %s, %s#%d %s", - interface->ifnum, - interface->name, - stoa((&interface->sin)), - ntohs( (u_short) port), - (interface->ignore_packets == ISC_FALSE) ? - "Enabled": "Disabled"); + msyslog(LOG_INFO, "%s on %d %s %s", + (ep->ignore_packets) + ? "Listen and drop" + : "Listen normally", + ep->ifnum, + ep->name, + sptoa(&ep->sin)); } + static void -create_wildcards(u_short port) { - isc_boolean_t okipv4 = ISC_TRUE; +create_wildcards( + u_short port + ) +{ + int v4wild; +#ifdef INCLUDE_IPV6_SUPPORT + int v6wild; +#endif + sockaddr_u wildaddr; + nic_rule_action action; + struct interface * wildif; + /* - * create pseudo-interface with wildcard IPv4 address + * silence "potentially uninitialized" warnings from VC9 + * failing to follow the logic. Ideally action could remain + * uninitialized, and the memset be the first statement under + * the first if (v4wild). */ -#ifdef IPV6_V6ONLY - if(isc_net_probeipv4() != ISC_R_SUCCESS) - okipv4 = ISC_FALSE; -#endif - - if(okipv4 == ISC_TRUE) { - struct interface *interface = new_interface(NULL); - - interface->family = AF_INET; - interface->sin.ss_family = AF_INET; - ((struct sockaddr_in*)&interface->sin)->sin_addr.s_addr = htonl(INADDR_ANY); - ((struct sockaddr_in*)&interface->sin)->sin_port = port; - (void) strncpy(interface->name, "wildcard", sizeof(interface->name)); - interface->mask.ss_family = AF_INET; - ((struct sockaddr_in*)&interface->mask)->sin_addr.s_addr = htonl(~(u_int32)0); - interface->flags = INT_BROADCAST | INT_UP | INT_WILDCARD; - interface->ignore_packets = ISC_TRUE; -#if defined(MCAST) - /* - * enable possible multicast reception on the broadcast socket - */ - interface->bcast.ss_family = AF_INET; - ((struct sockaddr_in*)&interface->bcast)->sin_port = port; - ((struct sockaddr_in*)&interface->bcast)->sin_addr.s_addr = htonl(INADDR_ANY); -#endif /* MCAST */ - interface->fd = open_socket(&interface->sin, - interface->flags, 1, interface); - - if (interface->fd != INVALID_SOCKET) { - wildipv4 = interface; - any_interface = interface; - - add_addr_to_list(&interface->sin, interface); - add_interface(interface); - list_if_listening(interface, port); + action = ACTION_LISTEN; + ZERO(wildaddr); + +#ifdef INCLUDE_IPV6_SUPPORT + /* + * create pseudo-interface with wildcard IPv6 address + */ + v6wild = ipv6_works; + if (v6wild) { + /* set wildaddr to the v6 wildcard address :: */ + ZERO(wildaddr); + AF(&wildaddr) = AF_INET6; + SET_ADDR6N(&wildaddr, in6addr_any); + SET_PORT(&wildaddr, port); + SET_SCOPE(&wildaddr, 0); + + /* check for interface/nic rules affecting the wildcard */ + action = interface_action(NULL, &wildaddr, 0); + v6wild = (ACTION_IGNORE != action); + } + if (v6wild) { + wildif = new_interface(NULL); + + strlcpy(wildif->name, "v6wildcard", sizeof(wildif->name)); + memcpy(&wildif->sin, &wildaddr, sizeof(wildif->sin)); + wildif->family = AF_INET6; + AF(&wildif->mask) = AF_INET6; + SET_ONESMASK(&wildif->mask); + + wildif->flags = INT_UP | INT_WILDCARD; + wildif->ignore_packets = (ACTION_DROP == action); + + wildif->fd = open_socket(&wildif->sin, 0, 1, wildif); + + if (wildif->fd != INVALID_SOCKET) { + wildipv6 = wildif; + any6_interface = wildif; + add_addr_to_list(&wildif->sin, wildif); + add_interface(wildif); + log_listen_address(wildif); } else { - msyslog(LOG_ERR, "unable to bind to wildcard socket address %s - another process may be running - EXITING", - stoa((&interface->sin))); + msyslog(LOG_ERR, + "unable to bind to wildcard address %s - another process may be running - EXITING", + stoa(&wildif->sin)); exit(1); } + DPRINT_INTERFACE(2, (wildif, "created ", "\n")); } +#endif -#ifdef INCLUDE_IPV6_SUPPORT /* - * create pseudo-interface with wildcard IPv6 address + * create pseudo-interface with wildcard IPv4 address */ - if (isc_net_probeipv6() == ISC_R_SUCCESS) { - struct interface *interface = new_interface(NULL); + v4wild = ipv4_works; + if (v4wild) { + /* set wildaddr to the v4 wildcard address 0.0.0.0 */ + AF(&wildaddr) = AF_INET; + SET_ADDR4N(&wildaddr, INADDR_ANY); + SET_PORT(&wildaddr, port); + + /* check for interface/nic rules affecting the wildcard */ + action = interface_action(NULL, &wildaddr, 0); + v4wild = (ACTION_IGNORE != action); + } + if (v4wild) { + wildif = new_interface(NULL); - interface->family = AF_INET6; - interface->sin.ss_family = AF_INET6; - ((struct sockaddr_in6*)&interface->sin)->sin6_addr = in6addr_any; - ((struct sockaddr_in6*)&interface->sin)->sin6_port = port; -# ifdef ISC_PLATFORM_HAVESCOPEID - ((struct sockaddr_in6*)&interface->sin)->sin6_scope_id = 0; -# endif - (void) strncpy(interface->name, "wildcard", sizeof(interface->name)); - interface->mask.ss_family = AF_INET6; - memset(&((struct sockaddr_in6*)&interface->mask)->sin6_addr.s6_addr, 0xff, sizeof(struct in6_addr)); - interface->flags = INT_UP | INT_WILDCARD; - interface->ignore_packets = ISC_TRUE; + strlcpy(wildif->name, "v4wildcard", sizeof(wildif->name)); + memcpy(&wildif->sin, &wildaddr, sizeof(wildif->sin)); + wildif->family = AF_INET; + AF(&wildif->mask) = AF_INET; + SET_ONESMASK(&wildif->mask); - interface->fd = open_socket(&interface->sin, - interface->flags, 1, interface); - - if (interface->fd != INVALID_SOCKET) { - wildipv6 = interface; - any6_interface = interface; - add_addr_to_list(&interface->sin, interface); - add_interface(interface); - list_if_listening(interface, port); + wildif->flags = INT_BROADCAST | INT_UP | INT_WILDCARD; + wildif->ignore_packets = (ACTION_DROP == action); +#if defined(MCAST) + /* + * enable multicast reception on the broadcast socket + */ + AF(&wildif->bcast) = AF_INET; + SET_ADDR4N(&wildif->bcast, INADDR_ANY); + SET_PORT(&wildif->bcast, port); +#endif /* MCAST */ + wildif->fd = open_socket(&wildif->sin, 0, 1, wildif); + + if (wildif->fd != INVALID_SOCKET) { + wildipv4 = wildif; + any_interface = wildif; + + add_addr_to_list(&wildif->sin, wildif); + add_interface(wildif); + log_listen_address(wildif); } else { - msyslog(LOG_ERR, "unable to bind to wildcard socket address %s - another process may be running - EXITING", - stoa((&interface->sin))); + msyslog(LOG_ERR, + "unable to bind to wildcard address %s - another process may be running - EXITING", + stoa(&wildif->sin)); exit(1); } + DPRINT_INTERFACE(2, (wildif, "created ", "\n")); } -#endif } -static isc_boolean_t -address_okay(struct interface *iface) { +/* + * add_nic_rule() -- insert a rule entry at the head of nic_rule_list. + */ +void +add_nic_rule( + nic_rule_match match_type, + const char * if_name, /* interface name or numeric address */ + int prefixlen, + nic_rule_action action + ) +{ + nic_rule * rule; + isc_boolean_t is_ip; + + rule = emalloc_zero(sizeof(*rule)); + rule->match_type = match_type; + rule->prefixlen = prefixlen; + rule->action = action; + + if (MATCH_IFNAME == match_type) { + NTP_REQUIRE(NULL != if_name); + rule->if_name = estrdup(if_name); + } else if (MATCH_IFADDR == match_type) { + NTP_REQUIRE(NULL != if_name); + /* set rule->addr */ + is_ip = is_ip_address(if_name, AF_UNSPEC, &rule->addr); + NTP_REQUIRE(is_ip); + } else + NTP_REQUIRE(NULL == if_name); + + LINK_SLIST(nic_rule_list, rule, next); +} + + +#ifdef DEBUG +static const char * +action_text( + nic_rule_action action + ) +{ + const char *t; + + switch (action) { + + default: + t = "ERROR"; /* quiet uninit warning */ + DPRINTF(1, ("fatal: unknown nic_rule_action %d\n", + action)); + NTP_ENSURE(0); + break; + + case ACTION_LISTEN: + t = "listen"; + break; + + case ACTION_IGNORE: + t = "ignore"; + break; + + case ACTION_DROP: + t = "drop"; + break; + } + + return t; +} +#endif /* DEBUG */ - DPRINTF(4, ("address_okay: listen Virtual: %d, IF name: %s\n", - listen_to_virtual_ips, iface->name)); + +static nic_rule_action +interface_action( + char * if_name, + sockaddr_u * if_addr, + u_int32 if_flags + ) +{ + nic_rule * rule; + int isloopback; + int iswildcard; + + DPRINTF(4, ("interface_action: interface %s ", + (if_name != NULL) ? if_name : "wildcard")); + + iswildcard = is_wildcard_addr(if_addr); + isloopback = !!(INT_LOOPBACK & if_flags); + + /* + * Find any matching NIC rule from --interface / -I or ntp.conf + * interface/nic rules. + */ + for (rule = nic_rule_list; rule != NULL; rule = rule->next) { + + switch (rule->match_type) { + + case MATCH_ALL: + /* loopback and wildcard excluded from "all" */ + if (isloopback || iswildcard) + break; + DPRINTF(4, ("nic all %s\n", + action_text(rule->action))); + return rule->action; + + case MATCH_IPV4: + if (IS_IPV4(if_addr)) { + DPRINTF(4, ("nic ipv4 %s\n", + action_text(rule->action))); + return rule->action; + } + break; + + case MATCH_IPV6: + if (IS_IPV6(if_addr)) { + DPRINTF(4, ("nic ipv6 %s\n", + action_text(rule->action))); + return rule->action; + } + break; + + case MATCH_WILDCARD: + if (iswildcard) { + DPRINTF(4, ("nic wildcard %s\n", + action_text(rule->action))); + return rule->action; + } + break; + + case MATCH_IFADDR: + if (rule->prefixlen != -1) { + if (addr_eqprefix(if_addr, &rule->addr, + rule->prefixlen)) { + + DPRINTF(4, ("subnet address match - %s\n", + action_text(rule->action))); + return rule->action; + } + } else + if (SOCK_EQ(if_addr, &rule->addr)) { + + DPRINTF(4, ("address match - %s\n", + action_text(rule->action))); + return rule->action; + } + break; + + case MATCH_IFNAME: + if (if_name != NULL +#if defined(HAVE_FNMATCH) && defined(FNM_CASEFOLD) + && !fnmatch(rule->if_name, if_name, FNM_CASEFOLD) +#else + && !strcasecmp(if_name, rule->if_name) +#endif + ) { + + DPRINTF(4, ("interface name match - %s\n", + action_text(rule->action))); + return rule->action; + } + break; + } + } /* - * Always allow the loopback + * Unless explicitly disabled such as with "nic ignore ::1" + * listen on loopback addresses. Since ntpq and ntpdc query + * "localhost" by default, which typically resolves to ::1 and + * 127.0.0.1, it's useful to default to listening on both. */ - if((iface->flags & INT_LOOPBACK) != 0) { - DPRINTF(4, ("address_okay: loopback - OK\n")); - return (ISC_TRUE); + if (isloopback) { + DPRINTF(4, ("default loopback listen\n")); + return ACTION_LISTEN; } /* - * Check if the interface is specified + * Treat wildcard addresses specially. If there is no explicit + * "nic ... wildcard" or "nic ... 0.0.0.0" or "nic ... ::" rule + * default to drop. */ - if (specific_interface != NULL) { - if (strcasecmp(iface->name, specific_interface) == 0) { - DPRINTF(4, ("address_okay: specific interface name matched - OK\n")); - return (ISC_TRUE); - } else { - DPRINTF(4, ("address_okay: specific interface name NOT matched - FAIL\n")); - return (ISC_FALSE); - } + if (iswildcard) { + DPRINTF(4, ("default wildcard drop\n")); + return ACTION_DROP; } - else { - if (listen_to_virtual_ips == 0 && - (strchr(iface->name, (int)':') != NULL)) { - DPRINTF(4, ("address_okay: virtual ip/alias - FAIL\n")); - return (ISC_FALSE); - } + + /* + * Check for "virtual IP" (colon in the interface name) after + * the rules so that "ntpd --interface eth0:1 -novirtualips" + * does indeed listen on eth0:1's addresses. + */ + if (!listen_to_virtual_ips && if_name != NULL + && (strchr(if_name, ':') != NULL)) { + + DPRINTF(4, ("virtual ip - ignore\n")); + return ACTION_IGNORE; + } + + /* + * If there are no --interface/-I command-line options and no + * interface/nic rules in ntp.conf, the default action is to + * listen. In the presence of rules from either, the default + * is to ignore. This implements ntpd's traditional listen- + * every default with no interface listen configuration, and + * ensures a single -I eth0 or "nic listen eth0" means do not + * listen on any other addresses. + */ + if (NULL == nic_rule_list) { + DPRINTF(4, ("default listen\n")); + return ACTION_LISTEN; } - DPRINTF(4, ("address_okay: OK\n")); - return (ISC_TRUE); + DPRINTF(4, ("implicit ignore\n")); + return ACTION_IGNORE; } + static void -convert_isc_if(isc_interface_t *isc_if, struct interface *itf, u_short port) +convert_isc_if( + isc_interface_t *isc_if, + endpt *itf, + u_short port + ) { - itf->scopeid = 0; - itf->family = (short) isc_if->af; - strcpy(itf->name, isc_if->name); - - if(isc_if->af == AF_INET) { - itf->sin.ss_family = (u_short) isc_if->af; - memcpy(&(((struct sockaddr_in*)&itf->sin)->sin_addr), - &(isc_if->address.type.in), - sizeof(struct in_addr)); - ((struct sockaddr_in*)&itf->sin)->sin_port = port; - - if((isc_if->flags & INTERFACE_F_BROADCAST) != 0) { + const u_char v6loop[16] = {0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 1}; + + strlcpy(itf->name, isc_if->name, sizeof(itf->name)); + itf->ifindex = isc_if->ifindex; + itf->family = (u_short)isc_if->af; + AF(&itf->sin) = itf->family; + AF(&itf->mask) = itf->family; + AF(&itf->bcast) = itf->family; + SET_PORT(&itf->sin, port); + SET_PORT(&itf->mask, port); + SET_PORT(&itf->bcast, port); + + if (IS_IPV4(&itf->sin)) { + NSRCADR(&itf->sin) = isc_if->address.type.in.s_addr; + NSRCADR(&itf->mask) = isc_if->netmask.type.in.s_addr; + + if (isc_if->flags & INTERFACE_F_BROADCAST) { itf->flags |= INT_BROADCAST; - itf->bcast.ss_family = itf->sin.ss_family; - memcpy(&(((struct sockaddr_in*)&itf->bcast)->sin_addr), - &(isc_if->broadcast.type.in), - sizeof(struct in_addr)); - ((struct sockaddr_in*)&itf->bcast)->sin_port = port; + NSRCADR(&itf->bcast) = + isc_if->broadcast.type.in.s_addr; } - - itf->mask.ss_family = itf->sin.ss_family; - memcpy(&(((struct sockaddr_in*)&itf->mask)->sin_addr), - &(isc_if->netmask.type.in), - sizeof(struct in_addr)); - ((struct sockaddr_in*)&itf->mask)->sin_port = port; } #ifdef INCLUDE_IPV6_SUPPORT - else if (isc_if->af == AF_INET6) { - itf->sin.ss_family = (u_short) isc_if->af; - memcpy(&(((struct sockaddr_in6 *)&itf->sin)->sin6_addr), - &(isc_if->address.type.in6), - sizeof(((struct sockaddr_in6 *)&itf->sin)->sin6_addr)); - ((struct sockaddr_in6 *)&itf->sin)->sin6_port = port; - -#ifdef ISC_PLATFORM_HAVESCOPEID - ((struct sockaddr_in6 *)&itf->sin)->sin6_scope_id = isc_netaddr_getzone(&isc_if->address); - itf->scopeid = isc_netaddr_getzone(&isc_if->address); -#endif - itf->mask.ss_family = itf->sin.ss_family; - memcpy(&(((struct sockaddr_in6 *)&itf->mask)->sin6_addr), - &(isc_if->netmask.type.in6), - sizeof(struct in6_addr)); - ((struct sockaddr_in6 *)&itf->mask)->sin6_port = port; - /* Copy the interface index */ - itf->ifindex = isc_if->ifindex; + else if (IS_IPV6(&itf->sin)) { + SET_ADDR6N(&itf->sin, isc_if->address.type.in6); + SET_ADDR6N(&itf->mask, isc_if->netmask.type.in6); + + SET_SCOPE(&itf->sin, isc_if->address.zone); } #endif /* INCLUDE_IPV6_SUPPORT */ /* Process the rest of the flags */ - if((isc_if->flags & INTERFACE_F_UP) != 0) - itf->flags |= INT_UP; - if((isc_if->flags & INTERFACE_F_LOOPBACK) != 0) - itf->flags |= INT_LOOPBACK; - if((isc_if->flags & INTERFACE_F_POINTTOPOINT) != 0) - itf->flags |= INT_PPP; - if((isc_if->flags & INTERFACE_F_MULTICAST) != 0) - itf->flags |= INT_MULTICAST; + itf->flags |= + ((INTERFACE_F_UP & isc_if->flags) + ? INT_UP : 0) + | ((INTERFACE_F_LOOPBACK & isc_if->flags) + ? INT_LOOPBACK : 0) + | ((INTERFACE_F_POINTTOPOINT & isc_if->flags) + ? INT_PPP : 0) + | ((INTERFACE_F_MULTICAST & isc_if->flags) + ? INT_MULTICAST : 0) + | ((INTERFACE_F_PRIVACY & isc_if->flags) + ? INT_PRIVACY : 0) + ; + /* + * Clear the loopback flag if the address is not localhost. + * http://bugs.ntp.org/1683 + */ + if (INT_LOOPBACK & itf->flags) { + if (AF_INET == itf->family) { + if (127 != (SRCADR(&itf->sin) >> 24)) + itf->flags &= ~INT_LOOPBACK; + } else { + if (memcmp(v6loop, NSRCADR6(&itf->sin), + sizeof(NSRCADR6(&itf->sin)))) + itf->flags &= ~INT_LOOPBACK; + } + } } + /* * refresh_interface * @@ -1057,27 +1525,34 @@ convert_isc_if(isc_interface_t *isc_if, struct interface *itf, u_short port) * the socket. */ static int -refresh_interface(struct interface * interface) +refresh_interface( + struct interface * interface + ) { #ifdef OS_MISSES_SPECIFIC_ROUTE_UPDATES - if (interface->fd != INVALID_SOCKET) - { + if (interface->fd != INVALID_SOCKET) { + int bcast = (interface->flags & INT_BCASTXMIT) != 0; + /* as we forcibly close() the socket remove the + broadcast permission indication */ + if (bcast) + socket_broadcast_disable(interface, &interface->sin); + close_and_delete_fd_from_list(interface->fd); + + /* create new socket picking up a new first hop binding + at connect() time */ interface->fd = open_socket(&interface->sin, - interface->flags, 0, interface); + bcast, 0, interface); /* - * reset TTL indication so TTL is is set again + * reset TTL indication so TTL is is set again * next time around */ interface->last_ttl = 0; - return interface->fd != INVALID_SOCKET; - } - else - { + return (interface->fd != INVALID_SOCKET); + } else return 0; /* invalid sockets are not refreshable */ - } #else /* !OS_MISSES_SPECIFIC_ROUTE_UPDATES */ - return interface->fd != INVALID_SOCKET; + return (interface->fd != INVALID_SOCKET); #endif /* !OS_MISSES_SPECIFIC_ROUTE_UPDATES */ } @@ -1085,113 +1560,125 @@ refresh_interface(struct interface * interface) * interface_update - externally callable update function */ void -interface_update(interface_receiver_t receiver, void *data) +interface_update( + interface_receiver_t receiver, + void * data) { - if (!disable_dynamic_updates) { - int new_interface_found; + int new_interface_found; - BLOCKIO(); - new_interface_found = update_interfaces(htons(NTP_PORT), receiver, data); - UNBLOCKIO(); + if (disable_dynamic_updates) + return; + + BLOCKIO(); + new_interface_found = update_interfaces(NTP_PORT, receiver, data); + UNBLOCKIO(); + + if (!new_interface_found) + return; - if (new_interface_found) { #ifdef DEBUG - msyslog(LOG_DEBUG, "new interface(s) found: waking up resolver"); + msyslog(LOG_DEBUG, "new interface(s) found: waking up resolver"); #endif -#ifdef SYS_WINNT - /* wake up the resolver thread */ - if (ResolverEventHandle != NULL) - SetEvent(ResolverEventHandle); -#else - /* write any single byte to the pipe to wake up the resolver process */ - write( resolver_pipe_fd[1], &new_interface_found, 1 ); -#endif - } - } + interrupt_worker_sleep(); } + /* - * find out if a given interface structure contains - * a wildcard address + * sau_from_netaddr() - convert network address on-wire formats. + * Convert from libisc's isc_netaddr_t to NTP's sockaddr_u */ +void +sau_from_netaddr( + sockaddr_u *psau, + const isc_netaddr_t *pna + ) +{ + ZERO_SOCK(psau); + AF(psau) = (u_short)pna->family; + switch (pna->family) { + + case AF_INET: + memcpy(&psau->sa4.sin_addr, &pna->type.in, + sizeof(psau->sa4.sin_addr)); + break; + + case AF_INET6: + memcpy(&psau->sa6.sin6_addr, &pna->type.in6, + sizeof(psau->sa6.sin6_addr)); + break; + } +} + + static int -is_wildcard_addr(struct sockaddr_storage *sas) +is_wildcard_addr( + const sockaddr_u *psau + ) { - if (sas->ss_family == AF_INET && - ((struct sockaddr_in*)sas)->sin_addr.s_addr == htonl(INADDR_ANY)) + if (IS_IPV4(psau) && !NSRCADR(psau)) return 1; #ifdef INCLUDE_IPV6_SUPPORT - if (sas->ss_family == AF_INET6 && - memcmp(&((struct sockaddr_in6*)sas)->sin6_addr, &in6addr_any, - sizeof(in6addr_any)) == 0) + if (IS_IPV6(psau) && S_ADDR6_EQ(psau, &in6addr_any)) return 1; #endif return 0; } + #ifdef OS_NEEDS_REUSEADDR_FOR_IFADDRBIND /* * enable/disable re-use of wildcard address socket */ static void -set_wildcard_reuse(int family, int on) +set_wildcard_reuse( + u_short family, + int on + ) { - int onvalue = 1; - int offvalue = 0; - int *onoff; + struct interface *any; SOCKET fd = INVALID_SOCKET; - onoff = on ? &onvalue : &offvalue; - - switch (family) { - case AF_INET: - if (any_interface) { - fd = any_interface->fd; - } - break; - -#ifdef INCLUDE_IPV6_SUPPORT - case AF_INET6: - if (any6_interface) { - fd = any6_interface->fd; - } - break; -#endif /* !INCLUDE_IPV6_SUPPORT */ - } + any = ANY_INTERFACE_BYFAM(family); + if (any != NULL) + fd = any->fd; if (fd != INVALID_SOCKET) { - if (setsockopt(fd, SOL_SOCKET, - SO_REUSEADDR, (char *)onoff, - sizeof(*onoff))) { - netsyslog(LOG_ERR, "set_wildcard_reuse: setsockopt(SO_REUSEADDR, %s) failed: %m", *onoff ? "on" : "off"); - } - DPRINTF(4, ("set SO_REUSEADDR to %s on %s\n", *onoff ? "ON" : "OFF", - stoa((family == AF_INET) ? - &any_interface->sin : &any6_interface->sin))); + if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, + (char *)&on, sizeof(on))) + msyslog(LOG_ERR, + "set_wildcard_reuse: setsockopt(SO_REUSEADDR, %s) failed: %m", + on ? "on" : "off"); + + DPRINTF(4, ("set SO_REUSEADDR to %s on %s\n", + on ? "on" : "off", + stoa(&any->sin))); } } #endif /* OS_NEEDS_REUSEADDR_FOR_IFADDRBIND */ -#ifdef INCLUDE_IPV6_SUPPORT + static isc_boolean_t -is_not_bindable(struct sockaddr *sa, char *name) +check_flags6( + sockaddr_u *psau, + const char *name, + u_int32 flags6 + ) { -#if defined(SIOCGIFAFLAG_IN6) && \ - (defined(IN6_IFF_ANYCAST) || defined(IN6_IFF_NOTREADY)) +#if defined(INCLUDE_IPV6_SUPPORT) && defined(SIOCGIFAFLAG_IN6) && \ + (defined(IN6_IFF_ANYCAST) || defined(IN6_IFF_NOTREADY)) struct in6_ifreq ifr6; int fd; - u_int32_t flags6, exclude = 0; + u_int32 exclude = 0; - if (sa->sa_family != AF_INET6) + if (psau->sa.sa_family != AF_INET6) return ISC_FALSE; if ((fd = socket(AF_INET6, SOCK_DGRAM, 0)) < 0) return ISC_FALSE; - memset(&ifr6, 0, sizeof(ifr6)); - memcpy(&ifr6.ifr_addr, (struct sockaddr_in6 *)sa, - sizeof(struct sockaddr_in6)); - strlcpy(ifr6.ifr_name, name, IF_NAMESIZE); + ZERO(ifr6); + memcpy(&ifr6.ifr_addr, &psau->sa6, sizeof(ifr6.ifr_addr)); + strlcpy(ifr6.ifr_name, name, sizeof(ifr6.ifr_name)); if (ioctl(fd, SIOCGIFAFLAG_IN6, &ifr6) < 0) { close(fd); return ISC_FALSE; @@ -1206,10 +1693,43 @@ is_not_bindable(struct sockaddr *sa, char *name) #endif /* !IN6_IFF_NOTREADY */ if ((flags6 & exclude) != 0) return ISC_TRUE; -#endif /* !SIOCGIFAFLAG_IN6 || !(IN6_IFF_ANYCAST && IN6_IFF_NOTREADY) */ +#endif /* INCLUDE_IPV6_SUPPORT && SIOCGIFAFLAG_IN6 && (IN6_IFF_ANYCAST && IN6_IFF_NOTREADY) */ return ISC_FALSE; } -#endif /* !INCLUDE_IPV6_SUPPORT */ + +static isc_boolean_t +is_not_bindable( + sockaddr_u *psau, + const char *name + ) +{ +#ifdef IN6_IFF_ANYCAST + return check_flags6(psau, name, IN6_IFF_ANYCAST); +#else + return ISC_FALSE; +#endif +} + +static isc_boolean_t +is_valid( + sockaddr_u *psau, + const char *name + ) +{ + u_int32 flags6; + + flags6 = 0; +#ifdef IN6_IFF_DEPARTED + flags6 |= IN6_IFF_DEPARTED; +#endif +#ifdef IN6_IFF_DETACHED + flags6 |= IN6_IFF_DETACHED; +#endif +#ifdef IN6_IFF_TENTATIVE + flags6 |= IN6_IFF_TENTATIVE; +#endif + return check_flags6(psau, name, flags6) ? ISC_FALSE : ISC_TRUE; +} /* * update_interface strategy @@ -1219,17 +1739,17 @@ is_not_bindable(struct sockaddr *sa, char *name) * Phase 1: * forall currently existing interfaces * if address is known: - * drop socket - rebind again + * drop socket - rebind again * * if address is NOT known: - * attempt to create a new interface entry + * attempt to create a new interface entry * * Phase 2: * forall currently known non MCAST and WILDCARD interfaces * if interface does not match configuration phase (not seen in phase 1): - * remove interface from known interface list - * forall peers associated with this interface - * disconnect peer from this interface + * remove interface from known interface list + * forall peers associated with this interface + * disconnect peer from this interface * * Phase 3: * attempt to re-assign interfaces to peers @@ -1238,67 +1758,47 @@ is_not_bindable(struct sockaddr *sa, char *name) static int update_interfaces( - u_short port, - interface_receiver_t receiver, - void *data + u_short port, + interface_receiver_t receiver, + void * data ) { - interface_info_t ifi; - isc_mem_t *mctx = NULL; - isc_interfaceiter_t *iter = NULL; - isc_boolean_t scan_ipv4 = ISC_FALSE; - isc_boolean_t scan_ipv6 = ISC_FALSE; - isc_result_t result; - int new_interface_found = 0; + isc_mem_t * mctx = (void *)-1; + interface_info_t ifi; + isc_interfaceiter_t * iter; + isc_result_t result; + isc_interface_t isc_if; + int new_interface_found; + unsigned int family; + endpt enumep; + endpt * ep; + endpt * next_ep; + + DPRINTF(3, ("update_interfaces(%d)\n", port)); - DPRINTF(3, ("update_interfaces(%d)\n", ntohs( (u_short) port))); - -#ifdef INCLUDE_IPV6_SUPPORT - if (isc_net_probeipv6() == ISC_R_SUCCESS) - scan_ipv6 = ISC_TRUE; -#if defined(DEBUG) - else - if (debug) - netsyslog(LOG_ERR, "no IPv6 interfaces found"); -#endif -#endif - if (isc_net_probeipv6() == ISC_R_SUCCESS) - scan_ipv6 = ISC_TRUE; -#if defined(ISC_PLATFORM_HAVEIPV6) && defined(DEBUG) - else - if (debug) - netsyslog(LOG_ERR, "no IPv6 interfaces found"); -#endif - - if (isc_net_probeipv4() == ISC_R_SUCCESS) - scan_ipv4 = ISC_TRUE; -#ifdef DEBUG - else - if(debug) - netsyslog(LOG_ERR, "no IPv4 interfaces found"); -#endif /* * phase one - scan interfaces * - create those that are not found * - update those that are found */ + new_interface_found = FALSE; + iter = NULL; result = isc_interfaceiter_create(mctx, &iter); if (result != ISC_R_SUCCESS) return 0; - sys_interphase ^= 0x1; /* toggle system phase for finding untouched (to be deleted) interfaces */ - + /* + * Toggle system interface scan phase to find untouched + * interfaces to be deleted. + */ + sys_interphase ^= 0x1; + for (result = isc_interfaceiter_first(iter); - result == ISC_R_SUCCESS; - result = isc_interfaceiter_next(iter)) - { - isc_interface_t isc_if; - unsigned int family; - struct interface interface; - struct interface *iface; - + ISC_R_SUCCESS == result; + result = isc_interfaceiter_next(iter)) { + result = isc_interfaceiter_current(iter, &isc_if); if (result != ISC_R_SUCCESS) @@ -1306,39 +1806,48 @@ update_interfaces( /* See if we have a valid family to use */ family = isc_if.address.family; - if (family != AF_INET && family != AF_INET6) + if (AF_INET != family && AF_INET6 != family) continue; - if (scan_ipv4 == ISC_FALSE && family == AF_INET) + if (AF_INET == family && !ipv4_works) continue; - if (scan_ipv6 == ISC_FALSE && family == AF_INET6) + if (AF_INET6 == family && !ipv6_works) continue; + /* create prototype */ + init_interface(&enumep); + + convert_isc_if(&isc_if, &enumep, port); + + DPRINT_INTERFACE(4, (&enumep, "examining ", "\n")); + /* - * create prototype + * Check if and how we are going to use the interface. */ - init_interface(&interface); + switch (interface_action(enumep.name, &enumep.sin, + enumep.flags)) { - convert_isc_if(&isc_if, &interface, port); + case ACTION_IGNORE: + DPRINTF(4, ("ignoring interface %s (%s) - by nic rules\n", + enumep.name, stoa(&enumep.sin))); + continue; - /* - * Check to see if we are going to use the interface - * If we don't use it we mark it to drop any packet - * received but we still must create the socket and - * bind to it. This prevents other apps binding to it - * and potentially causing problems with more than one - * process fiddling with the clock - */ - if (address_okay(&interface) == ISC_TRUE) { - interface.ignore_packets = ISC_FALSE; - } - else { - interface.ignore_packets = ISC_TRUE; - } + case ACTION_LISTEN: + DPRINTF(4, ("listen interface %s (%s) - by nic rules\n", + enumep.name, stoa(&enumep.sin))); + enumep.ignore_packets = ISC_FALSE; + break; - DPRINT_INTERFACE(4, (&interface, "examining ", "\n")); + case ACTION_DROP: + DPRINTF(4, ("drop on interface %s (%s) - by nic rules\n", + enumep.name, stoa(&enumep.sin))); + enumep.ignore_packets = ISC_TRUE; + break; + } - if (!(interface.flags & INT_UP)) { /* interfaces must be UP to be usable */ - DPRINTF(4, ("skipping interface %s (%s) - DOWN\n", interface.name, stoa(&interface.sin))); + /* interfaces must be UP to be usable */ + if (!(enumep.flags & INT_UP)) { + DPRINTF(4, ("skipping interface %s (%s) - DOWN\n", + enumep.name, stoa(&enumep.sin))); continue; } @@ -1347,62 +1856,119 @@ update_interfaces( * address - some dhcp clients produce that in the * wild */ - if (is_wildcard_addr(&interface.sin)) + if (is_wildcard_addr(&enumep.sin)) continue; -#ifdef INCLUDE_IPV6_SUPPORT - if (is_not_bindable((struct sockaddr *)&interface.sin, isc_if.name)) + if (is_not_bindable(&enumep.sin, isc_if.name)) continue; -#endif /* !INCLUDE_IPV6_SUPPORT */ /* - * map to local *address* in order - * to map all duplicate interfaces to an interface structure - * with the appropriate socket (our name space is - * (ip-address) - NOT (interface name, ip-address)) + * skip any address that is an invalid state to be used */ - iface = getinterface(&interface.sin, INT_WILDCARD); - - if (iface && refresh_interface(iface)) - { + if (!is_valid(&enumep.sin, isc_if.name)) + continue; + + /* + * map to local *address* in order to map all duplicate + * interfaces to an endpt structure with the appropriate + * socket. Our name space is (ip-address), NOT + * (interface name, ip-address). + */ + ep = getinterface(&enumep.sin, INT_WILDCARD); + + if (ep != NULL && refresh_interface(ep)) { /* - * found existing and up to date interface - mark present + * found existing and up to date interface - + * mark present. */ + if (ep->phase != sys_interphase) { + /* + * On a new round we reset the name so + * the interface name shows up again if + * this address is no longer shared. + * We reset ignore_packets from the + * new prototype to respect any runtime + * changes to the nic rules. + */ + strlcpy(ep->name, enumep.name, + sizeof(ep->name)); + ep->ignore_packets = + enumep.ignore_packets; + } else { + /* name collision - rename interface */ + strlcpy(ep->name, "*multiple*", + sizeof(ep->name)); + } + + DPRINT_INTERFACE(4, (ep, "updating ", + " present\n")); + + if (ep->ignore_packets != + enumep.ignore_packets) { + /* + * We have conflicting configurations + * for the interface address. This is + * caused by using -I <interfacename> + * for an interface that shares its + * address with other interfaces. We + * can not disambiguate incoming + * packets delivered to this socket + * without extra syscalls/features. + * These are not (commonly) available. + * Note this is a more unusual + * configuration where several + * interfaces share an address but + * filtering via interface name is + * attempted. We resolve the + * configuration conflict by disabling + * the processing of received packets. + * This leads to no service on the + * interface address where the conflict + * occurs. + */ + msyslog(LOG_ERR, + "WARNING: conflicting enable configuration for interfaces %s and %s for address %s - unsupported configuration - address DISABLED", + enumep.name, ep->name, + stoa(&enumep.sin)); + + ep->ignore_packets = ISC_TRUE; + } + + ep->phase = sys_interphase; - iface->phase = sys_interphase; - DPRINT_INTERFACE(4, (iface, "updating ", " present\n")); ifi.action = IFS_EXISTS; - ifi.interface = iface; - if (receiver) - receiver(data, &ifi); - } - else - { + ifi.ep = ep; + if (receiver != NULL) + (*receiver)(data, &ifi); + } else { /* - * this is new or refreshing failed - add to our interface list - * if refreshing failed we will delete the interface structure in - * phase 2 as the interface was not marked current. We can bind to - * the address as the refresh code already closed the offending socket + * This is new or refreshing failed - add to + * our interface list. If refreshing failed we + * will delete the interface structure in phase + * 2 as the interface was not marked current. + * We can bind to the address as the refresh + * code already closed the offending socket */ - - iface = create_interface(port, &interface); + ep = create_interface(port, &enumep); - if (iface) - { + if (ep != NULL) { ifi.action = IFS_CREATED; - ifi.interface = iface; - if (receiver) - receiver(data, &ifi); - - new_interface_found = 1; - - DPRINT_INTERFACE(3, (iface, "updating ", " new - created\n")); - } - else - { - DPRINT_INTERFACE(3, (&interface, "updating ", " new - creation FAILED")); - - msyslog(LOG_INFO, "failed to initialize interface for address %s", stoa(&interface.sin)); + ifi.ep = ep; + if (receiver != NULL) + (*receiver)(data, &ifi); + + new_interface_found = TRUE; + DPRINT_INTERFACE(3, + (ep, "updating ", + " new - created\n")); + } else { + DPRINT_INTERFACE(3, + (&enumep, "updating ", + " new - creation FAILED")); + + msyslog(LOG_INFO, + "failed to init interface for address %s", + stoa(&enumep.sin)); continue; } } @@ -1411,64 +1977,86 @@ update_interfaces( isc_interfaceiter_destroy(&iter); /* - * phase 2 - delete gone interfaces - reassigning peers to other interfaces + * phase 2 - delete gone interfaces - reassigning peers to + * other interfaces */ - { - struct interface *interf = ISC_LIST_HEAD(inter_list); + for (ep = ep_list; ep != NULL; ep = next_ep) { + next_ep = ep->elink; - while (interf != NULL) - { - struct interface *next = ISC_LIST_NEXT(interf, link); - - if (!