--- embedaddon/dnsmasq/src/dnsmasq.c 2016/11/02 09:57:01 1.1.1.3 +++ embedaddon/dnsmasq/src/dnsmasq.c 2021/03/17 00:56:46 1.1.1.4 @@ -1,4 +1,4 @@ -/* dnsmasq is Copyright (c) 2000-2016 Simon Kelley +/* dnsmasq is Copyright (c) 2000-2021 Simon Kelley This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by @@ -48,11 +48,17 @@ int main (int argc, char **argv) long i, max_fd = sysconf(_SC_OPEN_MAX); char *baduser = NULL; int log_err; + int chown_warn = 0; #if defined(HAVE_LINUX_NETWORK) cap_user_header_t hdr = NULL; cap_user_data_t data = NULL; + int need_cap_net_admin = 0; + int need_cap_net_raw = 0; + int need_cap_net_bind_service = 0; char *bound_device = NULL; int did_bind = 0; + struct server *serv; + char *netlink_warn; #endif #if defined(HAVE_DHCP) || defined(HAVE_DHCP6) struct dhcp_context *context; @@ -77,17 +83,22 @@ int main (int argc, char **argv) sigaction(SIGTERM, &sigact, NULL); sigaction(SIGALRM, &sigact, NULL); sigaction(SIGCHLD, &sigact, NULL); - + sigaction(SIGINT, &sigact, NULL); + /* ignore SIGPIPE */ sigact.sa_handler = SIG_IGN; sigaction(SIGPIPE, &sigact, NULL); umask(022); /* known umask, create leases and pid files as 0644 */ - + rand_init(); /* Must precede read_opts() */ read_opts(argc, argv, compile_opts); +#ifdef HAVE_LINUX_NETWORK + daemon->kernel_version = kernel_version(); +#endif + if (daemon->edns_pktsz < PACKETSZ) daemon->edns_pktsz = PACKETSZ; @@ -118,6 +129,9 @@ int main (int argc, char **argv) daemon->namebuff = safe_malloc(MAXDNAME * 2); daemon->keyname = safe_malloc(MAXDNAME * 2); daemon->workspacename = safe_malloc(MAXDNAME * 2); + /* one char flag per possible RR in answer section (may get extended). */ + daemon->rr_status_sz = 64; + daemon->rr_status = safe_malloc(sizeof(*daemon->rr_status) * daemon->rr_status_sz); } #endif @@ -129,20 +143,18 @@ int main (int argc, char **argv) } #endif - /* Close any file descriptors we inherited apart from std{in|out|err} - - Ensure that at least stdin, stdout and stderr (fd 0, 1, 2) exist, + /* Ensure that at least stdin, stdout and stderr (fd 0, 1, 2) exist, otherwise file descriptors we create can end up being 0, 1, or 2 and then get accidentally closed later when we make 0, 1, and 2 open to /dev/null. Normally we'll be started with 0, 1 and 2 open, but it's not guaranteed. By opening /dev/null three times, we ensure that we're not using those fds for real stuff. */ - for (i = 0; i < max_fd; i++) - if (i != STDOUT_FILENO && i != STDERR_FILENO && i != STDIN_FILENO) - close(i); - else - open("/dev/null", O_RDWR); - + for (i = 0; i < 3; i++) + open("/dev/null", O_RDWR); + + /* Close any file descriptors we inherited apart from std{in|out|err} */ + close_fds(max_fd, -1, -1, -1); + #ifndef HAVE_LINUX_NETWORK # if !(defined(IP_RECVDSTADDR) && defined(IP_RECVIF) && defined(IP_SENDSRCADDR)) if (!option_bool(OPT_NOWILD)) @@ -202,16 +214,16 @@ int main (int argc, char **argv) #ifdef HAVE_SOLARIS_NETWORK if (daemon->max_logs != 0) - die(_("asychronous logging is not available under Solaris"), NULL, EC_BADCONF); + die(_("asynchronous logging is not available under Solaris"), NULL, EC_BADCONF); #endif #ifdef __ANDROID__ if (daemon->max_logs != 0) - die(_("asychronous logging is not available under Android"), NULL, EC_BADCONF); + die(_("asynchronous logging is not available under Android"), NULL, EC_BADCONF); #endif #ifndef HAVE_AUTH - if (daemon->authserver) + if (daemon->auth_zones) die(_("authoritative DNS not available: set HAVE_AUTH in src/config.h"), NULL, EC_BADCONF); #endif @@ -220,21 +232,30 @@ int main (int argc, char **argv) die(_("loop detection not available: set HAVE_LOOP in src/config.h"), NULL, EC_BADCONF); #endif - if (daemon->max_port != MAX_PORT && daemon->min_port == 0) - daemon->min_port = 1024u; - +#ifndef HAVE_UBUS + if (option_bool(OPT_UBUS)) + die(_("Ubus not available: set HAVE_UBUS in src/config.h"), NULL, EC_BADCONF); +#endif + if (daemon->max_port < daemon->min_port) die(_("max_port cannot be smaller than min_port"), NULL, EC_BADCONF); now = dnsmasq_time(); - /* Create a serial at startup if not configured. */ - if (daemon->authinterface && daemon->soa_sn == 0) + if (daemon->auth_zones) + { + if (!daemon->authserver) + die(_("--auth-server required when an auth zone is defined."), NULL, EC_BADCONF); + + /* Create a serial at startup if not configured. */ #ifdef HAVE_BROKEN_RTC - die(_("zone serial must be configured in --auth-soa"), NULL, EC_BADCONF); + if (daemon->soa_sn == 0) + die(_("zone serial must be configured in --auth-soa"), NULL, EC_BADCONF); #else - daemon->soa_sn = now; + if (daemon->soa_sn == 0) + daemon->soa_sn = now; #endif + } #ifdef HAVE_DHCP6 if (daemon->dhcp6) @@ -271,11 +292,24 @@ int main (int argc, char **argv) } if (daemon->dhcp || daemon->relay4) - dhcp_init(); + { + dhcp_init(); +# ifdef HAVE_LINUX_NETWORK + if (!option_bool(OPT_NO_PING)) + need_cap_net_raw = 1; + need_cap_net_admin = 1; +# endif + } # ifdef HAVE_DHCP6 if (daemon->doing_ra || daemon->doing_dhcp6 || daemon->relay6) - ra_init(now); + { + ra_init(now); +# ifdef HAVE_LINUX_NETWORK + need_cap_net_raw = 1; + need_cap_net_admin = 1; +# endif + } if (daemon->doing_dhcp6 || daemon->relay6) dhcp6_init(); @@ -285,11 +319,16 @@ int main (int argc, char **argv) #ifdef HAVE_IPSET if (daemon->ipsets) - ipset_init(); + { + ipset_init(); +# ifdef HAVE_LINUX_NETWORK + need_cap_net_admin = 1; +# endif + } #endif #if defined(HAVE_LINUX_NETWORK) - netlink_init(); + netlink_warn = netlink_init(); #elif defined(HAVE_BSD_NETWORK) route_init(); #endif @@ -351,19 +390,27 @@ int main (int argc, char **argv) if (daemon->port != 0) { cache_init(); - -#ifdef HAVE_DNSSEC blockdata_init(); -#endif + hash_questions_init(); } #ifdef HAVE_INOTIFY - if (daemon->port != 0 || daemon->dhcp || daemon->doing_dhcp6) + if ((daemon->port != 0 || daemon->dhcp || daemon->doing_dhcp6) + && (!option_bool(OPT_NO_RESOLV) || daemon->dynamic_dirs)) inotify_dnsmasq_init(); else daemon->inotifyfd = -1; #endif - + + if (daemon->dump_file) +#ifdef HAVE_DUMPFILE + dump_init(); + else + daemon->dumpfd = -1; +#else + die(_("Packet dumps not available: set HAVE_DUMP in src/config.h"), NULL, EC_BADCONF); +#endif + if (option_bool(OPT_DBUS)) #ifdef HAVE_DBUS { @@ -377,6 +424,16 @@ int main (int argc, char **argv) die(_("DBus not available: set HAVE_DBUS in src/config.h"), NULL, EC_BADCONF); #endif + if (option_bool(OPT_UBUS)) +#ifdef HAVE_UBUS + { + daemon->ubus = NULL; + ubus_init(); + } +#else + die(_("UBus not available: set HAVE_UBUS in src/config.h"), NULL, EC_BADCONF); +#endif + if (daemon->port != 0) pre_allocate_sfds(); @@ -386,10 +443,12 @@ int main (int argc, char **argv) daemon->scriptuser && (daemon->lease_change_command || daemon->luascript)) { - if ((ent_pw = getpwnam(daemon->scriptuser))) + struct passwd *scr_pw; + + if ((scr_pw = getpwnam(daemon->scriptuser))) { - script_uid = ent_pw->pw_uid; - script_gid = ent_pw->pw_gid; + script_uid = scr_pw->pw_uid; + script_gid = scr_pw->pw_gid; } else baduser = daemon->scriptuser; @@ -416,28 +475,81 @@ int main (int argc, char **argv) } #if defined(HAVE_LINUX_NETWORK) + /* We keep CAP_NETADMIN (for ARP-injection) and + CAP_NET_RAW (for icmp) if we're doing dhcp, + if we have yet to bind ports because of DAD, + or we're doing it dynamically, we