(interf->flags & (INT_WILDCARD|INT_MCASTIF))) { - /* - * if phase does not match sys_phase this interface was not - * enumerated during interface scan - so it is gone and - * will be deleted here unless it is solely an MCAST/WILDCARD interface - */ - if (interf->phase != sys_interphase) { - struct peer *peer; - DPRINT_INTERFACE(3, (interf, "updating ", "GONE - deleting\n")); - remove_interface(interf); - - ifi.action = IFS_DELETED; - ifi.interface = interf; - if (receiver) - receiver(data, &ifi); - - peer = ISC_LIST_HEAD(interf->peers); - /* - * disconnect peer from deleted interface - */ - while (peer != NULL) { - struct peer *npeer = ISC_LIST_NEXT(peer, ilink); - - /* - * this one just lost it's interface - */ - set_peerdstadr(peer, NULL); - - peer = npeer; - } - - /* - * update globals in case we lose - * a loopback interface - */ - if (interf == loopback_interface) - loopback_interface = NULL; - - delete_interface(interf); - } - } - interf = next; - } + /* + * if phase does not match sys_phase this interface was + * not enumerated during the last interface scan - so it + * is gone and will be deleted here unless it did not + * originate from interface enumeration (INT_WILDCARD, + * INT_MCASTIF). + */ + if (((INT_WILDCARD | INT_MCASTIF) & ep->flags) || + ep->phase == sys_interphase) + continue; + + DPRINT_INTERFACE(3, (ep, "updating ", + "GONE - deleting\n")); + remove_interface(ep); + + ifi.action = IFS_DELETED; + ifi.ep = ep; + if (receiver != NULL) + (*receiver)(data, &ifi); + + /* disconnect peers from deleted endpt. */ + while (ep->peers != NULL) + set_peerdstadr(ep->peers, NULL); + + /* + * update globals in case we lose + * a loopback interface + */ + if (ep == loopback_interface) + loopback_interface = NULL; + + delete_interface(ep); } /* - * phase 3 - re-configure as the world has changed if necessary + * phase 3 - re-configure as the world has possibly changed + * + * never ever make this conditional again - it is needed to track + * routing updates. see bug #2506 */ refresh_all_peerinterfaces(); + + if (broadcast_client_enabled) + io_setbclient(); + + if (sys_bclient) + io_setbclient(); + + /* + * Check multicast interfaces and try to join multicast groups if + * not joined yet. + */ + for (ep = ep_list; ep != NULL; ep = ep->elink) { + remaddr_t *entry; + + if (!(INT_MCASTIF & ep->flags) || (INT_MCASTOPEN & ep->flags)) + continue; + + /* Find remote address that was linked to this interface */ + for (entry = remoteaddr_list; + entry != NULL; + entry = entry->link) { + if (entry->ep == ep) { + if (socket_multicast_enable(ep, &entry->addr)) { + msyslog(LOG_INFO, + "Joined %s socket to multicast group %s", + stoa(&ep->sin), + stoa(&entry->addr)); + } + break; + } + } + } + return new_interface_found; } @@ -1490,12 +2078,12 @@ create_sockets( FD_ZERO(&activefds); #endif - DPRINTF(2, ("create_sockets(%d)\n", ntohs( (u_short) port))); + DPRINTF(2, ("create_sockets(%d)\n", port)); create_wildcards(port); update_interfaces(port, NULL, NULL); - + /* * Now that we have opened all the sockets, turn off the reuse * flag for security. @@ -1513,87 +2101,137 @@ create_sockets( */ static struct interface * create_interface( - u_short port, - struct interface *iface - ) + u_short port, + struct interface * protot + ) { - struct sockaddr_storage resmask; - struct interface *interface; - - DPRINTF(2, ("create_interface(%s#%d)\n", stoa(&iface->sin), ntohs( (u_short) port))); + sockaddr_u resmask; + endpt * iface; +#if defined(MCAST) && defined(MULTICAST_NONEWSOCKET) + remaddr_t * entry; + remaddr_t * next_entry; +#endif + DPRINTF(2, ("create_interface(%s#%d)\n", stoa(&protot->sin), + port)); /* build an interface */ - interface = new_interface(iface); - + iface = new_interface(protot); + /* * create socket */ - interface->fd = open_socket(&interface->sin, - interface->flags, 0, interface); + iface->fd = open_socket(&iface->sin, 0, 0, iface); - if (interface->fd != INVALID_SOCKET) - list_if_listening(interface, port); + if (iface->fd != INVALID_SOCKET) + log_listen_address(iface); - if ((interface->flags & INT_BROADCAST) && - interface->bfd != INVALID_SOCKET) - msyslog(LOG_INFO, "Listening on broadcast address %s#%d", - stoa((&interface->bcast)), - ntohs( (u_short) port)); + if ((INT_BROADCAST & iface->flags) + && iface->bfd != INVALID_SOCKET) + msyslog(LOG_INFO, "Listening on broadcast address %s#%d", + stoa((&iface->bcast)), port); - if (interface->fd == INVALID_SOCKET && - interface->bfd == INVALID_SOCKET) { + if (INVALID_SOCKET == iface->fd + && INVALID_SOCKET == iface->bfd) { msyslog(LOG_ERR, "unable to create socket on %s (%d) for %s#%d", - interface->name, - interface->ifnum, - stoa((&interface->sin)), - ntohs( (u_short) port)); - delete_interface(interface); + iface->name, + iface->ifnum, + stoa((&iface->sin)), + port); + delete_interface(iface); return NULL; } - - /* - * Blacklist bound interface address + + /* + * Blacklist our own addresses, no use talking to ourself */ - - SET_HOSTMASK(&resmask, interface->sin.ss_family); - hack_restrict(RESTRICT_FLAGS, &interface->sin, &resmask, - RESM_NTPONLY|RESM_INTERFACE, RES_IGNORE); - + SET_HOSTMASK(&resmask, AF(&iface->sin)); + hack_restrict(RESTRICT_FLAGS, &iface->sin, &resmask, + RESM_NTPONLY | RESM_INTERFACE, RES_IGNORE, 0); + /* * set globals with the first found * loopback interface of the appropriate class */ - if ((loopback_interface == NULL) && - (interface->family == AF_INET) && - ((interface->flags & INT_LOOPBACK) != 0)) - { - loopback_interface = interface; - } + if (NULL == loopback_interface && AF_INET == iface->family + && (INT_LOOPBACK & iface->flags)) + loopback_interface = iface; /* * put into our interface list */ - add_addr_to_list(&interface->sin, interface); - add_interface(interface); + add_addr_to_list(&iface->sin, iface); + add_interface(iface); + +#if defined(MCAST) && defined(MULTICAST_NONEWSOCKET) + /* + * Join any previously-configured compatible multicast groups. + */ + if (INT_MULTICAST & iface->flags && + !((INT_LOOPBACK | INT_WILDCARD) & iface->flags) && + !iface->ignore_packets) { + for (entry = remoteaddr_list; + entry != NULL; + entry = next_entry) { + next_entry = entry->link; + if (AF(&iface->sin) != AF(&entry->addr) || + !IS_MCAST(&entry->addr)) + continue; + if (socket_multicast_enable(iface, + &entry->addr)) + msyslog(LOG_INFO, + "Joined %s socket to multicast group %s", + stoa(&iface->sin), + stoa(&entry->addr)); + else + msyslog(LOG_ERR, + "Failed to join %s socket to multicast group %s", + stoa(&iface->sin), + stoa(&entry->addr)); + } + } +#endif /* MCAST && MCAST_NONEWSOCKET */ - DPRINT_INTERFACE(2, (interface, "created ", "\n")); - return interface; + DPRINT_INTERFACE(2, (iface, "created ", "\n")); + return iface; } #ifdef SO_EXCLUSIVEADDRUSE static void -set_excladdruse(int fd) +set_excladdruse( + SOCKET fd + ) { int one = 1; int failed; +#ifdef SYS_WINNT + DWORD err; +#endif failed = setsockopt(fd, SOL_SOCKET, SO_EXCLUSIVEADDRUSE, (char *)&one, sizeof(one)); - if (failed) - netsyslog(LOG_ERR, - "setsockopt(%d, SO_EXCLUSIVEADDRUSE, on): %m", fd); + if (!failed) + return; + +#ifdef SYS_WINNT + /* + * Prior to Windows XP setting SO_EXCLUSIVEADDRUSE can fail with + * error WSAINVAL depending on service pack level and whether + * the user account is in the Administrators group. Do not + * complain if it fails that way on versions prior to XP (5.1). + */ + err = GetLastError(); + + if (isc_win32os_versioncheck(5, 1, 0, 0) < 0 /* < 5.1/XP */ + && WSAEINVAL == err) + return; + + SetLastError(err); +#endif + msyslog(LOG_ERR, + "setsockopt(%d, SO_EXCLUSIVEADDRUSE, on): %m", + (int)fd); } #endif /* SO_EXCLUSIVEADDRUSE */ @@ -1604,29 +2242,30 @@ set_excladdruse(int fd) * fd's also? */ static void -set_reuseaddr(int flag) { - struct interface *interf; - +set_reuseaddr( + int flag + ) +{ #ifndef SO_EXCLUSIVEADDRUSE + endpt *ep; - for (interf = ISC_LIST_HEAD(inter_list); - interf != NULL; - interf = ISC_LIST_NEXT(interf, link)) { - - if (interf->flags & INT_WILDCARD) + for (ep = ep_list; ep != NULL; ep = ep->elink) { + if (ep->flags & INT_WILDCARD) continue; - + /* - * if interf->fd is INVALID_SOCKET, we might have a adapter + * if ep->fd is INVALID_SOCKET, we might have a adapter * configured but not present */ - DPRINTF(4, ("setting SO_REUSEADDR on %.16s@%s to %s\n", interf->name, stoa(&interf->sin), flag ? "on" : "off")); - - if (interf->fd != INVALID_SOCKET) { - if (setsockopt(interf->fd, SOL_SOCKET, - SO_REUSEADDR, (char *)&flag, - sizeof(flag))) { - netsyslog(LOG_ERR, "set_reuseaddr: setsockopt(SO_REUSEADDR, %s) failed: %m", flag ? "on" : "off"); + DPRINTF(4, ("setting SO_REUSEADDR on %.16s@%s to %s\n", + ep->name, stoa(&ep->sin), + flag ? "on" : "off")); + + if (ep->fd != INVALID_SOCKET) { + if (setsockopt(ep->fd, SOL_SOCKET, SO_REUSEADDR, + (char *)&flag, sizeof(flag))) { + msyslog(LOG_ERR, "set_reuseaddr: setsockopt(%s, SO_REUSEADDR, %s) failed: %m", + stoa(&ep->sin), flag ? "on" : "off"); } } } @@ -1638,117 +2277,114 @@ set_reuseaddr(int flag) { * make other changes as necessary later on */ void -enable_broadcast(struct interface *iface, struct sockaddr_storage *baddr) +enable_broadcast( + struct interface * iface, + sockaddr_u * baddr + ) { -#ifdef SO_BROADCAST +#ifdef OPEN_BCAST_SOCKET socket_broadcast_enable(iface, iface->fd, baddr); #endif } -#ifdef OPEN_BCAST_SOCKET +#ifdef OPEN_BCAST_SOCKET /* * Enable a broadcast address to a given socket - * The socket is in the inter_list all we need to do is enable + * The socket is in the ep_list all we need to do is enable * broadcasting. It is not this function's job to select the socket */ static isc_boolean_t -socket_broadcast_enable(struct interface *iface, SOCKET fd, struct sockaddr_storage *maddr) +socket_broadcast_enable( + struct interface * iface, + SOCKET fd, + sockaddr_u * baddr + ) { #ifdef SO_BROADCAST int on = 1; - if (maddr->ss_family == AF_INET) - { + if (IS_IPV4(baddr)) { /* if this interface can support broadcast, set SO_BROADCAST */ if (setsockopt(fd, SOL_SOCKET, SO_BROADCAST, (char *)&on, sizeof(on))) - { - netsyslog(LOG_ERR, "setsockopt(SO_BROADCAST) enable failure on address %s: %m", - stoa(maddr)); - } -#ifdef DEBUG - else if (debug > 1) { - printf("Broadcast enabled on socket %d for address %s\n", - fd, stoa(maddr)); - } -#endif + msyslog(LOG_ERR, + "setsockopt(SO_BROADCAST) enable failure on address %s: %m", + stoa(baddr)); + else + DPRINTF(2, ("Broadcast enabled on socket %d for address %s\n", + fd, stoa(baddr))); } - iface->flags |= INT_BCASTOPEN; + iface->flags |= INT_BCASTXMIT; return ISC_TRUE; #else return ISC_FALSE; #endif /* SO_BROADCAST */ } +#ifdef OS_MISSES_SPECIFIC_ROUTE_UPDATES /* * Remove a broadcast address from a given socket - * The socket is in the inter_list all we need to do is disable + * The socket is in the ep_list all we need to do is disable * broadcasting. It is not this function's job to select the socket */ static isc_boolean_t -socket_broadcast_disable(struct interface *iface, struct sockaddr_storage *maddr) +socket_broadcast_disable( + struct interface * iface, + sockaddr_u * baddr + ) { #ifdef SO_BROADCAST int off = 0; /* This seems to be OK as an int */ - if (maddr->ss_family == AF_INET) - { - if (setsockopt(iface->fd, SOL_SOCKET, SO_BROADCAST, - (char *)&off, sizeof(off))) - { - netsyslog(LOG_ERR, "setsockopt(SO_BROADCAST) disable failure on address %s: %m", - stoa(maddr)); - } - } - iface->flags &= ~INT_BCASTOPEN; + if (IS_IPV4(baddr) && setsockopt(iface->fd, SOL_SOCKET, + SO_BROADCAST, (char *)&off, sizeof(off))) + msyslog(LOG_ERR, + "setsockopt(SO_BROADCAST) disable failure on address %s: %m", + stoa(baddr)); + + iface->flags &= ~INT_BCASTXMIT; return ISC_TRUE; #else return ISC_FALSE; #endif /* SO_BROADCAST */ } +#endif /* OS_MISSES_SPECIFIC_ROUTE_UPDATES */ #endif /* OPEN_BCAST_SOCKET */ + +/* + * return the broadcast client flag value + */ +isc_boolean_t +get_broadcastclient_flag(void) +{ + return (broadcast_client_enabled); +} /* * Check to see if the address is a multicast address */ static isc_boolean_t -addr_ismulticast(struct sockaddr_storage *maddr) +addr_ismulticast( + sockaddr_u *maddr + ) { - switch (maddr->ss_family) - { - case AF_INET : - if (!IN_CLASSD(ntohl(((struct sockaddr_in*)maddr)->sin_addr.s_addr))) { - DPRINTF(4, ("multicast address %s not class D\n", stoa(maddr))); - return (ISC_FALSE); - } - else - { - return (ISC_TRUE); - } + isc_boolean_t result; - case AF_INET6 : -#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT - if (!IN6_IS_ADDR_MULTICAST(&((struct sockaddr_in6*)maddr)->sin6_addr)) { - DPRINTF(4, ("address %s not IPv6 multicast address\n", stoa(maddr))); - return (ISC_FALSE); - } - else - { - return (ISC_TRUE); - } - -/* - * If we don't have IPV6 support any IPV6 address is not multicast - */ -#else - return (ISC_FALSE); -#endif +#ifndef INCLUDE_IPV6_MULTICAST_SUPPORT /* - * Never valid + * If we don't have IPV6 support any IPV6 addr is not multicast */ - default: - return (ISC_FALSE); - } + if (IS_IPV6(maddr)) + result = ISC_FALSE; + else +#endif + result = IS_MCAST(maddr); + + if (!result) + DPRINTF(4, ("address %s is not multicast\n", + stoa(maddr))); + + return result; } /* @@ -1757,67 +2393,57 @@ addr_ismulticast(struct sockaddr_storage *maddr) * send the multicast packet. */ void -enable_multicast_if(struct interface *iface, struct sockaddr_storage *maddr) +enable_multicast_if( + struct interface * iface, + sockaddr_u * maddr + ) { #ifdef MCAST #ifdef IP_MULTICAST_LOOP - /*u_char*/ TYPEOF_IP_MULTICAST_LOOP off = 0; + TYPEOF_IP_MULTICAST_LOOP off = 0; #endif -#ifdef IPV6_MULTICAST_LOOP - u_int off6 = 0; /* RFC 3493, 5.2. defines type unsigned int */ +#if defined(INCLUDE_IPV6_MULTICAST_SUPPORT) && defined(IPV6_MULTICAST_LOOP) + u_int off6 = 0; #endif - switch (maddr->ss_family) - { + NTP_REQUIRE(AF(maddr) == AF(&iface->sin)); + + switch (AF(&iface->sin)) { + case AF_INET: - if (setsockopt(iface->fd, IPPROTO_IP, IP_MULTICAST_IF, - (char *)&(((struct sockaddr_in*)&iface->sin)->sin_addr.s_addr), - sizeof(struct in_addr)) == -1) { - netsyslog(LOG_ERR, - "setsockopt IP_MULTICAST_IF failure: %m on socket %d, addr %s for multicast address %s", - iface->fd, stoa(&iface->sin), stoa(maddr)); - return; - } #ifdef IP_MULTICAST_LOOP /* - * Don't send back to itself, but allow it to fail to set it + * Don't send back to itself, but allow failure to set */ - if (setsockopt(iface->fd, IPPROTO_IP, IP_MULTICAST_LOOP, - SETSOCKOPT_ARG_CAST &off, sizeof(off)) == -1) { - netsyslog(LOG_ERR, - "setsockopt IP_MULTICAST_LOOP failure: %m on socket %d, addr %s for multicast address %s", - iface->fd, stoa(&iface->sin), stoa(maddr)); + if (setsockopt(iface->fd, IPPROTO_IP, + IP_MULTICAST_LOOP, + SETSOCKOPT_ARG_CAST &off, + sizeof(off))) { + + msyslog(LOG_ERR, + "setsockopt IP_MULTICAST_LOOP failed: %m on socket %d, addr %s for multicast address %s", + iface->fd, stoa(&iface->sin), + stoa(maddr)); } #endif - DPRINTF(4, ("Added IPv4 multicast interface on socket %d, addr %s for multicast address %s\n", - iface->fd, stoa(&iface->sin), - stoa(maddr))); break; case AF_INET6: #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT - if (setsockopt(iface->fd, IPPROTO_IPV6, IPV6_MULTICAST_IF, - (char *) &iface->scopeid, sizeof(iface->scopeid)) == -1) { - netsyslog(LOG_ERR, - "setsockopt IPV6_MULTICAST_IF failure: %m on socket %d, addr %s, scope %d for multicast address %s", - iface->fd, stoa(&iface->sin), iface->scopeid, - stoa(maddr)); - return; - } #ifdef IPV6_MULTICAST_LOOP /* - * Don't send back to itself, but allow it to fail to set it + * Don't send back to itself, but allow failure to set */ - if (setsockopt(iface->fd, IPPROTO_IPV6, IPV6_MULTICAST_LOOP, - (char *) &off6, sizeof(off6)) == -1) { - netsyslog(LOG_ERR, - "setsockopt IPV6_MULTICAST_LOOP failure: %m on socket %d, addr %s for multicast address %s", - iface->fd, stoa(&iface->sin), stoa(maddr)); + if (setsockopt(iface->fd, IPPROTO_IPV6, + IPV6_MULTICAST_LOOP, + (char *) &off6, sizeof(off6))) { + + msyslog(LOG_ERR, + "setsockopt IPV6_MULTICAST_LOOP failed: %m on socket %d, addr %s for multicast address %s", + iface->fd, stoa(&iface->sin), + stoa(maddr)); } #endif - DPRINTF(4, ("Added IPv6 multicast interface on socket %d, addr %s, scope %d for multicast address %s\n", - iface->fd, stoa(&iface->sin), iface->scopeid, - stoa(maddr))); break; #else return; @@ -1829,37 +2455,37 @@ enable_multicast_if(struct interface *iface, struct sockaddr_storage *maddr) /* * Add a multicast address to a given socket - * The socket is in the inter_list all we need to do is enable + * The socket is in the ep_list all we need to do is enable * multicasting. It is not this function's job to select the socket */ +#if defined(MCAST) static isc_boolean_t -socket_multicast_enable(struct interface *iface, int lscope, struct sockaddr_storage *maddr) +socket_multicast_enable( + endpt * iface, + sockaddr_u * maddr + ) { + struct ip_mreq mreq; #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT - struct ipv6_mreq mreq6; - struct in6_addr iaddr6; -#endif /* INCLUDE_IPV6_MULTICAST_SUPPORT */ - - struct ip_mreq mreq; - - if (find_addr_in_list(maddr)) { - DPRINTF(4, ("socket_multicast_enable(%s): already enabled\n", stoa(maddr))); - return ISC_TRUE; - } + struct ipv6_mreq mreq6; +#endif + switch (AF(maddr)) { - switch (maddr->ss_family) - { case AF_INET: - memset((char *)&mreq, 0, sizeof(mreq)); - mreq.imr_multiaddr = (((struct sockaddr_in*)maddr)->sin_addr); + ZERO(mreq); + mreq.imr_multiaddr = SOCK_ADDR4(maddr); mreq.imr_interface.s_addr = htonl(INADDR_ANY); - if (setsockopt(iface->fd, IPPROTO_IP, IP_ADD_MEMBERSHIP, - (char *)&mreq, sizeof(mreq)) == -1) { - netsyslog(LOG_ERR, - "setsockopt IP_ADD_MEMBERSHIP failure: %m on socket %d, addr %s for %x / %x (%s)", - iface->fd, stoa(&iface->sin), - mreq.imr_multiaddr.s_addr, - mreq.imr_interface.s_addr, stoa(maddr)); + if (setsockopt(iface->fd, + IPPROTO_IP, + IP_ADD_MEMBERSHIP, + (char *)&mreq, + sizeof(mreq))) { + DPRINTF(2, ( + "setsockopt IP_ADD_MEMBERSHIP failed: %m on socket %d, addr %s for %x / %x (%s)", + iface->fd, stoa(&iface->sin), + mreq.imr_multiaddr.s_addr, + mreq.imr_interface.s_addr, + stoa(maddr))); return ISC_FALSE; } DPRINTF(4, ("Added IPv4 multicast membership on socket %d, addr %s for %x / %x (%s)\n", @@ -1871,71 +2497,79 @@ socket_multicast_enable(struct interface *iface, int lscope, struct sockaddr_sto case AF_INET6: #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT /* - * Enable reception of multicast packets - * If the address is link-local we can get the interface index - * from the scope id. Don't do this for other types of multicast - * addresses. For now let the kernel figure it out. + * Enable reception of multicast packets. + * If the address is link-local we can get the + * interface index from the scope id. Don't do this + * for other types of multicast addresses. For now let + * the kernel figure it out. */ - memset((char *)&mreq6, 0, sizeof(mreq6)); - iaddr6 = ((struct sockaddr_in6*)maddr)->sin6_addr; - mreq6.ipv6mr_multiaddr = iaddr6; - mreq6.ipv6mr_interface = lscope; - - if (setsockopt(iface->fd, IPPROTO_IPV6, IPV6_JOIN_GROUP, - (char *)&mreq6, sizeof(mreq6)) == -1) { - netsyslog(LOG_ERR, - "setsockopt IPV6_JOIN_GROUP failure: %m on socket %d, addr %s for interface %d(%s)", - iface->fd, stoa(&iface->sin), - mreq6.ipv6mr_interface, stoa(maddr)); + ZERO(mreq6); + mreq6.ipv6mr_multiaddr = SOCK_ADDR6(maddr); + mreq6.ipv6mr_interface = iface->ifindex; + + if (setsockopt(iface->fd, IPPROTO_IPV6, + IPV6_JOIN_GROUP, (char *)&mreq6, + sizeof(mreq6))) { + DPRINTF(2, ( + "setsockopt IPV6_JOIN_GROUP failed: %m on socket %d, addr %s for interface %u (%s)", + iface->fd, stoa(&iface->sin), + mreq6.ipv6mr_interface, stoa(maddr))); return ISC_FALSE; } - DPRINTF(4, ("Added IPv6 multicast group on socket %d, addr %s for interface %d(%s)\n", + DPRINTF(4, ("Added IPv6 multicast group on socket %d, addr %s for interface %u (%s)\n", iface->fd, stoa(&iface->sin), mreq6.ipv6mr_interface, stoa(maddr))); - break; #else return ISC_FALSE; #endif /* INCLUDE_IPV6_MULTICAST_SUPPORT */ } iface->flags |= INT_MCASTOPEN; iface->num_mcast++; - add_addr_to_list(maddr, iface); + return ISC_TRUE; } +#endif /* MCAST */ + /* * Remove a multicast address from a given socket - * The socket is in the inter_list all we need to do is disable + * The socket is in the ep_list all we need to do is disable * multicasting. It is not this function's job to select the socket */ +#ifdef MCAST static isc_boolean_t -socket_multicast_disable(struct interface *iface, struct sockaddr_storage *maddr) +socket_multicast_disable( + struct interface * iface, + sockaddr_u * maddr + ) { #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT struct ipv6_mreq mreq6; - struct in6_addr iaddr6; -#endif /* INCLUDE_IPV6_MULTICAST_SUPPORT */ - +#endif struct ip_mreq mreq; - memset((char *)&mreq, 0, sizeof(mreq)); + + ZERO(mreq); if (find_addr_in_list(maddr) == NULL) { - DPRINTF(4, ("socket_multicast_disable(%s): not enabled\n", stoa(maddr))); + DPRINTF(4, ("socket_multicast_disable(%s): not found\n", + stoa(maddr))); return ISC_TRUE; } - switch (maddr->ss_family) - { + switch (AF(maddr)) { + case AF_INET: - mreq.imr_multiaddr = (((struct sockaddr_in*)&maddr)->sin_addr); - mreq.imr_interface.s_addr = ((struct sockaddr_in*)&iface->sin)->sin_addr.s_addr; - if (setsockopt(iface->fd, IPPROTO_IP, IP_DROP_MEMBERSHIP, - (char *)&mreq, sizeof(mreq)) == -1) { - netsyslog(LOG_ERR, - "setsockopt IP_DROP_MEMBERSHIP failure: %m on socket %d, addr %s for %x / %x (%s)", - iface->fd, stoa(&iface->sin), - mreq.imr_multiaddr.s_addr, - mreq.imr_interface.s_addr, stoa(maddr)); + mreq.imr_multiaddr = SOCK_ADDR4(maddr); + mreq.imr_interface = SOCK_ADDR4(&iface->sin); + if (setsockopt(iface->fd, IPPROTO_IP, + IP_DROP_MEMBERSHIP, (char *)&mreq, + sizeof(mreq))) { + + msyslog(LOG_ERR, + "setsockopt IP_DROP_MEMBERSHIP failed: %m on socket %d, addr %s for %x / %x (%s)", + iface->fd, stoa(&iface->sin), + SRCADR(maddr), SRCADR(&iface->sin), + stoa(maddr)); return ISC_FALSE; } break; @@ -1943,35 +2577,37 @@ socket_multicast_disable(struct interface *iface, struct sockaddr_storage *maddr #ifdef INCLUDE_IPV6_MULTICAST_SUPPORT /* * Disable reception of multicast packets - * If the address is link-local we can get the interface index - * from the scope id. Don't do this for other types of multicast - * addresses. For now let the kernel figure it out. + * If the address is link-local we can get the + * interface index from the scope id. Don't do this + * for other types of multicast addresses. For now let + * the kernel figure it out. */ - iaddr6 = ((struct sockaddr_in6*)&maddr)->sin6_addr; - mreq6.ipv6mr_multiaddr = iaddr6; - mreq6.ipv6mr_interface = iface->scopeid; - - if (setsockopt(iface->fd, IPPROTO_IPV6, IPV6_LEAVE_GROUP, - (char *)&mreq6, sizeof(mreq6)) == -1) { - netsyslog(LOG_ERR, - "setsockopt IPV6_LEAVE_GROUP failure: %m on socket %d, addr %s for %d(%s)", - iface->fd, stoa(&iface->sin), - mreq6.ipv6mr_interface, stoa(maddr)); + mreq6.ipv6mr_multiaddr = SOCK_ADDR6(maddr); + mreq6.ipv6mr_interface = iface->ifindex; + + if (setsockopt(iface->fd, IPPROTO_IPV6, + IPV6_LEAVE_GROUP, (char *)&mreq6, + sizeof(mreq6))) { + + msyslog(LOG_ERR, + "setsockopt IPV6_LEAVE_GROUP failure: %m on socket %d, addr %s for %d (%s)", + iface->fd, stoa(&iface->sin), + iface->ifindex, stoa(maddr)); return ISC_FALSE; } break; #else return ISC_FALSE; #endif /* INCLUDE_IPV6_MULTICAST_SUPPORT */ - } + iface->num_mcast--; - if (iface->num_mcast <= 0) { - iface->num_mcast = 0; + if (!iface->num_mcast) iface->flags &= ~INT_MCASTOPEN; - } + return ISC_TRUE; } +#endif /* MCAST */ /* * io_setbclient - open the broadcast client sockets @@ -1979,38 +2615,37 @@ socket_multicast_disable(struct interface *iface, struct sockaddr_storage *maddr void io_setbclient(void) { -#ifdef OPEN_BCAST_SOCKET - struct interface *interf; - int nif = 0; - isc_boolean_t jstatus; - SOCKET fd; +#ifdef OPEN_BCAST_SOCKET + struct interface * interf; + int nif; + nif = 0; set_reuseaddr(1); - for (interf = ISC_LIST_HEAD(inter_list); + for (interf = ep_list; interf != NULL; - interf = ISC_LIST_NEXT(interf, link)) { - if (interf->flags & INT_WILDCARD) - continue; - - /* use only allowed addresses */ - if (interf->ignore_packets == ISC_TRUE) + interf = interf->elink) { + + if (interf->flags & (INT_WILDCARD | INT_LOOPBACK)) continue; - /* Only IPv4 addresses are valid for broadcast */ - if (interf->sin.ss_family != AF_INET) + + /* use only allowed addresses */ + if (interf->ignore_packets) continue; - /* Is this a broadcast address? */ + /* Need a broadcast-capable interface */ if (!(interf->flags & INT_BROADCAST)) continue; - /* Skip the loopback addresses */ - if (interf->flags & INT_LOOPBACK) - continue; + /* Only IPv4 addresses are valid for broadcast */ + NTP_REQUIRE(IS_IPV4(&interf->sin)); /* Do we already have the broadcast address open? */ if (interf->flags & INT_BCASTOPEN) { - /* account for already open interfaces to aviod misleading warning below */ + /* + * account for already open interfaces to avoid + * misleading warning below + */ nif++; continue; } @@ -2019,41 +2654,41 @@ io_setbclient(void) * Try to open the broadcast address */ interf->family = AF_INET; - interf->bfd = open_socket(&interf->bcast, - INT_BROADCAST, 0, interf); + interf->bfd = open_socket(&interf->bcast, 1, 0, interf); - /* - * If we succeeded then we use it otherwise - * enable the underlying address + /* + * If we succeeded then we use it otherwise enable + * broadcast on the interface address */ - if (interf->bfd == INVALID_SOCKET) { - fd = interf->fd; - } - else { - fd = interf->bfd; - } - - /* Enable Broadcast on socket */ - jstatus = socket_broadcast_enable(interf, fd, &interf->sin); - if (jstatus == ISC_TRUE) - { + if (interf->bfd != INVALID_SOCKET) { nif++; - netsyslog(LOG_INFO,"io_setbclient: Opened broadcast client on interface #%d %s, socket: %d", - interf->ifnum, interf->name, fd); - interf->addr_refid = addr2refid(&interf->sin); + interf->flags |= INT_BCASTOPEN; + msyslog(LOG_INFO, + "Listen for broadcasts to %s on interface #%d %s", + stoa(&interf->bcast), interf->ifnum, interf->name); + } else { + /* silently ignore EADDRINUSE as we probably opened + the socket already for an address in the same network */ + if (errno != EADDRINUSE) + msyslog(LOG_INFO, + "failed to listen for broadcasts to %s on interface #%d %s", + stoa(&interf->bcast), interf->ifnum, interf->name); } } set_reuseaddr(0); -#ifdef DEBUG - if (debug) - if (nif > 0) - printf("io_setbclient: Opened broadcast clients\n"); -#endif - if (nif == 0) - netsyslog(LOG_ERR, "Unable to listen for broadcasts, no broadcast interfaces available"); + if (nif > 0) { + broadcast_client_enabled = ISC_TRUE; + DPRINTF(1, ("io_setbclient: listening to %d broadcast addresses\n", nif)); + } + else if (!nif) { + broadcast_client_enabled = ISC_FALSE; + msyslog(LOG_ERR, + "Unable to listen for broadcasts, no broadcast interfaces available"); + } #else - netsyslog(LOG_ERR, "io_setbclient: Broadcast Client disabled by build"); -#endif + msyslog(LOG_ERR, + "io_setbclient: Broadcast Client disabled by build"); +#endif /* OPEN_BCAST_SOCKET */ } /* @@ -2062,20 +2697,25 @@ io_setbclient(void) void io_unsetbclient(void) { - struct interface *interf; - isc_boolean_t lstatus; + endpt *ep; - for (interf = ISC_LIST_HEAD(inter_list); - interf != NULL; - interf = ISC_LIST_NEXT(interf, link)) - { - if (interf->flags & INT_WILDCARD) - continue; - - if (!(interf->flags & INT_BCASTOPEN)) - continue; - lstatus = socket_broadcast_disable(interf, &interf->sin); + for (ep = ep_list; ep != NULL; ep = ep->elink) { + if (INT_WILDCARD & ep->flags) + continue; + if (!(INT_BCASTOPEN & ep->flags)) + continue; + + if (ep->bfd != INVALID_SOCKET) { + /* destroy broadcast listening socket */ + msyslog(LOG_INFO, + "stop listening for broadcasts to %s on interface #%d %s", + stoa(&ep->bcast), ep->ifnum, ep->name); + close_and_delete_fd_from_list(ep->bfd); + ep->bfd = INVALID_SOCKET; + ep->flags &= ~INT_BCASTOPEN; + } } + broadcast_client_enabled = ISC_FALSE; } /* @@ -2083,279 +2723,146 @@ io_unsetbclient(void) */ void io_multicast_add( - struct sockaddr_storage addr + sockaddr_u *addr ) { #ifdef MCAST - struct interface *interface; -#ifndef MULTICAST_NONEWSOCKET - struct interface *iface; -#endif - int lscope = 0; - + endpt * ep; + endpt * one_ep; + /* * Check to see if this is a multicast address */ - if (addr_ismulticast(&addr) == ISC_FALSE) + if (!addr_ismulticast(addr)) return; /* If we already have it we can just return */ - if (find_flagged_addr_in_list(&addr, INT_MCASTOPEN|INT_MCASTIF) != NULL) - { - netsyslog(LOG_INFO, "Duplicate request found for multicast address %s", - stoa(&addr)); + if (NULL != find_flagged_addr_in_list(addr, INT_MCASTOPEN)) { + msyslog(LOG_INFO, + "Duplicate request found for multicast address %s", + stoa(addr)); return; } #ifndef MULTICAST_NONEWSOCKET - interface = new_interface(NULL); - + ep = new_interface(NULL); + /* * Open a new socket for the multicast address */ - interface->sin.ss_family = addr.ss_family; - interface->family = addr.ss_family; + ep->sin = *addr; + SET_PORT(&ep->sin, NTP_PORT); + ep->family = AF(&ep->sin); + AF(&ep->mask) = ep->family; + SET_ONESMASK(&ep->mask); - switch(addr.ss_family) { - case AF_INET: - memcpy(&(((struct sockaddr_in *)&interface->sin)->sin_addr), - &(((struct sockaddr_in*)&addr)->sin_addr), - sizeof(struct in_addr)); - ((struct sockaddr_in*)&interface->sin)->sin_port = htons(NTP_PORT); - memset(&((struct sockaddr_in*)&interface->mask)->sin_addr.s_addr, 0xff, sizeof(struct in_addr)); - break; - case AF_INET6: -#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT - memcpy(&(((struct sockaddr_in6 *)&interface->sin)->sin6_addr), - &((struct sockaddr_in6*)&addr)->sin6_addr, - sizeof(struct in6_addr)); - ((struct sockaddr_in6*)&interface->sin)->sin6_port = htons(NTP_PORT); -#ifdef ISC_PLATFORM_HAVESCOPEID - ((struct sockaddr_in6*)&interface->sin)->sin6_scope_id = ((struct sockaddr_in6*)&addr)->sin6_scope_id; -#endif - memset(&((struct sockaddr_in6*)&interface->mask)->sin6_addr.s6_addr, 0xff, sizeof(struct in6_addr)); -#endif - iface = findlocalcastinterface(&addr, INT_MULTICAST); - if (iface) { -# ifdef ISC_PLATFORM_HAVESCOPEID - lscope = ((struct sockaddr_in6*)&iface->sin)->sin6_scope_id; -# endif - DPRINTF(4, ("Found interface #%d %s, scope: %d for address %s\n", iface->ifnum, iface->name, lscope, stoa(&addr))); - } - break; - } - set_reuseaddr(1); - interface->bfd = INVALID_SOCKET; - interface->fd = open_socket(&interface->sin, - INT_MULTICAST, 0, interface); + ep->bfd = INVALID_SOCKET; + ep->fd = open_socket(&ep->sin, 0, 0, ep); + if (ep->fd != INVALID_SOCKET) { + ep->ignore_packets = ISC_FALSE; + ep->flags |= INT_MCASTIF; + + strlcpy(ep->name, "multicast", sizeof(ep->name)); + DPRINT_INTERFACE(2, (ep, "multicast add ", "\n")); + add_interface(ep); + log_listen_address(ep); + } else { + /* bind failed, re-use wildcard interface */ + delete_interface(ep); - if (interface->fd != INVALID_SOCKET) - { - interface->bfd = INVALID_SOCKET; - interface->ignore_packets = ISC_FALSE; - interface->flags |= INT_MCASTIF; - - (void) strncpy(interface->name, "multicast", - sizeof(interface->name)); - ((struct sockaddr_in*)&interface->mask)->sin_addr.s_addr = - htonl(~(u_int32)0); - DPRINT_INTERFACE(2, (interface, "multicast add ", "\n")); - /* socket_multicast_enable() will add this address to the addresslist */ - add_interface(interface); - list_if_listening(interface, htons(NTP_PORT)); - } - else - { - delete_interface(interface); /* re-use existing interface */ - interface = NULL; - if (addr.ss_family == AF_INET) - interface = wildipv4; - else if (addr.ss_family == AF_INET6) - interface = wildipv6; - - if (interface != NULL) { + if (IS_IPV4(addr)) + ep = wildipv4; + else if (IS_IPV6(addr)) + ep = wildipv6; + else + ep = NULL; + + if (ep != NULL) { /* HACK ! -- stuff in an address */ - interface->bcast = addr; - netsyslog(LOG_ERR, - "...multicast address %s using wildcard interface #%d %s", - stoa(&addr), interface->ifnum, interface->name); + /* because we don't bind addr? DH */ + ep->bcast = *addr; + msyslog(LOG_ERR, + "multicast address %s using wildcard interface #%d %s", + stoa(addr), ep->ifnum, ep->name); } else { - netsyslog(LOG_ERR, - "No multicast socket available to use for address %s", - stoa(&addr)); + msyslog(LOG_ERR, + "No multicast socket available to use for address %s", + stoa(addr)); return; } } -#else - /* - * For the case where we can't use a separate socket - */ - interface = findlocalcastinterface(&addr, INT_MULTICAST); + { /* in place of the { following for in #else clause */ + one_ep = ep; +#else /* MULTICAST_NONEWSOCKET follows */ /* - * If we don't have a valid socket, just return + * For the case where we can't use a separate socket (Windows) + * join each applicable endpoint socket to the group address. */ - if (!interface) - { - netsyslog(LOG_ERR, - "Cannot add multicast address %s: Cannot find slot", - stoa(&addr)); - return; + if (IS_IPV4(addr)) + one_ep = wildipv4; + else + one_ep = wildipv6; + for (ep = ep_list; ep != NULL; ep = ep->elink) { + if (ep->ignore_packets || AF(&ep->sin) != AF(addr) || + !(INT_MULTICAST & ep->flags) || + (INT_LOOPBACK | INT_WILDCARD) & ep->flags) + continue; + one_ep = ep; +#endif /* MULTICAST_NONEWSOCKET */ + if (socket_multicast_enable(ep, addr)) + msyslog(LOG_INFO, + "Joined %s socket to multicast group %s", + stoa(&ep->sin), + stoa(addr)); } + add_addr_to_list(addr, one_ep); +#else /* !MCAST follows*/ + msyslog(LOG_ERR, + "Can not add multicast address %s: no multicast support", + stoa(addr)); #endif - { - isc_boolean_t jstatus; - jstatus = socket_multicast_enable(interface, lscope, &addr); - - if (jstatus == ISC_TRUE) - netsyslog(LOG_INFO, "Added Multicast Listener %s on interface #%d %s\n", stoa(&addr), interface->ifnum, interface->name); - else - netsyslog(LOG_ERR, "Failed to add Multicast Listener %s\n", stoa(&addr)); - } -#else /* MCAST */ - netsyslog(LOG_ERR, - "Cannot add multicast address %s: no Multicast support", - stoa(&addr)); -#endif /* MCAST */ return; } + /* * io_multicast_del() - delete multicast group address */ void io_multicast_del( - struct sockaddr_storage addr + sockaddr_u * addr ) { #ifdef MCAST - struct interface *interface; - isc_boolean_t lstatus; + endpt *iface; /* * Check to see if this is a multicast address */ - if (addr_ismulticast(&addr) == ISC_FALSE) - { - netsyslog(LOG_ERR, - "invalid multicast address %s", stoa(&addr)); + if (!addr_ismulticast(addr)) { + msyslog(LOG_ERR, "invalid multicast address %s", + stoa(addr)); return; } - switch (addr.ss_family) - { - case AF_INET : - /* - * Disable reception of multicast packets - */ - interface = find_flagged_addr_in_list(&addr, INT_MCASTOPEN); - while ( interface != NULL) { - lstatus = socket_multicast_disable(interface, &addr); - interface = find_flagged_addr_in_list(&addr, INT_MCASTOPEN); - } - break; - -#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT - case AF_INET6 : - /* - * Disable reception of multicast packets - */ - for (interface = ISC_LIST_HEAD(inter_list); - interface != NULL; - interface = ISC_LIST_NEXT(interface, link)) - { - if (interface->flags & INT_WILDCARD) - continue; - - /* Be sure it's the correct family */ - if (interface->sin.ss_family != AF_INET6) - continue; - if (!(interface->flags & INT_MCASTOPEN)) - continue; - if (!(interface->fd < 0)) - continue; - if (!SOCKCMP(&addr, &interface->sin)) - continue; - lstatus = socket_multicast_disable(interface, &addr); - } - break; -#endif /* INCLUDE_IPV6_MULTICAST_SUPPORT */ - - }/* switch */ + /* + * Disable reception of multicast packets + */ + while ((iface = find_flagged_addr_in_list(addr, INT_MCASTOPEN)) + != NULL) + socket_multicast_disable(iface, addr); - delete_addr_from_list(&addr); + delete_addr_from_list(addr); #else /* not MCAST */ - netsyslog(LOG_ERR, "this function requires multicast kernel"); + msyslog(LOG_ERR, + "Can not delete multicast address %s: no multicast support", + stoa(addr)); #endif /* not MCAST */ } -/* - * init_nonblocking_io() - set up descriptor to be non blocking - */ -static void init_nonblocking_io(SOCKET fd) -{ - /* - * set non-blocking, - */ - -#ifdef USE_FIONBIO - /* in vxWorks we use FIONBIO, but the others are defined for old systems, so - * all hell breaks loose if we leave them defined - */ -#undef O_NONBLOCK -#undef FNDELAY -#undef O_NDELAY -#endif - -#if defined(O_NONBLOCK) /* POSIX */ - if (fcntl(fd, F_SETFL, O_NONBLOCK) < 0) - { - netsyslog(LOG_ERR, "fcntl(O_NONBLOCK) fails on fd #%d: %m", - fd); - exit(1); - /*NOTREACHED*/ - } -#elif defined(FNDELAY) - if (fcntl(fd, F_SETFL, FNDELAY) < 0) - { - netsyslog(LOG_ERR, "fcntl(FNDELAY) fails on fd #%d: %m", - fd); - exit(1); - /*NOTREACHED*/ - } -#elif defined(O_NDELAY) /* generally the same as FNDELAY */ - if (fcntl(fd, F_SETFL, O_NDELAY) < 0) - { - netsyslog(LOG_ERR, "fcntl(O_NDELAY) fails on fd #%d: %m", - fd); - exit(1); - /*NOTREACHED*/ - } -#elif defined(FIONBIO) - { - int on = 1; - if (ioctl(fd,FIONBIO,&on) < 0) - { - netsyslog(LOG_ERR, "ioctl(FIONBIO) fails on fd #%d: %m", - fd); - exit(1); - /*NOTREACHED*/ - } - } -#elif defined(FIOSNBIO) - if (ioctl(fd,FIOSNBIO,&on) < 0) - { - netsyslog(LOG_ERR, "ioctl(FIOSNBIO) fails on fd #%d: %m", - fd); - exit(1); - /*NOTREACHED*/ - } -#else -# include "Bletch: Need non-blocking I/O!" -#endif -} /* * open_socket - open a socket, returning the file descriptor @@ -2363,48 +2870,42 @@ static void init_nonblocking_io(SOCKET fd) static SOCKET open_socket( - struct sockaddr_storage *addr, - int flags, - int turn_off_reuse, - struct interface *interf + sockaddr_u * addr, + int bcast, + int turn_off_reuse, + endpt * interf ) { - int errval; - SOCKET fd; + SOCKET fd; + int errval; /* - * int is OK for REUSEADR per + * int is OK for REUSEADR per * http://www.kohala.com/start/mcast.api.txt */ - int on = 1; - int off = 0; - -#if defined(IPTOS_LOWDELAY) && defined(IPPROTO_IP) && defined(IP_TOS) - int tos; -#endif /* IPTOS_LOWDELAY && IPPROTO_IP && IP_TOS */ + int on = 1; + int off = 0; - if ((addr->ss_family == AF_INET6) && (isc_net_probeipv6() != ISC_R_SUCCESS)) - return (INVALID_SOCKET); + if (IS_IPV6(addr) && !ipv6_works) + return INVALID_SOCKET; /* create a datagram (UDP) socket */ - fd = socket(addr->ss_family, SOCK_DGRAM, 0); + fd = socket(AF(addr), SOCK_DGRAM, 0); if (INVALID_SOCKET == fd) { -#ifndef SYS_WINNT - errval = errno; -#else - errval = WSAGetLastError(); -#endif - netsyslog(LOG_ERR, - "socket(AF_INET%s, SOCK_DGRAM, 0) failed on address %s: %m", - (addr->ss_family == AF_INET6) ? "6" : "", - stoa(addr)); + errval = socket_errno(); + msyslog(LOG_ERR, + "socket(AF_INET%s, SOCK_DGRAM, 0) failed on address %s: %m", + IS_IPV6(addr) ? "6" : "", stoa(addr)); - if (errval == EPROTONOSUPPORT || + if (errval == EPROTONOSUPPORT || errval == EAFNOSUPPORT || errval == EPFNOSUPPORT) return (INVALID_SOCKET); - msyslog(LOG_ERR, "unexpected error code %d (not PROTONOSUPPORT|AFNOSUPPORT|FPNOSUPPORT) - exiting", errval); + + errno = errval; + msyslog(LOG_ERR, + "unexpected socket() error %m code %d (not EPROTONOSUPPORT nor EAFNOSUPPORT nor EPFNOSUPPORT) - exiting", + errno); exit(1); - /*NOTREACHED*/ } #ifdef SYS_WINNT @@ -2426,17 +2927,17 @@ open_socket( if (isc_win32os_versioncheck(5, 1, 0, 0) < 0) /* before 5.1 */ #endif if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, - (char *)(turn_off_reuse - ? &off - : &on), + (char *)((turn_off_reuse) + ? &off + : &on), sizeof(on))) { - netsyslog(LOG_ERR, "setsockopt SO_REUSEADDR %s" - " fails for address %s: %m", - turn_off_reuse - ? "off" - : "on", - stoa(addr)); + msyslog(LOG_ERR, + "setsockopt SO_REUSEADDR %s fails for address %s: %m", + (turn_off_reuse) + ? "off" + : "on", + stoa(addr)); closesocket(fd); return INVALID_SOCKET; } @@ -2452,39 +2953,44 @@ open_socket( /* * IPv4 specific options go here */ - if (addr->ss_family == AF_INET) { -#if defined(IPTOS_LOWDELAY) && defined(IPPROTO_IP) && defined(IP_TOS) - /* set IP_TOS to minimize packet delay */ - tos = IPTOS_LOWDELAY; - if (setsockopt(fd, IPPROTO_IP, IP_TOS, (char *) &tos, sizeof(tos)) < 0) - { - netsyslog(LOG_ERR, "setsockopt IPTOS_LOWDELAY on fails on address %s: %m", - stoa(addr)); - } -#endif /* IPTOS_LOWDELAY && IPPROTO_IP && IP_TOS */ + if (IS_IPV4(addr)) { +#if defined(IPPROTO_IP) && defined(IP_TOS) + if (setsockopt(fd, IPPROTO_IP, IP_TOS, (char*)&qos, + sizeof(qos))) + msyslog(LOG_ERR, + "setsockopt IP_TOS (%02x) fails on address %s: %m", + qos, stoa(addr)); +#endif /* IPPROTO_IP && IP_TOS */ + if (bcast) + socket_broadcast_enable(interf, fd, addr); } /* * IPv6 specific options go here */ - if (addr->ss_family == AF_INET6) { -#if defined(IPV6_V6ONLY) - if (setsockopt(fd, IPPROTO_IPV6, IPV6_V6ONLY, - (char*)&on, sizeof(on))) - { - netsyslog(LOG_ERR, "setsockopt IPV6_V6ONLY on fails on address %s: %m", + if (IS_IPV6(addr)) { +#if defined(IPPROTO_IPV6) && defined(IPV6_TCLASS) + if (setsockopt(fd, IPPROTO_IPV6, IPV6_TCLASS, (char*)&qos, + sizeof(qos))) + msyslog(LOG_ERR, + "setsockopt IPV6_TCLASS (%02x) fails on address %s: %m", + qos, stoa(addr)); +#endif /* IPPROTO_IPV6 && IPV6_TCLASS */ +#ifdef IPV6_V6ONLY + if (isc_net_probe_ipv6only() == ISC_R_SUCCESS + && setsockopt(fd, IPPROTO_IPV6, IPV6_V6ONLY, + (char*)&on, sizeof(on))) + msyslog(LOG_ERR, + "setsockopt IPV6_V6ONLY on fails on address %s: %m", stoa(addr)); - } -#endif /* IPV6_V6ONLY */ -#if defined(IPV6_BINDV6ONLY) - if (setsockopt(fd, IPPROTO_IPV6, IPV6_BINDV6ONLY, - (char*)&on, sizeof(on))) - { - netsyslog(LOG_ERR, - "setsockopt IPV6_BINDV6ONLY on fails on address %s: %m", - stoa(addr)); - } -#endif /* IPV6_BINDV6ONLY */ +#endif +#ifdef IPV6_BINDV6ONLY + if (setsockopt(fd, IPPROTO_IPV6, IPV6_BINDV6ONLY, + (char*)&on, sizeof(on))) + msyslog(LOG_ERR, + "setsockopt IPV6_BINDV6ONLY on fails on address %s: %m", + stoa(addr)); +#endif } #ifdef OS_NEEDS_REUSEADDR_FOR_IFADDRBIND @@ -2493,25 +2999,18 @@ open_socket( * addresses if a wildcard address already bound * to the port and SO_REUSEADDR is not set */ - if (!is_wildcard_addr(addr)) { - set_wildcard_reuse(addr->ss_family, 1); - } + if (!is_wildcard_addr(addr)) + set_wildcard_reuse(AF(addr), 1); #endif /* * bind the local address. */ - errval = bind(fd, (struct sockaddr *)addr, SOCKLEN(addr)); + errval = bind(fd, &addr->sa, SOCKLEN(addr)); #ifdef OS_NEEDS_REUSEADDR_FOR_IFADDRBIND - /* - * some OSes don't allow binding to more specific - * addresses if a wildcard address already bound - * to the port and REUSE_ADDR is not set - */ - if (!is_wildcard_addr(addr)) { - set_wildcard_reuse(addr->ss_family, 0); - } + if (!is_wildcard_addr(addr)) + set_wildcard_reuse(AF(addr), 0); #endif if (errval < 0) { @@ -2522,32 +3021,17 @@ open_socket( #ifdef DEBUG || debug > 1 #endif - ) { - if (addr->ss_family == AF_INET) - netsyslog(LOG_ERR, - "bind() fd %d, family AF_INET, port %d, addr %s, in_classd=%d flags=0x%x fails: %m", - fd, (int)ntohs(((struct sockaddr_in*)addr)->sin_port), - stoa(addr), - IN_CLASSD(ntohl(((struct sockaddr_in*)addr)->sin_addr.s_addr)), - flags); -#ifdef INCLUDE_IPV6_SUPPORT - else if (addr->ss_family == AF_INET6) - netsyslog(LOG_ERR, - "bind() fd %d, family AF_INET6, port %d, scope %d, addr %s, mcast=%d flags=0x%x fails: %m", - fd, (int)ntohs(((struct sockaddr_in6*)addr)->sin6_port), -# ifdef ISC_PLATFORM_HAVESCOPEID - ((struct sockaddr_in6*)addr)->sin6_scope_id -# else - -1 -# endif - , stoa(addr), - IN6_IS_ADDR_MULTICAST(&((struct sockaddr_in6*)addr)->sin6_addr), - flags); -#endif + ) { + msyslog(LOG_ERR, + "bind(%d) AF_INET%s %s#%d%s flags 0x%x failed: %m", + fd, IS_IPV6(addr) ? "6" : "", + stoa(addr), SRCPORT(addr), + IS_MCAST(addr) ? " (multicast)" : "", + interf->flags); } closesocket(fd); - + return INVALID_SOCKET; } @@ -2555,28 +3039,45 @@ open_socket( { if (setsockopt(fd, SOL_SOCKET, SO_TIMESTAMP, (char*)&on, sizeof(on))) - { - netsyslog(LOG_DEBUG, - "setsockopt SO_TIMESTAMP on fails on address %s: %m", - stoa(addr)); - } -#ifdef DEBUG + msyslog(LOG_DEBUG, + "setsockopt SO_TIMESTAMP on fails on address %s: %m", + stoa(addr)); else - { - DPRINTF(4, ("setsockopt SO_TIMESTAMP enabled on fd %d address %s\n", fd, stoa(addr))); - } + DPRINTF(4, ("setsockopt SO_TIMESTAMP enabled on fd %d address %s\n", + fd, stoa(addr))); + } #endif - } +#ifdef HAVE_TIMESTAMPNS + { + if (setsockopt(fd, SOL_SOCKET, SO_TIMESTAMPNS, + (char*)&on, sizeof(on))) + msyslog(LOG_DEBUG, + "setsockopt SO_TIMESTAMPNS on fails on address %s: %m", + stoa(addr)); + else + DPRINTF(4, ("setsockopt SO_TIMESTAMPNS enabled on fd %d address %s\n", + fd, stoa(addr))); + } #endif - DPRINTF(4, ("bind() fd %d, family %d, port %d, addr %s, flags=0x%x\n", - fd, - addr->ss_family, - (int)ntohs(((struct sockaddr_in*)addr)->sin_port), - stoa(addr), - interf->flags)); +#ifdef HAVE_BINTIME + { + if (setsockopt(fd, SOL_SOCKET, SO_BINTIME, + (char*)&on, sizeof(on))) + msyslog(LOG_DEBUG, + "setsockopt SO_BINTIME on fails on address %s: %m", + stoa(addr)); + else + DPRINTF(4, ("setsockopt SO_BINTIME enabled on fd %d address %s\n", + fd, stoa(addr))); + } +#endif + + DPRINTF(4, ("bind(%d) AF_INET%s, addr %s%%%d#%d, flags 0x%x\n", + fd, IS_IPV6(addr) ? "6" : "", stoa(addr), + SCOPE(addr), SRCPORT(addr), interf->flags)); + + make_socket_nonblocking(fd); - init_nonblocking_io(fd); - #ifdef HAVE_SIGNALED_IO init_socket_sig(fd); #endif /* not HAVE_SIGNALED_IO */ @@ -2592,8 +3093,7 @@ open_socket( /* * Add the socket to the completion port */ - if (io_completion_port_add_socket(fd, interf)) - { + if (io_completion_port_add_socket(fd, interf)) { msyslog(LOG_ERR, "unable to set up io completion port - EXITING"); exit(1); } @@ -2601,6 +3101,12 @@ open_socket( return fd; } + +#ifdef SYS_WINNT +#define sendto(fd, buf, len, flags, dest, destsz) \ + io_completion_port_sendto(fd, buf, len, (sockaddr_u *)(dest)) +#endif + /* XXX ELIMINATE sendpkt similar in ntpq.c, ntpdc.c, ntp_io.c, ntptrace.c */ /* * sendpkt - send a packet to the specified destination. Maintain a @@ -2609,216 +3115,101 @@ open_socket( */ void sendpkt( - struct sockaddr_storage *dest, - struct interface *inter, - int ttl, - struct pkt *pkt, - int len + sockaddr_u * dest, + struct interface * ep, + int ttl, + struct pkt * pkt, + int len ) { - int cc, slot; - - /* - * Send error caches. Empty slots have port == 0 - * Set ERRORCACHESIZE to 0 to disable - */ - struct cache { - u_short port; - struct in_addr addr; - }; - -#ifdef INCLUDE_IPV6_SUPPORT - struct cache6 { - u_short port; - struct in6_addr addr; - }; -#endif /* INCLUDE_IPV6_SUPPORT */ - - -#ifndef ERRORCACHESIZE -#define ERRORCACHESIZE 8 -#endif -#if ERRORCACHESIZE > 0 - static struct cache badaddrs[ERRORCACHESIZE]; -#ifdef INCLUDE_IPV6_SUPPORT - static struct cache6 badaddrs6[ERRORCACHESIZE]; -#endif /* INCLUDE_IPV6_SUPPORT */ -#else -#define badaddrs ((struct cache *)0) /* Only used in empty loops! */ -#ifdef INCLUDE_IPV6_SUPPORT -#define badaddrs6 ((struct cache6 *)0) /* Only used in empty loops! */ -#endif /* INCLUDE_IPV6_SUPPORT */ -#endif -#ifdef DEBUG - if (debug > 1) - { - if (inter != NULL) - { - printf("%ssendpkt(fd=%d dst=%s, src=%s, ttl=%d, len=%d)\n", - (ttl > 0) ? "\tMCAST\t***** " : "", - inter->fd, stoa(dest), - stoa(&inter->sin), ttl, len); - } - else - { - printf("%ssendpkt(dst=%s, ttl=%d, len=%d): no interface - IGNORED\n", - (ttl > 0) ? "\tMCAST\t***** " : "", - stoa(dest), - ttl, len); - } - } -#endif - - if (inter == NULL) /* unbound peer - drop request and wait for better network conditions */ - return; - -#ifdef MCAST + endpt * src; + int ismcast; + int cc; + int rc; + u_char cttl; + + ismcast = IS_MCAST(dest); + if (!ismcast) + src = ep; + else + src = (IS_IPV4(dest)) + ? mc4_list + : mc6_list; - /* - * for the moment we use the bcast option to set multicast ttl - */ - if (ttl > 0 && ttl != inter->last_ttl) { - + if (NULL == src) { /* - * set the multicast ttl for outgoing packets + * unbound peer - drop request and wait for better + * network conditions */ - int rtc; - - switch (inter->sin.ss_family) { - - case AF_INET : - { - u_char mttl = (u_char) ttl; - - rtc = setsockopt(inter->fd, IPPROTO_IP, IP_MULTICAST_TTL, - (const void *) &mttl, sizeof(mttl)); - break; - } - -#ifdef INCLUDE_IPV6_SUPPORT - case AF_INET6 : - { - u_int ittl = (u_char) ttl; - - rtc = setsockopt(inter->fd, IPPROTO_IPV6, IPV6_MULTICAST_HOPS, - (const void *) &ittl, sizeof(ittl)); - break; - } - -#endif /* INCLUDE_IPV6_SUPPORT */ - default: /* just NOP if not supported */ - rtc = 0; - break; - } - - if (rtc != 0) { - netsyslog(LOG_ERR, "setsockopt IP_MULTICAST_TTL/IPV6_MULTICAST_HOPS fails on address %s: %m", - stoa(&inter->sin)); - } - else - inter->last_ttl = ttl; + DPRINTF(2, ("%ssendpkt(dst=%s, ttl=%d, len=%d): no interface - IGNORED\n", + ismcast ? "\tMCAST\t***** " : "", + stoa(dest), ttl, len)); + return; } -#endif /* MCAST */ - - for (slot = ERRORCACHESIZE; --slot >= 0; ) - if(dest->ss_family == AF_INET) { - if (badaddrs[slot].port == SRCPORT(dest) && - badaddrs[slot].addr.s_addr == ((struct sockaddr_in*)dest)->sin_addr.s_addr) - break; - } -#ifdef INCLUDE_IPV6_SUPPORT - else if (dest->ss_family == AF_INET6) { - if (badaddrs6[slot].port == SRCPORT(dest) && - !memcmp(&badaddrs6[slot].addr, &((struct sockaddr_in6*)dest)->sin6_addr, sizeof(struct in6_addr))) - break; - } -#endif /* INCLUDE_IPV6_SUPPORT */ - -#if defined(HAVE_IO_COMPLETION_PORT) - cc = io_completion_port_sendto(inter, pkt, len, dest); - if (cc != ERROR_SUCCESS) -#else -#ifdef SIM - cc = srvr_rply(&ntp_node, dest, inter, pkt); -#else /* SIM */ - cc = sendto(inter->fd, (char *)pkt, (unsigned int)len, 0, (struct sockaddr *)dest, - SOCKLEN(dest)); -#endif /* SIM */ - if (cc == -1) -#endif - { - inter->notsent++; - packets_notsent++; - -#if defined(HAVE_IO_COMPLETION_PORT) - if (cc != WSAEWOULDBLOCK && cc != WSAENOBUFS && slot < 0) -#else - if (errno != EWOULDBLOCK && errno != ENOBUFS && slot < 0) -#endif - { + do { + DPRINTF(2, ("%ssendpkt(%d, dst=%s, src=%s, ttl=%d, len=%d)\n", + ismcast ? "\tMCAST\t***** " : "", src->fd, + stoa(dest), stoa(&src->sin), ttl, len)); +#ifdef MCAST + /* + * for the moment we use the bcast option to set multicast ttl + */ + if (ismcast && ttl > 0 && ttl != src->last_ttl) { /* - * Remember this, if there's an empty slot + * set the multicast ttl for outgoing packets */ - switch (dest->ss_family) { + switch (AF(&src->sin)) { case AF_INET : - - for (slot = ERRORCACHESIZE; --slot >= 0; ) - if (badaddrs[slot].port == 0) - { - badaddrs[slot].port = SRCPORT(dest); - badaddrs[slot].addr = ((struct sockaddr_in*)dest)->sin_addr; - break; - } + cttl = (u_char)ttl; + rc = setsockopt(src->fd, IPPROTO_IP, + IP_MULTICAST_TTL, + (void *)&cttl, + sizeof(cttl)); break; -#ifdef INCLUDE_IPV6_SUPPORT +# ifdef INCLUDE_IPV6_SUPPORT case AF_INET6 : - - for (slot = ERRORCACHESIZE; --slot >= 0; ) - if (badaddrs6[slot].port == 0) - { - badaddrs6[slot].port = SRCPORT(dest); - badaddrs6[slot].addr = ((struct sockaddr_in6*)dest)->sin6_addr; - break; - } - break; -#endif /* INCLUDE_IPV6_SUPPORT */ - default: /* don't care if not supported */ + rc = setsockopt(src->fd, IPPROTO_IPV6, + IPV6_MULTICAST_HOPS, + (void *)&ttl, + sizeof(ttl)); break; +# endif /* INCLUDE_IPV6_SUPPORT */ + + default: + rc = 0; } - netsyslog(LOG_ERR, "sendto(%s) (fd=%d): %m", - stoa(dest), inter->fd); + if (!rc) + src->last_ttl = ttl; + else + msyslog(LOG_ERR, + "setsockopt IP_MULTICAST_TTL/IPV6_MULTICAST_HOPS fails on address %s: %m", + stoa(&src->sin)); } - } - else - { - inter->sent++; - packets_sent++; - /* - * He's