need CAP_NET_BIND_SERVICE. */ + if ((is_dad_listeners() || option_bool(OPT_CLEVERBIND)) && + (option_bool(OPT_TFTP) || (daemon->port != 0 && daemon->port <= 1024))) + need_cap_net_bind_service = 1; + + /* usptream servers which bind to an interface call SO_BINDTODEVICE + for each TCP connection, so need CAP_NET_RAW */ + for (serv = daemon->servers; serv; serv = serv->next) + if (serv->interface[0] != 0) + need_cap_net_raw = 1; + + /* If we're doing Dbus or UBus, the above can be set dynamically, + (as can ports) so always (potentially) needed. */ +#ifdef HAVE_DBUS + if (option_bool(OPT_DBUS)) + { + need_cap_net_bind_service = 1; + need_cap_net_raw = 1; + } +#endif + +#ifdef HAVE_UBUS + if (option_bool(OPT_UBUS)) + { + need_cap_net_bind_service = 1; + need_cap_net_raw = 1; + } +#endif + /* determine capability API version here, while we can still call safe_malloc */ - if (ent_pw && ent_pw->pw_uid != 0) + int capsize = 1; /* for header version 1 */ + char *fail = NULL; + + hdr = safe_malloc(sizeof(*hdr)); + + /* find version supported by kernel */ + memset(hdr, 0, sizeof(*hdr)); + capget(hdr, NULL); + + if (hdr->version != LINUX_CAPABILITY_VERSION_1) { - int capsize = 1; /* for header version 1 */ - hdr = safe_malloc(sizeof(*hdr)); - - /* find version supported by kernel */ - memset(hdr, 0, sizeof(*hdr)); - capget(hdr, NULL); - - if (hdr->version != LINUX_CAPABILITY_VERSION_1) - { - /* if unknown version, use largest supported version (3) */ - if (hdr->version != LINUX_CAPABILITY_VERSION_2) - hdr->version = LINUX_CAPABILITY_VERSION_3; - capsize = 2; - } - - data = safe_malloc(sizeof(*data) * capsize); - memset(data, 0, sizeof(*data) * capsize); + /* if unknown version, use largest supported version (3) */ + if (hdr->version != LINUX_CAPABILITY_VERSION_2) + hdr->version = LINUX_CAPABILITY_VERSION_3; + capsize = 2; } + + data = safe_malloc(sizeof(*data) * capsize); + capget(hdr, data); /* Get current values, for verification */ + + if (need_cap_net_admin && !(data->permitted & (1 << CAP_NET_ADMIN))) + fail = "NET_ADMIN"; + else if (need_cap_net_raw && !(data->permitted & (1 << CAP_NET_RAW))) + fail = "NET_RAW"; + else if (need_cap_net_bind_service && !(data->permitted & (1 << CAP_NET_BIND_SERVICE))) + fail = "NET_BIND_SERVICE"; + + if (fail) + die(_("process is missing required capability %s"), fail, EC_MISC); + + /* Now set bitmaps to set caps after daemonising */ + memset(data, 0, sizeof(*data) * capsize); + + if (need_cap_net_admin) + data->effective |= (1 << CAP_NET_ADMIN); + if (need_cap_net_raw) + data->effective |= (1 << CAP_NET_RAW); + if (need_cap_net_bind_service) + data->effective |= (1 << CAP_NET_BIND_SERVICE); + + data->permitted = data->effective; #endif /* Use a pipe to carry signals and other events back to the event loop @@ -459,7 +571,6 @@ int main (int argc, char **argv) if (chdir("/") != 0) die(_("cannot chdir to filesystem root: %s"), NULL, EC_MISC); -#ifndef NO_FORK if (!option_bool(OPT_NO_FORK)) { pid_t pid; @@ -478,7 +589,7 @@ int main (int argc, char **argv) char *msg; /* close our copy of write-end */ - while (retry_send(close(err_pipe[1]))); + close(err_pipe[1]); /* check for errors after the fork */ if (read_event(err_pipe[0], &ev, &msg)) @@ -487,7 +598,7 @@ int main (int argc, char **argv) _exit(EC_GOOD); } - while (retry_send(close(err_pipe[0]))); + close(err_pipe[0]); /* NO calls to die() from here on. */ @@ -499,7 +610,6 @@ int main (int argc, char **argv) if (pid != 0) _exit(0); } -#endif /* write pidfile _after_ forking ! */ if (daemon->runfile) @@ -514,7 +624,7 @@ int main (int argc, char **argv) extent that an attacker running