not bad any more - */ - if (slot >= 0) - { - netsyslog(LOG_INFO, "Connection re-established to %s", stoa(dest)); - switch (dest->ss_family) { - case AF_INET : - badaddrs[slot].port = 0; - break; -#ifdef INCLUDE_IPV6_SUPPORT - case AF_INET6 : - badaddrs6[slot].port = 0; - break; -#endif /* INCLUDE_IPV6_SUPPORT */ - default: /* don't care if not supported */ - break; - } +#endif /* MCAST */ + +#ifdef SIM + cc = simulate_server(dest, src, pkt); +#else + cc = sendto(src->fd, (char *)pkt, (u_int)len, 0, + &dest->sa, SOCKLEN(dest)); +#endif + if (cc == -1) { + src->notsent++; + packets_notsent++; + } else { + src->sent++; + packets_sent++; } - } + if (ismcast) + src = src->mclink; + } while (ismcast && src != NULL); } + #if !defined(HAVE_IO_COMPLETION_PORT) /* * fdbits - generate ascii representation of fd_set (FAU debug support) @@ -2833,10 +3224,9 @@ fdbits( static char buffer[256]; char * buf = buffer; - count = (count < 256) ? count : 255; + count = min(count, 255); - while (count >= 0) - { + while (count >= 0) { *buf++ = FD_ISSET(count, set) ? '#' : '-'; count--; } @@ -2845,22 +3235,29 @@ fdbits( return buffer; } + +#ifdef REFCLOCK /* * Routine to read the refclock packets for a specific interface * Return the number of bytes read. That way we know if we should * read it again or go on to the next one if no bytes returned */ static inline int -read_refclock_packet(SOCKET fd, struct refclockio *rp, l_fp ts) +read_refclock_packet( + SOCKET fd, + struct refclockio * rp, + l_fp ts + ) { - int i; - int buflen; - register struct recvbuf *rb; + int i; + int buflen; + int saved_errno; + int consumed; + struct recvbuf * rb; rb = get_free_recv_buffer(); - if (rb == NULL) - { + if (NULL == rb) { /* * No buffer space available - just drop the packet */ @@ -2872,17 +3269,18 @@ read_refclock_packet(SOCKET fd, struct refclockio *rp, l_fp ts) } i = (rp->datalen == 0 - || rp->datalen > sizeof(rb->recv_space)) - ? sizeof(rb->recv_space) : rp->datalen; - buflen = read(fd, (char *)&rb->recv_space, (unsigned)i); + || rp->datalen > (int)sizeof(rb->recv_space)) + ? (int)sizeof(rb->recv_space) + : rp->datalen; + do { + buflen = read(fd, (char *)&rb->recv_space, (u_int)i); + } while (buflen < 0 && EINTR == errno); - if (buflen < 0) - { - if (errno != EINTR && errno != EAGAIN) { - netsyslog(LOG_ERR, "clock read fd %d: %m", fd); - } + if (buflen <= 0) { + saved_errno = errno; freerecvbuf(rb); - return (buflen); + errno = saved_errno; + return buflen; } /* @@ -2890,85 +3288,141 @@ read_refclock_packet(SOCKET fd, struct refclockio *rp, l_fp ts) * put it on the full list and do bookkeeping. */ rb->recv_length = buflen; - rb->recv_srcclock = rp->srcclock; + rb->recv_peer = rp->srcclock; rb->dstadr = 0; rb->fd = fd; rb->recv_time = ts; rb->receiver = rp->clock_recv; - if (rp->io_input) - { - /* - * have direct input routine for refclocks - */ - if (rp->io_input(rb) == 0) - { - /* - * data was consumed - nothing to pass up - * into block input machine - */ - freerecvbuf(rb); - return (buflen); - } + consumed = indicate_refclock_packet(rp, rb); + if (!consumed) { + rp->recvcount++; + packets_received++; } - - add_full_recv_buffer(rb); - rp->recvcount++; - packets_received++; - return (buflen); + return buflen; } +#endif /* REFCLOCK */ -#ifdef HAVE_TIMESTAMP + +#ifdef HAVE_PACKET_TIMESTAMP /* * extract timestamps from control message buffer */ static l_fp - fetch_timestamp(struct recvbuf *rb, struct msghdr *msghdr, l_fp ts) +fetch_timestamp( + struct recvbuf * rb, + struct msghdr * msghdr, + l_fp ts + ) { -#ifdef USE_TIMESTAMP_CMSG - struct cmsghdr *cmsghdr; + struct cmsghdr * cmsghdr; +#ifdef HAVE_BINTIME + struct bintime * btp; +#endif +#ifdef HAVE_TIMESTAMPNS + struct timespec * tsp; +#endif +#ifdef HAVE_TIMESTAMP + struct timeval * tvp; +#endif + unsigned long ticks; + double fuzz; + l_fp lfpfuzz; + l_fp nts; +#ifdef DEBUG_TIMING + l_fp dts; +#endif cmsghdr = CMSG_FIRSTHDR(msghdr); while (cmsghdr != NULL) { switch (cmsghdr->cmsg_type) { +#ifdef HAVE_BINTIME + case SCM_BINTIME: +#endif /* HAVE_BINTIME */ +#ifdef HAVE_TIMESTAMPNS + case SCM_TIMESTAMPNS: +#endif /* HAVE_TIMESTAMPNS */ +#ifdef HAVE_TIMESTAMP case SCM_TIMESTAMP: - { - struct timeval *tvp = (struct timeval *)CMSG_DATA(cmsghdr); - double dtemp; - l_fp nts; - DPRINTF(4, ("fetch_timestamp: system network time stamp: %ld.%06ld\n", tvp->tv_sec, tvp->tv_usec)); - nts.l_i = tvp->tv_sec + JAN_1970; - dtemp = tvp->tv_usec / 1e6; - - /* fuzz lower bits not covered by precision */ - if (sys_precision != 0) - dtemp += (ntp_random() / FRAC - .5) / (1 << - -sys_precision); - - nts.l_uf = (u_int32)(dtemp*FRAC); -#ifdef DEBUG_TIMING +#endif /* HAVE_TIMESTAMP */ +#if defined(HAVE_BINTIME) || defined (HAVE_TIMESTAMPNS) || defined(HAVE_TIMESTAMP) + switch (cmsghdr->cmsg_type) { - l_fp dts = ts; - L_SUB(&dts, &nts); - collect_timing(rb, "input processing delay", 1, &dts); - DPRINTF(4, ("fetch_timestamp: timestamp delta: %s (incl. prec fuzz)\n", lfptoa(&dts, 9))); +#ifdef HAVE_BINTIME + case SCM_BINTIME: + btp = (struct bintime *)CMSG_DATA(cmsghdr); + /* + * bintime documentation is at http://phk.freebsd.dk/pubs/timecounter.pdf + */ + nts.l_i = btp->sec + JAN_1970; + nts.l_uf = (u_int32)(btp->frac >> 32); + if (sys_tick > measured_tick && + sys_tick > 1e-9) { + ticks = (unsigned long)(nts.l_uf / (unsigned long)(sys_tick * FRAC)); + nts.l_uf = (unsigned long)(ticks * (unsigned long)(sys_tick * FRAC)); + } + DPRINTF(4, ("fetch_timestamp: system bintime network time stamp: %ld.%09lu\n", + btp->sec, (unsigned long)((nts.l_uf / FRAC) * 1e9))); + break; +#endif /* HAVE_BINTIME */ +#ifdef HAVE_TIMESTAMPNS + case SCM_TIMESTAMPNS: + tsp = (struct timespec *)CMSG_DATA(cmsghdr); + if (sys_tick > measured_tick && + sys_tick > 1e-9) { + ticks = (unsigned long)((tsp->tv_nsec * 1e-9) / + sys_tick); + tsp->tv_nsec = (long)(ticks * 1e9 * + sys_tick); + } + DPRINTF(4, ("fetch_timestamp: system nsec network time stamp: %ld.%09ld\n", + tsp->tv_sec, tsp->tv_nsec)); + nts = tspec_stamp_to_lfp(*tsp); + break; +#endif /* HAVE_TIMESTAMPNS */ +#ifdef HAVE_TIMESTAMP + case SCM_TIMESTAMP: + tvp = (struct timeval *)CMSG_DATA(cmsghdr); + if (sys_tick > measured_tick && + sys_tick > 1e-6) { + ticks = (unsigned long)((tvp->tv_usec * 1e-6) / + sys_tick); + tvp->tv_usec = (long)(ticks * 1e6 * + sys_tick); + } + DPRINTF(4, ("fetch_timestamp: system usec network time stamp: %jd.%06ld\n", + (intmax_t)tvp->tv_sec, (long)tvp->tv_usec)); + nts = tval_stamp_to_lfp(*tvp); + break; +#endif /* HAVE_TIMESTAMP */ } -#endif + fuzz = ntp_random() * 2. / FRAC * sys_fuzz; + DTOLFP(fuzz, &lfpfuzz); + L_ADD(&nts, &lfpfuzz); +#ifdef DEBUG_TIMING + dts = ts; + L_SUB(&dts, &nts); + collect_timing(rb, "input processing delay", 1, + &dts); + DPRINTF(4, ("fetch_timestamp: timestamp delta: %s (incl. fuzz)\n", + lfptoa(&dts, 9))); +#endif /* DEBUG_TIMING */ ts = nts; /* network time stamp */ break; - } +#endif /* HAVE_BINTIME || HAVE_TIMESTAMPNS || HAVE_TIMESTAMP */ + default: - DPRINTF(4, ("fetch_timestamp: skipping control message 0x%x\n", cmsghdr->cmsg_type)); - break; + DPRINTF(4, ("fetch_timestamp: skipping control message 0x%x\n", + cmsghdr->cmsg_type)); } cmsghdr = CMSG_NXTHDR(msghdr, cmsghdr); } -#endif return ts; } -#endif +#endif /* HAVE_PACKET_TIMESTAMP */ + /* * Routine to read the network NTP packets for a specific interface @@ -2976,15 +3430,19 @@ static l_fp * read it again or go on to the next one if no bytes returned */ static inline int -read_network_packet(SOCKET fd, struct interface *itf, l_fp ts) +read_network_packet( + SOCKET fd, + struct interface * itf, + l_fp ts + ) { GETSOCKNAME_SOCKLEN_TYPE fromlen; int buflen; register struct recvbuf *rb; -#ifdef HAVE_TIMESTAMP +#ifdef HAVE_PACKET_TIMESTAMP struct msghdr msghdr; struct iovec iovec; - char control[TIMESTAMP_CTLMSGBUF_SIZE]; /* pick up control messages */ + char control[CMSG_BUFSIZE]; #endif /* @@ -2995,41 +3453,39 @@ read_network_packet(SOCKET fd, struct interface *itf, l_fp ts) */ rb = get_free_recv_buffer(); - - if (rb == NULL || itf->ignore_packets == ISC_TRUE) - { + if (NULL == rb || itf->ignore_packets) { char buf[RX_BUFF_SIZE]; - struct sockaddr_storage from; + sockaddr_u from; + if (rb != NULL) freerecvbuf(rb); fromlen = sizeof(from); buflen = recvfrom(fd, buf, sizeof(buf), 0, - (struct sockaddr*)&from, &fromlen); + &from.sa, &fromlen); DPRINTF(4, ("%s on (%lu) fd=%d from %s\n", - (itf->ignore_packets == ISC_TRUE) ? "ignore" : "drop", - free_recvbuffs(), fd, - stoa(&from))); - if (itf->ignore_packets == ISC_TRUE) + (itf->ignore_packets) + ? "ignore" + : "drop", + free_recvbuffs(), fd, stoa(&from))); + if (itf->ignore_packets) packets_ignored++; else packets_dropped++; return (buflen); } - fromlen = sizeof(struct sockaddr_storage); + fromlen = sizeof(rb->recv_srcadr); -#ifndef HAVE_TIMESTAMP - rb->recv_length = recvfrom(fd, - (char *)&rb->recv_space, - sizeof(rb->recv_space), 0, - (struct sockaddr *)&rb->recv_srcadr, - &fromlen); +#ifndef HAVE_PACKET_TIMESTAMP + rb->recv_length = recvfrom(fd, (char *)&rb->recv_space, + sizeof(rb->recv_space), 0, + &rb->recv_srcadr.sa, &fromlen); #else - iovec.iov_base = (void *)&rb->recv_space; + iovec.iov_base = &rb->recv_space; iovec.iov_len = sizeof(rb->recv_space); - msghdr.msg_name = (void *)&rb->recv_srcadr; - msghdr.msg_namelen = sizeof(rb->recv_srcadr); + msghdr.msg_name = &rb->recv_srcadr; + msghdr.msg_namelen = fromlen; msghdr.msg_iov = &iovec; msghdr.msg_iovlen = 1; msghdr.msg_control = (void *)&control; @@ -3040,38 +3496,48 @@ read_network_packet(SOCKET fd, struct interface *itf, l_fp ts) buflen = rb->recv_length; - if (buflen == 0 || (buflen == -1 && - (errno==EWOULDBLOCK + if (buflen == 0 || (buflen == -1 && + (EWOULDBLOCK == errno #ifdef EAGAIN - || errno==EAGAIN + || EAGAIN == errno #endif - ))) { + ))) { freerecvbuf(rb); return (buflen); - } - else if (buflen < 0) - { - netsyslog(LOG_ERR, "recvfrom(%s) fd=%d: %m", - stoa(&rb->recv_srcadr), fd); - DPRINTF(5, ("read_network_packet: fd=%d dropped (bad recvfrom)\n", fd)); + } else if (buflen < 0) { + msyslog(LOG_ERR, "recvfrom(%s) fd=%d: %m", + stoa(&rb->recv_srcadr), fd); + DPRINTF(5, ("read_network_packet: fd=%d dropped (bad recvfrom)\n", + fd)); freerecvbuf(rb); return (buflen); } -#ifdef DEBUG - if (debug > 2) { - if(rb->recv_srcadr.ss_family == AF_INET) - printf("read_network_packet: fd=%d length %d from %08lx %s\n", - fd, buflen, - (u_long)ntohl(((struct sockaddr_in*)&rb->recv_srcadr)->sin_addr.s_addr) & - 0x00000000ffffffff, - stoa(&rb->recv_srcadr)); - else - printf("read_network_packet: fd=%d length %d from %s\n", - fd, buflen, - stoa(&rb->recv_srcadr)); + DPRINTF(3, ("read_network_packet: fd=%d length %d from %s\n", + fd, buflen, stoa(&rb->recv_srcadr))); + + /* + ** Bug 2672: Some OSes (MacOSX and Linux) don't block spoofed ::1 + */ + + if (AF_INET6 == itf->family) { + DPRINTF(2, ("Got an IPv6 packet, from <%s> (%d) to <%s> (%d)\n", + stoa(&rb->recv_srcadr), + IN6_IS_ADDR_LOOPBACK(PSOCK_ADDR6(&rb->recv_srcadr)), + stoa(&itf->sin), + !IN6_IS_ADDR_LOOPBACK(PSOCK_ADDR6(&itf->sin)) + )); + + if ( IN6_IS_ADDR_LOOPBACK(PSOCK_ADDR6(&rb->recv_srcadr)) + && !IN6_IS_ADDR_LOOPBACK(PSOCK_ADDR6(&itf->sin)) + ) { + packets_dropped++; + DPRINTF(2, ("DROPPING that packet\n")); + freerecvbuf(rb); + return buflen; + } + DPRINTF(2, ("processing that packet\n")); } -#endif /* * Got one. Mark how and when it got here, @@ -3079,8 +3545,9 @@ read_network_packet(SOCKET fd, struct interface *itf, l_fp ts) */ rb->dstadr = itf; rb->fd = fd; -#ifdef HAVE_TIMESTAMP - ts = fetch_timestamp(rb, &msghdr, ts); /* pick up a network time stamp if possible */ +#ifdef HAVE_PACKET_TIMESTAMP + /* pick up a network time stamp if possible */ + ts = fetch_timestamp(rb, &msghdr, ts); #endif rb->recv_time = ts; rb->receiver = receive; @@ -3093,35 +3560,106 @@ read_network_packet(SOCKET fd, struct interface *itf, l_fp ts) } /* - * input_handler - receive packets asynchronously + * attempt to handle io (select()/signaled IO) */ void +io_handler(void) +{ +# ifndef HAVE_SIGNALED_IO + fd_set rdfdes; + int nfound; + + /* + * Use select() on all on all input fd's for unlimited + * time. select() will terminate on SIGALARM or on the + * reception of input. Using select() means we can't do + * robust signal handling and we get a potential race + * between checking for alarms and doing the select(). + * Mostly harmless, I think. + */ + /* + * On VMS, I suspect that select() can't be interrupted + * by a "signal" either, so I take the easy way out and + * have select() time out after one second. + * System clock updates really aren't time-critical, + * and - lacking a hardware reference clock - I have + * yet to learn about anything else that is. + */ + rdfdes = activefds; +# if !defined(VMS) && !defined(SYS_VXWORKS) + nfound = select(maxactivefd + 1, &rdfdes, NULL, + NULL, NULL); +# else /* VMS, VxWorks */ + /* make select() wake up after one second */ + { + struct timeval t1; + + t1.tv_sec = 1; + t1.tv_usec = 0; + nfound = select(maxactivefd + 1, + &rdfdes, NULL, NULL, + &t1); + } +# endif /* VMS, VxWorks */ + if (nfound > 0) { + l_fp ts; + + get_systime(&ts); + + input_handler(&ts); + } else if (nfound == -1 && errno != EINTR) { + msyslog(LOG_ERR, "select() error: %m"); + } +# ifdef DEBUG + else if (debug > 4) { + msyslog(LOG_DEBUG, "select(): nfound=%d, error: %m", nfound); + } else { + DPRINTF(3, ("select() returned %d: %m\n", nfound)); + } +# endif /* DEBUG */ +# else /* HAVE_SIGNALED_IO */ + wait_for_signal(); +# endif /* HAVE_SIGNALED_IO */ +} + +/* + * input_handler - receive packets asynchronously + */ +static void input_handler( - l_fp *cts + l_fp * cts ) { - - int buflen; - int n; - int doing; - SOCKET fd; - struct timeval tvzero; - l_fp ts; /* Timestamp at BOselect() gob */ + int buflen; + int n; + u_int idx; + int doing; + SOCKET fd; + blocking_child *c; + struct timeval tvzero; + l_fp ts; /* Timestamp at BOselect() gob */ #ifdef DEBUG_TIMING - l_fp ts_e; /* Timestamp at EOselect() gob */ + l_fp ts_e; /* Timestamp at EOselect() gob */ #endif - fd_set fds; - int select_count = 0; - struct interface *interface; -#if defined(HAS_ROUTING_SOCKET) - struct asyncio_reader *asyncio_reader; + fd_set fds; + size_t select_count; + endpt * ep; +#ifdef REFCLOCK + struct refclockio *rp; + int saved_errno; + const char * clk; +#endif +#ifdef HAS_ROUTING_SOCKET + struct asyncio_reader * asyncio_reader; + struct asyncio_reader * next_asyncio_reader; #endif handler_calls++; + select_count = 0; /* * If we have something to do, freeze a timestamp. - * See below for the other cases (nothing (left) to do or error) + * See below for the other cases (nothing left to do or error) */ ts = *cts; @@ -3132,33 +3670,43 @@ input_handler( fds = activefds; tvzero.tv_sec = tvzero.tv_usec = 0; - n = select(maxactivefd+1, &fds, (fd_set *)0, (fd_set *)0, &tvzero); + n = select(maxactivefd + 1, &fds, (fd_set *)0, (fd_set *)0, + &tvzero); /* * If there are no packets waiting just return */ - if (n < 0) - { + if (n < 0) { int err = errno; + int j, b, prior; /* * extended FAU debugging output */ if (err != EINTR) - netsyslog(LOG_ERR, - "select(%d, %s, 0L, 0L, &0.0) error: %m", - maxactivefd+1, - fdbits(maxactivefd, &activefds)); - if (err == EBADF) { - int j, b; - fds = activefds; - for (j = 0; j <= maxactivefd; j++) - if ((FD_ISSET(j, &fds) && (read(j, &b, 0) == -1))) - netsyslog(LOG_ERR, "Bad file descriptor %d", j); + msyslog(LOG_ERR, + "select(%d, %s, 0L, 0L, &0.0) error: %m", + maxactivefd + 1, + fdbits(maxactivefd, &activefds)); + if (err != EBADF) + goto ih_return; + for (j = 0, prior = 0; j <= maxactivefd; j++) { + if (FD_ISSET(j, &activefds)) { + if (-1 != read(j, &b, 0)) { + prior = j; + continue; + } + msyslog(LOG_ERR, + "Removing bad file descriptor %d from select set", + j); + FD_CLR(j, &activefds); + if (j == maxactivefd) + maxactivefd = prior; + } } - return; + goto ih_return; } else if (n == 0) - return; + goto ih_return; ++handler_pkts; @@ -3167,55 +3715,64 @@ input_handler( * Check out the reference clocks first, if any */ - if (refio != NULL) - { - register struct refclockio *rp; - - for (rp = refio; rp != NULL; rp = rp->next) - { + if (refio != NULL) { + for (rp = refio; rp != NULL; rp = rp->next) { fd = rp->fd; - if (FD_ISSET(fd, &fds)) - { + if (!FD_ISSET(fd, &fds)) + continue; + ++select_count; + buflen = read_refclock_packet(fd, rp, ts); + /* + * The first read must succeed after select() + * indicates readability, or we've reached + * a permanent EOF. http://bugs.ntp.org/1732 + * reported ntpd munching CPU after a USB GPS + * was unplugged because select was indicating + * EOF but ntpd didn't remove the descriptor + * from the activefds set. + */ + if (buflen < 0 && EAGAIN != errno) { + saved_errno = errno; + clk = refnumtoa(&rp->srcclock->srcadr); + errno = saved_errno; + msyslog(LOG_ERR, "%s read: %m", clk); + maintain_activefds(fd, TRUE); + } else if (0 == buflen) { + clk = refnumtoa(&rp->srcclock->srcadr); + msyslog(LOG_ERR, "%s read EOF", clk); + maintain_activefds(fd, TRUE); + } else { + /* drain any remaining refclock input */ do { - ++select_count; buflen = read_refclock_packet(fd, rp, ts); } while (buflen > 0); - - } /* End if (FD_ISSET(fd, &fds)) */ - } /* End for (rp = refio; rp != 0 && n > 0; rp = rp->next) */ - } /* End if (refio != 0) */ - + } + } + } #endif /* REFCLOCK */ /* * Loop through the interfaces looking for data to read. */ - for (interface = ISC_LIST_TAIL(inter_list); - interface != NULL; - interface = ISC_LIST_PREV(interface, link)) - { - for (doing = 0; (doing < 2); doing++) - { - if (doing == 0) - { - fd = interface->fd; - } - else - { - if (!(interface->flags & INT_BCASTOPEN)) - break; - fd = interface->bfd; + for (ep = ep_list; ep != NULL; ep = ep->elink) { + for (doing = 0; doing < 2; doing++) { + if (!doing) { + fd = ep->fd; + } else { + if (!(ep->flags & INT_BCASTOPEN)) + break; + fd = ep->bfd; } - if (fd < 0) continue; + if (fd < 0) + continue; if (FD_ISSET(fd, &fds)) - { do { ++select_count; - buflen = read_network_packet(fd, interface, ts); + buflen = read_network_packet( + fd, ep, ts); } while (buflen > 0); - } - /* Check more interfaces */ + /* Check more interfaces */ } } @@ -3223,36 +3780,45 @@ input_handler( /* * scan list of asyncio readers - currently only used for routing sockets */ - asyncio_reader = ISC_LIST_TAIL(asyncio_reader_list); + asyncio_reader = asyncio_reader_list; - while (asyncio_reader != NULL) - { - struct asyncio_reader *next = ISC_LIST_PREV(asyncio_reader, link); + while (asyncio_reader != NULL) { + /* callback may unlink and free asyncio_reader */ + next_asyncio_reader = asyncio_reader->link; if (FD_ISSET(asyncio_reader->fd, &fds)) { ++select_count; - asyncio_reader->receiver(asyncio_reader); + (*asyncio_reader->receiver)(asyncio_reader); } - asyncio_reader = next; + asyncio_reader = next_asyncio_reader; } #endif /* HAS_ROUTING_SOCKET */ - + /* - * Done everything from that select. + * Check for a response from a blocking child */ + for (idx = 0; idx < blocking_children_alloc; idx++) { + c = blocking_children[idx]; + if (NULL == c || -1 == c->resp_read_pipe) + continue; + if (FD_ISSET(c->resp_read_pipe, &fds)) { + select_count++; + process_blocking_resp(c); + } + } /* + * Done everything from that select. * If nothing to do, just return. * If an error occurred, complain and return. */ - if (select_count == 0) /* We really had nothing to do */ - { + if (select_count == 0) { /* We really had nothing to do */ #ifdef DEBUG if (debug) - netsyslog(LOG_DEBUG, "input_handler: select() returned 0"); -#endif - return; + msyslog(LOG_DEBUG, "input_handler: select() returned 0"); +#endif /* DEBUG */ + goto ih_return; } - /* We've done our work */ + /* We've done our work */ #ifdef DEBUG_TIMING get_systime(&ts_e); /* @@ -3263,44 +3829,109 @@ input_handler( L_SUB(&ts_e, &ts); collect_timing(NULL, "input handler", 1, &ts_e); if (debug > 3) - netsyslog(LOG_INFO, "input_handler: Processed a gob of fd's in %s msec", lfptoms(&ts_e, 6)); -#endif - /* just bail. */ + msyslog(LOG_DEBUG, + "input_handler: Processed a gob of fd's in %s msec", + lfptoms(&ts_e, 6)); +#endif /* DEBUG_TIMING */ + /* We're done... */ + ih_return: return; } +#endif /* !HAVE_IO_COMPLETION_PORT */ + +/* + * find an interface suitable for the src address + */ +endpt * +select_peerinterface( + struct peer * peer, + sockaddr_u * srcadr, + endpt * dstadr + ) +{ + endpt *ep; +#ifndef SIM + endpt *wild; + + wild = ANY_INTERFACE_CHOOSE(srcadr); + + /* + * Initialize the peer structure and dance the interface jig. + * Reference clocks step the loopback waltz, the others + * squaredance around the interface list looking for a buddy. If + * the dance peters out, there is always the wildcard interface. + * This might happen in some systems and would preclude proper + * operation with public key cryptography. + */ + if (ISREFCLOCKADR(srcadr)) { + ep = loopback_interface; + } else if (peer->cast_flags & + (MDF_BCLNT | MDF_ACAST | MDF_MCAST | MDF_BCAST)) { + ep = findbcastinter(srcadr); + if (ep != NULL) + DPRINTF(4, ("Found *-cast interface %s for address %s\n", + stoa(&ep->sin), stoa(srcadr))); + else + DPRINTF(4, ("No *-cast local address found for address %s\n", + stoa(srcadr))); + } else { + ep = dstadr; + if (NULL == ep) + ep = wild; + } + /* + * If it is a multicast address, findbcastinter() may not find + * it. For unicast, we get to find the interface when dstadr is + * given to us as the wildcard (ANY_INTERFACE_CHOOSE). Either + * way, try a little harder. + */ + if (wild == ep) + ep = findinterface(srcadr); + /* + * we do not bind to the wildcard interfaces for output + * as our (network) source address would be undefined and + * crypto will not work without knowing the own transmit address + */ + if (ep != NULL && INT_WILDCARD & ep->flags) + if (!accept_wildcard_if_for_winnt) + ep = NULL; +#else /* SIM follows */ + ep = loopback_interface; #endif + return ep; +} + + /* * findinterface - find local interface corresponding to address */ -struct interface * +endpt * findinterface( - struct sockaddr_storage *addr + sockaddr_u *addr ) { - struct interface *interface; - - interface = findlocalinterface(addr, INT_WILDCARD); + endpt *iface; - if (interface == NULL) - { + iface = findlocalinterface(addr, INT_WILDCARD, 0); + + if (NULL == iface) { DPRINTF(4, ("Found no interface for address %s - returning wildcard\n", stoa(addr))); - return (ANY_INTERFACE_CHOOSE(addr)); - } - else - { + iface = ANY_INTERFACE_CHOOSE(addr); + } else DPRINTF(4, ("Found interface #%d %s for address %s\n", - interface->ifnum, interface->name, stoa(addr))); + iface->ifnum, iface->name, stoa(addr))); - return (interface); - } + return iface; } /* - * findlocalinterface - find local interface index corresponding to address + * findlocalinterface - find local interface corresponding to addr, + * which does not have any of flags set. If bast is nonzero, addr is + * a broadcast address. * * This code attempts to find the local sending address for an outgoing * address by connecting a new socket to destinationaddress:NTP_PORT @@ -3312,246 +3943,307 @@ findinterface( * logic in ntpd which would be a silly and really unportable thing to do. * */ -static struct interface * +static endpt * findlocalinterface( - struct sockaddr_storage *addr, - int flags + sockaddr_u * addr, + int flags, + int bcast ) { - SOCKET s; - int rtn; - struct sockaddr_storage saddr; - GETSOCKNAME_SOCKLEN_TYPE saddrlen = SOCKLEN(addr); - struct interface *iface; + GETSOCKNAME_SOCKLEN_TYPE sockaddrlen; + endpt * iface; + sockaddr_u saddr; + SOCKET s; + int rtn; + int on; DPRINTF(4, ("Finding interface for addr %s in list of addresses\n", stoa(addr))); - memset(&saddr, 0, sizeof(saddr)); - saddr.ss_family = addr->ss_family; - if(addr->ss_family == AF_INET) { - memcpy(&((struct sockaddr_in*)&saddr)->sin_addr, &((struct sockaddr_in*)addr)->sin_addr, sizeof(struct in_addr)); - ((struct sockaddr_in*)&saddr)->sin_port = htons(NTP_PORT); - } -#ifdef INCLUDE_IPV6_SUPPORT - else if(addr->ss_family == AF_INET6) { - memcpy(&((struct sockaddr_in6*)&saddr)->sin6_addr, &((struct sockaddr_in6*)addr)->sin6_addr, sizeof(struct in6_addr)); - ((struct sockaddr_in6*)&saddr)->sin6_port = htons(NTP_PORT); -# ifdef ISC_PLATFORM_HAVESCOPEID - ((struct sockaddr_in6*)&saddr)->sin6_scope_id = ((struct sockaddr_in6*)addr)->sin6_scope_id; -# endif - } -#endif - - s = socket(addr->ss_family, SOCK_DGRAM, 0); - if (s == INVALID_SOCKET) + s = socket(AF(addr), SOCK_DGRAM, 0); + if (INVALID_SOCKET == s) return NULL; - rtn = connect(s, (struct sockaddr *)&saddr, SOCKLEN(&saddr)); -#ifndef SYS_WINNT - if (rtn < 0) -#else - if (rtn == SOCKET_ERROR) -#endif - { + /* + * If we are looking for broadcast interface we need to set this + * socket to allow broadcast + */ + if (bcast) { + on = 1; + if (SOCKET_ERROR == setsockopt(s, SOL_SOCKET, + SO_BROADCAST, + (char *)&on, + sizeof(on))) { + closesocket(s); + return NULL; + } + } + + rtn = connect(s, &addr->sa, SOCKLEN(addr)); + if (SOCKET_ERROR == rtn) { closesocket(s); return NULL; } - rtn = getsockname(s, (struct sockaddr *)&saddr, &saddrlen); + sockaddrlen = sizeof(saddr); + rtn = getsockname(s, &saddr.sa, &sockaddrlen); closesocket(s); -#ifndef SYS_WINNT - if (rtn < 0) -#else - if (rtn == SOCKET_ERROR) -#endif + if (SOCKET_ERROR == rtn) return NULL; - DPRINTF(4, ("findlocalinterface: kernel maps %s to %s\n", stoa(addr), stoa(&saddr))); - + DPRINTF(4, ("findlocalinterface: kernel maps %s to %s\n", + stoa(addr), stoa(&saddr))); + iface = getinterface(&saddr, flags); - /* Don't both with ignore interfaces */ - if (iface != NULL && iface->ignore_packets == ISC_TRUE) - { - return NULL; - } - else - { - return iface; - } + /* + * if we didn't find an exact match on saddr, find the closest + * available local address. This handles the case of the + * address suggested by the kernel being excluded by nic rules + * or the user's -I and -L options to ntpd. + * See http://bugs.ntp.org/1184 and http://bugs.ntp.org/1683 + * for more background. + */ + if (NULL == iface || iface->ignore_packets) + iface = findclosestinterface(&saddr, + flags | INT_LOOPBACK); + + /* Don't use an interface which will ignore replies */ + if (iface != NULL && iface->ignore_packets) + iface = NULL; + + return iface; } + /* - * fetch an interface structure the matches the - * address is has the given flags not set + * findclosestinterface + * + * If there are -I/--interface or -L/novirtualips command-line options, + * or "nic" or "interface" rules in ntp.conf, findlocalinterface() may + * find the kernel's preferred local address for a given peer address is + * administratively unavailable to ntpd, and punt to this routine's more + * expensive search. + * + * Find the numerically closest local address to the one connect() + * suggested. This matches an address on the same subnet first, as + * needed by Bug 1184, and provides a consistent choice if there are + * multiple feasible local addresses, regardless of the order ntpd + * enumerated them. */ -static struct interface * -getinterface(struct sockaddr_storage *addr, int flags) +endpt * +findclosestinterface( + sockaddr_u * addr, + int flags + ) { - struct interface *interface = find_addr_in_list(addr); + endpt * ep; + endpt * winner; + sockaddr_u addr_dist; + sockaddr_u min_dist; + + ZERO_SOCK(&min_dist); + winner = NULL; + + for (ep = ep_list; ep != NULL; ep = ep->elink) { + if (ep->ignore_packets || + AF(addr) != ep->family || + flags & ep->flags) + continue; - if (interface != NULL && interface->flags & flags) - { - return NULL; + calc_addr_distance(&addr_dist, addr, &ep->sin); + if (NULL == winner || + -1 == cmp_addr_distance(&addr_dist, &min_dist)) { + min_dist = addr_dist; + winner = ep; + } } + if (NULL == winner) + DPRINTF(4, ("findclosestinterface(%s) failed\n", + stoa(addr))); else - { - return interface; - } + DPRINTF(4, ("findclosestinterface(%s) -> %s\n", + stoa(addr), stoa(&winner->sin))); + + return winner; } + /* - * findlocalcastinterface - find local *cast interface index corresponding to address - * depending on the flags passed + * calc_addr_distance - calculate the distance between two addresses, + * the absolute value of the difference between + * the addresses numerically, stored as an address. */ -static struct interface * -findlocalcastinterface( - struct sockaddr_storage *addr, int flags +static void +calc_addr_distance( + sockaddr_u * dist, + const sockaddr_u * a1, + const sockaddr_u * a2 ) { - struct interface *interface; - struct interface *nif = NULL; -#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT - isc_boolean_t want_linklocal; -#endif - - /* - * see how kernel maps the mcast address - */ - nif = findlocalinterface(addr, 0); + u_int32 a1val; + u_int32 a2val; + u_int32 v4dist; + int found_greater; + int a1_greater; + int i; + + NTP_REQUIRE(AF(a1) == AF(a2)); + + ZERO_SOCK(dist); + AF(dist) = AF(a1); + + /* v4 can be done a bit simpler */ + if (IS_IPV4(a1)) { + a1val = SRCADR(a1); + a2val = SRCADR(a2); + v4dist = (a1val > a2val) + ? a1val - a2val + : a2val - a1val; + SET_ADDR4(dist, v4dist); - if (nif) { - DPRINTF(2, ("findlocalcastinterface: kernel recommends interface #%d %s\n", nif->ifnum, nif->name)); - return nif; + return; } -#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT - want_linklocal = ISC_FALSE; - if (addr_ismulticast(addr) && flags == INT_MULTICAST) - { - if (IN6_IS_ADDR_MC_LINKLOCAL(&((struct sockaddr_in6*)addr)->sin6_addr)) - { - want_linklocal = ISC_TRUE; + found_greater = FALSE; + a1_greater = FALSE; /* suppress pot. uninit. warning */ + for (i = 0; i < (int)sizeof(NSRCADR6(a1)); i++) { + if (!found_greater && + NSRCADR6(a1)[i] != NSRCADR6(a2)[i]) { + found_greater = TRUE; + a1_greater = (NSRCADR6(a1)[i] > NSRCADR6(a2)[i]); } - else if (IN6_IS_ADDR_MC_SITELOCAL(&((struct sockaddr_in6*)addr)->sin6_addr)) - { - want_linklocal = ISC_TRUE; + if (!found_greater) { + NSRCADR6(dist)[i] = 0; + } else { + if (a1_greater) + NSRCADR6(dist)[i] = NSRCADR6(a1)[i] - + NSRCADR6(a2)[i]; + else + NSRCADR6(dist)[i] = NSRCADR6(a2)[i] - + NSRCADR6(a1)[i]; } } -#endif +} - for (interface = ISC_LIST_HEAD(inter_list); - interface != NULL; - interface = ISC_LIST_NEXT(interface, link)) - { - /* use only allowed addresses */ - if (interface->ignore_packets == ISC_TRUE) - continue; - /* Skip the loopback and wildcard addresses */ - if (interface->flags & (INT_LOOPBACK|INT_WILDCARD)) - continue; - - /* Skip if different family */ - if(interface->sin.ss_family != addr->ss_family) - continue; +/* + * cmp_addr_distance - compare two address distances, returning -1, 0, + * 1 to indicate their relationship. + */ +static int +cmp_addr_distance( + const sockaddr_u * d1, + const sockaddr_u * d2 + ) +{ + int i; - /* Is this it one of these based on flags? */ - if (!(interface->flags & flags)) - continue; + NTP_REQUIRE(AF(d1) == AF(d2)); - /* for IPv6 multicast check the address for linklocal */ -#ifdef INCLUDE_IPV6_MULTICAST_SUPPORT - if (flags == INT_MULTICAST && interface->sin.ss_family == AF_INET6 && - (IN6_IS_ADDR_LINKLOCAL(&((struct sockaddr_in6*)&interface->sin)->sin6_addr)) - && want_linklocal == ISC_TRUE) - { - nif = interface; - break; - } - /* If we want a linklocal address and this isn't it, skip */\ - if (want_linklocal == ISC_TRUE) - continue; -#endif - /* Otherwise just look for the flag */ - if((interface->flags & flags)) - { - nif = interface; - break; - } - } -#ifdef DEBUG - if (debug > 2) - { - if (nif) - printf("findlocalcastinterface: found interface #%d %s\n", nif->ifnum, nif->name); + if (IS_IPV4(d1)) { + if (SRCADR(d1) < SRCADR(d2)) + return -1; + else if (SRCADR(d1) == SRCADR(d2)) + return 0; else - printf("findlocalcastinterface: no interface found for %s flags 0x%x\n", stoa(addr), flags); + return 1; } -#endif - return (nif); + + for (i = 0; i < (int)sizeof(NSRCADR6(d1)); i++) { + if (NSRCADR6(d1)[i] < NSRCADR6(d2)[i]) + return -1; + else if (NSRCADR6(d1)[i] > NSRCADR6(d2)[i]) + return 1; + } + + return 0; } + + +/* + * fetch an interface structure the matches the + * address and has the given flags NOT set + */ +endpt * +getinterface( + sockaddr_u * addr, + u_int32 flags + ) +{ + endpt *iface; + + iface = find_addr_in_list(addr); + + if (iface != NULL && (iface->flags & flags)) + iface = NULL; + + return iface; +} + + /* * findbcastinter - find broadcast interface corresponding to address */ -struct interface * +endpt * findbcastinter( - struct sockaddr_storage *addr + sockaddr_u *addr ) { + endpt * iface; + + iface = NULL; #if !defined(MPE) && (defined(SIOCGIFCONF) || defined(SYS_WINNT)) - struct interface *interface; - - DPRINTF(4, ("Finding broadcast/multicast interface for addr %s in list of addresses\n", stoa(addr))); - interface = findlocalinterface(addr, INT_LOOPBACK|INT_WILDCARD); - - if (interface != NULL) - { - DPRINTF(4, ("Found bcast-/mcast- interface index #%d %s\n", interface->ifnum, interface->name)); - return interface; + iface = findlocalinterface(addr, INT_LOOPBACK | INT_WILDCARD, + 1); + if (iface != NULL) { + DPRINTF(4, ("Easily found bcast-/mcast- interface index #%d %s\n", + iface->ifnum, iface->name)); + return iface; } - /* plan B - try to find something reasonable in our lists in case kernel lookup doesn't help */ + /* + * plan B - try to find something reasonable in our lists in + * case kernel lookup doesn't help + */ + for (iface = ep_list; iface != NULL; iface = iface->elink) { + if (iface->flags & INT_WILDCARD) + continue; - for (interface = ISC_LIST_HEAD(inter_list); - interface != NULL; - interface = ISC_LIST_NEXT(interface, link)) - { - if (interface->flags & INT_WILDCARD) - continue; - /* Don't bother with ignored interfaces */ - if (interface->ignore_packets == ISC_TRUE) + if (iface->ignore_packets) continue; - + /* * First look if this is the correct family */ - if(interface->sin.ss_family != addr->ss_family) - continue; + if(AF(&iface->sin) != AF(addr)) + continue; /* Skip the loopback addresses */ - if (interface->flags & INT_LOOPBACK) + if (iface->flags & INT_LOOPBACK) continue; /* - * If we are looking to match a multicast address grab it. + * If we are looking to match a multicast address and + * this interface is one... */ - if (addr_ismulticast(addr) == ISC_TRUE && interface->flags & INT_MULTICAST) - { + if (addr_ismulticast(addr) + && (iface->flags & INT_MULTICAST)) { #ifdef INCLUDE_IPV6_SUPPORT - if(addr->ss_family == AF_INET6) { - /* Only use link-local address for link-scope mcast */ - if(IN6_IS_ADDR_MC_LINKLOCAL(&((struct sockaddr_in6*)addr)->sin6_addr) && - !IN6_IS_ADDR_LINKLOCAL(&((struct sockaddr_in6*)&interface->sin)->sin6_addr)) { - continue; - } - } + /* + * ...it is the winner unless we're looking for + * an interface to use for link-local multicast + * and its address is not link-local. + */ + if (IS_IPV6(addr) + && IN6_IS_ADDR_MC_LINKLOCAL(PSOCK_ADDR6(addr)) + && !IN6_IS_ADDR_LINKLOCAL(PSOCK_ADDR6(&iface->sin))) + continue; #endif break; } @@ -3562,34 +4254,35 @@ findbcastinter( * address or the network portion of the IP address. * Sloppy. */ - if(addr->ss_family == AF_INET) { - if (SOCKCMP(&interface->bcast, addr)) { + if (IS_IPV4(addr)) { + if (SOCK_EQ(&iface->bcast, addr)) break; - } - if ((NSRCADR(&interface->sin) & - NSRCADR(&interface->mask)) == (NSRCADR(addr) & - NSRCADR(&interface->mask))) + + if ((NSRCADR(&iface->sin) & NSRCADR(&iface->mask)) + == (NSRCADR(addr) & NSRCADR(&iface->mask))) break; } #ifdef INCLUDE_IPV6_SUPPORT - else if(addr->ss_family == AF_INET6) { - if (SOCKCMP(&interface->bcast, addr)) { + else if (IS_IPV6(addr)) { + if (SOCK_EQ(&iface->bcast, addr)) break; - } - if (SOCKCMP(netof(&interface->sin), netof(addr))) { + + if (SOCK_EQ(netof(&iface->sin), netof(addr))) break; - } } #endif } #endif /* SIOCGIFCONF */ - if (interface == NULL) { - DPRINTF(4, ("No bcast interface found for %s\n", stoa(addr))); - return ANY_INTERFACE_CHOOSE(addr); + if (NULL == iface) { + DPRINTF(4, ("No bcast interface found for %s\n", + stoa(addr))); + iface = ANY_INTERFACE_CHOOSE(addr); } else { - DPRINTF(4, ("Found bcast-/mcast- interface index #%d %s\n", interface->ifnum, interface->name)); - return interface; + DPRINTF(4, ("Found bcast-/mcast- interface index #%d %s\n", + iface->ifnum, iface->name)); } + + return iface; } @@ -3622,21 +4315,20 @@ io_addclock( ) { BLOCKIO(); + /* * Stuff the I/O structure in the list and mark the descriptor - * in use. There is a harmless (I hope) race condition here. + * in use. There is a harmless (I hope) race condition here. */ - rio->next = refio; + rio->active = TRUE; # ifdef HAVE_SIGNALED_IO - if (init_clock_sig(rio)) - { + if (init_clock_sig(rio)) { UNBLOCKIO(); return 0; } # elif defined(HAVE_IO_COMPLETION_PORT) - if (io_completion_port_add_clock_io(rio)) - { + if (io_completion_port_add_clock_io(rio)) { UNBLOCKIO(); return 0; } @@ -3645,9 +4337,9 @@ io_addclock( /* * enqueue */ - refio = rio; + LINK_SLIST(refio, rio, next); - /* + /* * register fd */ add_fd_to_list(rio->fd, FD_TYPE_FILE); @@ -3656,6 +4348,7 @@ io_addclock( return 1; } + /* * io_closeclock - close the clock in the I/O structure given */ @@ -3664,267 +4357,270 @@ io_closeclock( struct refclockio *rio ) { + struct refclockio *unlinked; + BLOCKIO(); + /* * Remove structure from the list */ - if (refio == rio) - { - refio = rio->next; - } - else - { - register struct refclockio *rp; - - for (rp = refio; rp != NULL; rp = rp->next) - if (rp->next == rio) - { - rp->next = rio->next; - break; - } - - if (rp == NULL) { - UNBLOCKIO(); - return; - } + rio->active = FALSE; + UNLINK_SLIST(unlinked, refio, rio, next, struct refclockio); + if (NULL != unlinked) { + purge_recv_buffers_for_fd(rio->fd); + /* + * Close the descriptor. + */ + close_and_delete_fd_from_list(rio->fd); } + rio->fd = -1; - /* - * Close the descriptor. - */ - close_and_delete_fd_from_list(rio->fd); UNBLOCKIO(); } #endif /* REFCLOCK */ + /* * On NT a SOCKET is an unsigned int so we cannot possibly keep it in * an array. So we use one of the ISC_LIST functions to hold the * socket value and use that when we want to enumerate it. + * + * This routine is called by the forked intres child process to close + * all open sockets. On Windows there's no need as intres runs in + * the same process as a thread. */ +#ifndef SYS_WINNT void -kill_asyncio(int startfd) +kill_asyncio( + int startfd + ) { - vsock_t *lsock; - vsock_t *next; - BLOCKIO(); - lsock = ISC_LIST_HEAD(fd_list); - while (lsock != NULL) { - /* - * careful here - list is being dismantled while - * we scan it - setting next here insures that - * we are able to correctly scan the list - */ - next = ISC_LIST_NEXT(lsock, link); - /* - * will remove socket from list - */ - close_and_delete_fd_from_list(lsock->fd); - lsock = next; - } + /* + * In the child process we do not maintain activefds and + * maxactivefd. Zeroing maxactivefd disables code which + * maintains it in close_and_delete_fd_from_list(). + */ + maxactivefd = 0; + + while (fd_list != NULL) + close_and_delete_fd_from_list(fd_list->fd); UNBLOCKIO(); } +#endif /* !SYS_WINNT */ + /* * Add and delete functions for the list of open sockets */ static void -add_fd_to_list(SOCKET fd, enum desc_type type) { - vsock_t *lsock = (vsock_t *)emalloc(sizeof(vsock_t)); +add_fd_to_list( + SOCKET fd, + enum desc_type type + ) +{ + vsock_t *lsock = emalloc(sizeof(*lsock)); + lsock->fd = fd; lsock->type = type; - ISC_LIST_APPEND(fd_list, lsock, link); - /* - * I/O Completion Ports don't care about the select and FD_SET - */ -#ifndef HAVE_IO_COMPLETION_PORT - if (fd < 0 || fd >= FD_SETSIZE) { - msyslog(LOG_ERR, "Too many sockets in use, FD_SETSIZE %d exceeded", - FD_SETSIZE); - exit(1); - } - /* - * keep activefds in sync - */ - if (fd > maxactivefd) - maxactivefd = fd; - FD_SET( (u_int)fd, &activefds); -#endif + LINK_SLIST(fd_list, lsock, link); + maintain_activefds(fd, 0); } + static void -close_and_delete_fd_from_list(SOCKET fd) { +close_and_delete_fd_from_list( + SOCKET fd + ) +{ + vsock_t *lsock; - vsock_t *next; - vsock_t *lsock = ISC_LIST_HEAD(fd_list); + UNLINK_EXPR_SLIST(lsock, fd_list, fd == + UNLINK_EXPR_SLIST_CURRENT()->fd, link, vsock_t); - while(lsock != NULL) { - next = ISC_LIST_NEXT(lsock, link); - if(lsock->fd == fd) { - ISC_LIST_DEQUEUE_TYPE(fd_list, lsock, link, vsock_t); + if (NULL == lsock) + return; - switch (lsock->type) { - case FD_TYPE_SOCKET: -#ifdef SYS_WINNT - closesocket(lsock->fd); - break; -#endif - case FD_TYPE_FILE: - (void) close(lsock->fd); - break; - default: - msyslog(LOG_ERR, "internal error - illegal descriptor type %d - EXITING", (int)lsock->type); - exit(1); - } + switch (lsock->type) { - free(lsock); - /* - * I/O Completion Ports don't care about select and fd_set - */ -#ifndef HAVE_IO_COMPLETION_PORT - /* - * remove from activefds - */ - FD_CLR( (u_int) fd, &activefds); - - if (fd == maxactivefd) { - int i, newmax = 0; - for (i = 0; i < maxactivefd; i++) - if (FD_ISSET(i, &activefds)) - newmax = i; - maxactivefd = newmax; - } -#endif - break; - } - lsock = next; + case FD_TYPE_SOCKET: + closesocket(lsock->fd); + break; + + case FD_TYPE_FILE: + closeserial(lsock->fd); + break; + + default: + msyslog(LOG_ERR, + "internal error - illegal descriptor type %d - EXITING", + (int)lsock->type); + exit(1); } + + free(lsock); + /* + * remove from activefds + */ + maintain_activefds(fd, 1); } + static void -add_addr_to_list(struct sockaddr_storage *addr, struct interface *interface){ +add_addr_to_list( + sockaddr_u * addr, + endpt * ep + ) +{ + remaddr_t *laddr; + #ifdef DEBUG if (find_addr_in_list(addr) == NULL) { #endif /* not there yet - add to list */ - remaddr_t *laddr = (remaddr_t *)emalloc(sizeof(remaddr_t)); - memcpy(&laddr->addr, addr, sizeof(struct sockaddr_storage)); - laddr->interface = interface; - - ISC_LIST_APPEND(remoteaddr_list, laddr, link); - + laddr = emalloc(sizeof(*laddr)); + laddr->addr = *addr; + laddr->ep = ep; + + LINK_SLIST(remoteaddr_list, laddr, link); + DPRINTF(4, ("Added addr %s to list of addresses\n", stoa(addr))); #ifdef DEBUG - } else { + } else DPRINTF(4, ("WARNING: Attempt to add duplicate addr %s to address list\n", stoa(addr))); - } #endif } + static void -delete_addr_from_list(struct sockaddr_storage *addr) { - - remaddr_t *next; - remaddr_t *laddr = ISC_LIST_HEAD(remoteaddr_list); - - while(laddr != NULL) { - next = ISC_LIST_NEXT(laddr, link); - if(SOCKCMP(&laddr->addr, addr)) { - ISC_LIST_DEQUEUE_TYPE(remoteaddr_list, laddr, link, remaddr_t); - DPRINTF(4, ("Deleted addr %s from list of addresses\n", - stoa(addr))); - free(laddr); - break; - } - laddr = next; +delete_addr_from_list( + sockaddr_u *addr + ) +{ + remaddr_t *unlinked; + + UNLINK_EXPR_SLIST(unlinked, remoteaddr_list, SOCK_EQ(addr, + &(UNLINK_EXPR_SLIST_CURRENT()->addr)), link, remaddr_t); + + if (unlinked != NULL) { + DPRINTF(4, ("Deleted addr %s from list of addresses\n", + stoa(addr))); + free(unlinked); } } + static void -delete_interface_from_list(struct interface *iface) { - remaddr_t *next; - remaddr_t *laddr = ISC_LIST_HEAD(remoteaddr_list); - - while(laddr != NULL) { - next = ISC_LIST_NEXT(laddr, link); - if (laddr->interface == iface) { - ISC_LIST_DEQUEUE_TYPE(remoteaddr_list, laddr, link, remaddr_t); - DPRINTF(4, ("Deleted addr %s for interface #%d %s from list of addresses\n", - stoa(&laddr->addr), iface->ifnum, iface->name)); - free(laddr); - } - laddr = next; +delete_interface_from_list( + endpt *iface + ) +{ + remaddr_t *unlinked; + + for (;;) { + UNLINK_EXPR_SLIST(unlinked, remoteaddr_list, iface == + UNLINK_EXPR_SLIST_CURRENT()->ep, link, + remaddr_t); + + if (unlinked == NULL) + break; + DPRINTF(4, ("Deleted addr %s for interface #%d %s from list of addresses\n", + stoa(&unlinked->addr), iface->ifnum, + iface->name)); + free(unlinked); } } + static struct interface * -find_addr_in_list(struct sockaddr_storage *addr) { +find_addr_in_list( + sockaddr_u *addr + ) +{ + remaddr_t *entry; - remaddr_t *next; - remaddr_t *laddr = ISC_LIST_HEAD(remoteaddr_list); DPRINTF(4, ("Searching for addr %s in list of addresses - ", stoa(addr))); - while(laddr != NULL) { - next = ISC_LIST_NEXT(laddr, link); - if(SOCKCMP(&laddr->addr, addr)) { + for (entry = remoteaddr_list; + entry != NULL; + entry = entry->link) + if (SOCK_EQ(&entry->addr, addr)) { DPRINTF(4, ("FOUND\n")); - return laddr->interface; + return entry->ep; } - else - laddr = next; - } + DPRINTF(4, ("NOT FOUND\n")); - return NULL; /* Not found */ + return NULL; } + /* - * Find the given address with the associated flag in the list + * Find the given address with the all given flags set in the list */ -static struct interface * -find_flagged_addr_in_list(struct sockaddr_storage *addr, int flag) { +static endpt * +find_flagged_addr_in_list( + sockaddr_u * addr, + u_int32 flags + ) +{ + remaddr_t *entry; - remaddr_t *next; - remaddr_t *laddr = ISC_LIST_HEAD(remoteaddr_list); - DPRINTF(4, ("Finding addr %s in list of addresses\n", - stoa(addr))); + DPRINTF(4, ("Finding addr %s with flags %d in list: ", + stoa(addr), flags)); - while(laddr != NULL) { - next = ISC_LIST_NEXT(laddr, link); - if(SOCKCMP(&laddr->addr, addr) && (laddr->interface->flags & flag)) { - return laddr->interface; - break; + for (entry = remoteaddr_list; + entry != NULL; + entry = entry->link) + + if (SOCK_EQ(&entry->addr, addr) + && (entry->ep->flags & flags) == flags) { + + DPRINTF(4, ("FOUND\n")); + return entry->ep; } - else - laddr = next; - } - return NULL; /* Not found */ + + DPRINTF(4, ("NOT FOUND\n")); + return NULL; } -#ifdef HAS_ROUTING_SOCKET -#include <net/route.h> -#ifndef UPDATE_GRACE -#define UPDATE_GRACE 2 /* wait UPDATE_GRACE seconds before scanning */ -#endif +const char * +localaddrtoa( + endpt *la + ) +{ + return (NULL == la) + ? "<null>" + : stoa(&la->sin); +} + + +#ifdef HAS_ROUTING_SOCKET +# ifndef UPDATE_GRACE +# define UPDATE_GRACE 2 /* wait UPDATE_GRACE seconds before scanning */ +# endif static void process_routing_msgs(struct asyncio_reader *reader) { char buffer[5120]; - char *p = buffer; + int cnt, msg_type; +#ifdef HAVE_RTNETLINK + struct nlmsghdr *nh; +#else + struct rt_msghdr rtm; + char *p; +#endif - int cnt; - if (disable_dynamic_updates) { /* - * discard ourselves if we are not need any more + * discard ourselves if we are not needed any more * usually happens when running unprivileged */ remove_asyncio_reader(reader); @@ -3933,9 +4629,10 @@ process_routing_msgs(struct asyncio_reader *reader) } cnt = read(reader->fd, buffer, sizeof(buffer)); - + if (cnt < 0) { - msyslog(LOG_ERR, "i/o error on routing socket %m - disabling"); + msyslog(LOG_ERR, + "i/o error on routing socket %m - disabling"); remove_asyncio_reader(reader); delete_asyncio_reader(reader); return; @@ -3944,19 +4641,28 @@ process_routing_msgs(struct asyncio_reader *reader) /* * process routing message */ - while ((p + sizeof(struct rt_msghdr)) <= (buffer + cnt)) - { - struct rt_msghdr *rtm; - - rtm = (struct rt_msghdr *)p; - if (rtm->rtm_version != RTM_VERSION) { - msyslog(LOG_ERR, "version mismatch on routing socket %m - disabling"); +#ifdef HAVE_RTNETLINK + for (nh = (struct nlmsghdr *)buffer; + NLMSG_OK(nh, cnt); + nh = NLMSG_NEXT(nh, cnt)) { + msg_type = nh->nlmsg_type; +#else + for (p = buffer; + (p + sizeof(struct rt_msghdr)) <= (buffer + cnt); + p += rtm.rtm_msglen) { + memcpy(&rtm, p, sizeof(rtm)); + if (rtm.rtm_version != RTM_VERSION) { + msyslog(LOG_ERR, + "version mismatch (got %d - expected %d) on routing socket - disabling", + rtm.rtm_version, RTM_VERSION); + remove_asyncio_reader(reader); delete_asyncio_reader(reader); return; } - - switch (rtm->rtm_type) { + msg_type = rtm.rtm_type; +#endif + switch (msg_type) { #ifdef RTM_NEWADDR case RTM_NEWADDR: #endif @@ -3984,20 +4690,40 @@ process_routing_msgs(struct asyncio_reader *reader) #ifdef RTM_IFANNOUNCE case RTM_IFANNOUNCE: #endif +#ifdef RTM_NEWLINK + case RTM_NEWLINK: +#endif +#ifdef RTM_DELLINK + case RTM_DELLINK: +#endif +#ifdef RTM_NEWROUTE + case RTM_NEWROUTE: +#endif +#ifdef RTM_DELROUTE + case RTM_DELROUTE: +#endif /* - * we are keen on new and deleted addresses and if an interface goes up and down or routing changes + * we are keen on new and deleted addresses and + * if an interface goes up and down or routing + * changes */ - DPRINTF(3, ("routing message op = %d: scheduling interface update\n", rtm->rtm_type)); + DPRINTF(3, ("routing message op = %d: scheduling interface update\n", + msg_type)); timer_interfacetimeout(current_time + UPDATE_GRACE); break; +#ifdef HAVE_RTNETLINK + case NLMSG_DONE: + /* end of multipart message */ + return; +#endif default: /* * the rest doesn't bother us. */ - DPRINTF(4, ("routing message op = %d: ignored\n", rtm->rtm_type)); + DPRINTF(4, ("routing message op = %d: ignored\n", + msg_type)); break; } - p += rtm->rtm_msglen; } } @@ -4008,29 +4734,52 @@ static void init_async_notifications() { struct asyncio_reader *reader; +#ifdef HAVE_RTNETLINK + int fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE); + struct sockaddr_nl sa; +#else int fd = socket(PF_ROUTE, SOCK_RAW, 0); - - if (fd >= 0) { - fd = move_fd(fd); - init_nonblocking_io(fd); +#endif + if (fd < 0) { + msyslog(LOG_ERR, + "unable to open routing socket (%m) - using polled interface update"); + return; + } + + fd = move_fd(fd); +#ifdef HAVE_RTNETLINK + ZERO(sa); + sa.nl_family = PF_NETLINK; + sa.nl_groups = RTMGRP_LINK | RTMGRP_IPV4_IFADDR + | RTMGRP_IPV6_IFADDR | RTMGRP_IPV4_ROUTE + | RTMGRP_IPV4_MROUTE | RTMGRP_IPV6_ROUTE + | RTMGRP_IPV6_MROUTE; + if (bind(fd, (struct sockaddr *)&sa, sizeof(sa)) < 0) { + msyslog(LOG_ERR, + "bind failed on routing socket (%m) - using polled interface update"); + return; + } +#endif + make_socket_nonblocking(fd); #if defined(HAVE_SIGNALED_IO) - init_socket_sig(fd); + init_socket_sig(fd); #endif /* HAVE_SIGNALED_IO */ - - reader = new_asyncio_reader(); - - reader->fd = fd; - reader->receiver = process_routing_msgs; - - add_asyncio_reader(reader, FD_TYPE_SOCKET); - msyslog(LOG_INFO, "Listening on routing socket on fd #%d for interface updates", fd); - } else { - msyslog(LOG_ERR, "unable to open routing socket (%m) - using polled interface update"); - } + + reader = new_asyncio_reader(); + + reader->fd = fd; + reader->receiver = process_routing_msgs; + + add_asyncio_reader(reader, FD_TYPE_SOCKET); + msyslog(LOG_INFO, + "Listening on routing socket on fd #%d for interface updates", + fd); } #else +/* HAS_ROUTING_SOCKET not defined */ static void -init_async_notifications() +init_async_notifications(void) { } #endif + |