as the unprivileged user could replace the pidfile with a symlink, and have the target of that symlink overwritten as root next time dnsmasq starts. - The folowing code first deletes any existing file, and then opens it with the O_EXCL flag, + The following code first deletes any existing file, and then opens it with the O_EXCL flag, ensuring that the open() fails should there be any existing file (because the unlink() failed, or an attacker exploited the race between unlink() and open()). This ensures that no symlink attack can succeed. @@ -537,12 +647,20 @@ int main (int argc, char **argv) } else { + /* We're still running as root here. Change the ownership of the PID file + to the user we will be running as. Note that this is not to allow + us to delete the file, since that depends on the permissions + of the directory containing the file. That directory will + need to by owned by the dnsmasq user, and the ownership of the + file has to match, to keep systemd >273 happy. */ + if (getuid() == 0 && ent_pw && ent_pw->pw_uid != 0 && fchown(fd, ent_pw->pw_uid, ent_pw->pw_gid) == -1) + chown_warn = errno; + if (!read_write(fd, (unsigned char *)daemon->namebuff, strlen(daemon->namebuff), 0)) err = 1; - else + else { - while (retry_send(close(fd))); - if (errno != 0) + if (close(fd) == -1) err = 1; } } @@ -583,7 +701,7 @@ int main (int argc, char **argv) int bad_capabilities = 0; gid_t dummy; - /* remove all supplimentary groups */ + /* remove all supplementary groups */ if (gp && (setgroups(0, &dummy) == -1 || setgid(gp->gr_gid) == -1)) @@ -595,18 +713,9 @@ int main (int argc, char **argv) if (ent_pw && ent_pw->pw_uid != 0) { #if defined(HAVE_LINUX_NETWORK) - /* On linux, we keep CAP_NETADMIN (for ARP-injection) and - CAP_NET_RAW (for icmp) if we're doing dhcp. If we have yet to bind - ports because of DAD, or we're doing it dynamically, - we need CAP_NET_BIND_SERVICE too. */ - if (is_dad_listeners() || option_bool(OPT_CLEVERBIND)) - data->effective = data->permitted = data->inheritable = - (1 << CAP_NET_ADMIN) | (1 << CAP_NET_RAW) | - (1 << CAP_SETUID) | (1 << CAP_NET_BIND_SERVICE); - else - data->effective = data->permitted = data->inheritable = - (1 << CAP_NET_ADMIN) | (1 << CAP_NET_RAW) | (1 << CAP_SETUID); - + /* Need to be able to drop root. */ + data->effective |= (1 << CAP_SETUID); + data->permitted |= (1 << CAP_SETUID); /* Tell kernel to not clear capabilities when dropping root */ if (capset(hdr, data) == -1 || prctl(PR_SET_KEEPCAPS, 1, 0, 0, 0) == -1) bad_capabilities = errno; @@ -647,15 +756,10 @@ int main (int argc, char **argv) } #ifdef HAVE_LINUX_NETWORK - if (is_dad_listeners() || option_bool(OPT_CLEVERBIND)) - data->effective = data->permitted = - (1 << CAP_NET_ADMIN) | (1 << CAP_NET_RAW) | (1 << CAP_NET_BIND_SERVICE); - else - data->effective = data->permitted = - (1 << CAP_NET_ADMIN) | (1 << CAP_NET_RAW); - data->inheritable = 0; + data->effective &= ~(1 << CAP_SETUID); + data->permitted &= ~(1 << CAP_SETUID); - /* lose the setuid and setgid capbilities */ + /* lose the setuid capability */ if (capset(hdr, data) == -1) { send_event(err_pipe[1], EVENT_CAP_ERR, errno, NULL); @@ -714,12 +818,25 @@ int main (int argc, char **argv) if (daemon->port == 0) my_syslog(LOG_INFO, _("started, version %s DNS disabled"), VERSION); - else if (daemon->cachesize != 0) - my_syslog(LOG_INFO, _("started, version %s cachesize %d"), VERSION, daemon->cachesize); - else - my_syslog(LOG_INFO, _("started, version %s cache disabled"), VERSION); + else + { + if (daemon->cachesize != 0) + { + my_syslog(LOG_INFO, _("started, version %s cachesize %d"), VERSION, daemon->cachesize); + if (daemon->cachesize > 10000) + my_syslog(LOG_WARNING, _("cache size greater than 10000 may cause performance issues, and is unlikely to be useful.")); + } + else + my_syslog(LOG_INFO, _("started, version %s cache disabled"), VERSION); + + if (option_bool(OPT_LOCAL_SERVICE)) + my_syslog(LOG_INFO, _("DNS service limited to local subnets")); + } my_syslog(LOG_INFO, _("compile time options: %s"), compile_opts); + + if (chown_warn != 0) + my_syslog(LOG_WARNING, "chown of PID file %s failed: %s", daemon->runfile, strerror(chown_warn)); #ifdef HAVE_DBUS if (option_bool(OPT_DBUS)) @@ -731,14 +848,22 @@ int main (int argc, char **argv) } #endif - if (option_bool(OPT_LOCAL_SERVICE)) - my_syslog(LOG_INFO, _("DNS service limited to local subnets")); - +#ifdef HAVE_UBUS + if (option_bool(OPT_UBUS)) + { + if (daemon->ubus) + my_syslog(LOG_INFO, _("UBus support enabled: connected to system bus")); + else + my_syslog(LOG_INFO, _("UBus support enabled: bus connection pending")); + } +#endif + #ifdef HAVE_DNSSEC if (option_bool(OPT_DNSSEC_VALID)) { int rc; - + struct ds_config *ds; + /* Delay creating the timestamp file until here, after we've changed user, so that it has the correct owner to allow updating the mtime later. This means we have to report fatal errors via the pipe. */ @@ -748,13 +873,21 @@ int main (int argc, char **argv) _exit(0); } - my_syslog(LOG_INFO, _("DNSSEC validation enabled")); + if (option_bool(OPT_DNSSEC_IGN_NS)) + my_syslog(LOG_INFO, _("DNSSEC validation enabled but all unsigned answers are trusted")); + else + my_syslog(LOG_INFO, _("DNSSEC validation enabled")); - if (option_bool(OPT_DNSSEC_TIME)) - my_syslog(LOG_INFO, _("DNSSEC signature timestamps not checked until first cache reload")); + daemon->dnssec_no_time_check = option_bool(OPT_DNSSEC_TIME); + if (option_bool(OPT_DNSSEC_TIME) && !daemon->back_to_the_future) + my_syslog(LOG_INFO, _("DNSSEC signature timestamps not checked until receipt of SIGINT")); if (rc == 1) my_syslog(LOG_INFO, _("DNSSEC signature timestamps not checked until system time valid")); + + for (ds = daemon->ds; ds; ds = ds->next) + my_syslog(LOG_INFO, _("configured with trust anchor for %s keytag %u"), + ds->name[0] == 0 ? "" : ds->name, ds->keytag); } #endif @@ -767,6 +900,8 @@ int main (int argc, char **argv) if (option_bool(OPT_NOWILD)) warn_bound_listeners(); + else if (!option_bool(OPT_CLEVERBIND)) + warn_wild_labels(); warn_int_names(); @@ -812,9 +947,12 @@ int main (int argc, char **argv) # ifdef HAVE_LINUX_NETWORK if (did_bind) my_syslog(MS_DHCP | LOG_INFO, _("DHCP, sockets bound exclusively to interface %s"), bound_device); + + if (netlink_warn) + my_syslog(LOG_WARNING, netlink_warn); # endif - /* after dhcp_contruct_contexts */ + /* after dhcp_construct_contexts */ if (daemon->dhcp || daemon->doing_dhcp6) lease_find_interfaces(now); #endif @@ -824,10 +962,11 @@ int main (int argc, char **argv) { struct tftp_prefix *p; - my_syslog(MS_TFTP | LOG_INFO, "TFTP %s%s %s", + my_syslog(MS_TFTP | LOG_INFO, "TFTP %s%s %s %s", daemon->tftp_prefix ? _("root is ") : _("enabled"), - daemon->tftp_prefix ? daemon->tftp_prefix: "", - option_bool(OPT_TFTP_SECURE) ? _("secure mode") : ""); + daemon->tftp_prefix ? daemon->tftp_prefix : "", + option_bool(OPT_TFTP_SECURE) ? _("secure mode") : "", + option_bool(OPT_SINGLE_PORT) ? _("single port mode") : ""); if (tftp_prefix_missing) my_syslog(MS_TFTP | LOG_WARNING, _("warning: %s inaccessible"), daemon->tftp_prefix); @@ -845,7 +984,7 @@ int main (int argc, char **argv) if (max_fd < 0) max_fd = 5; - else if (max_fd < 100) + else if (max_fd < 100 && !option_bool(OPT_SINGLE_PORT)) max_fd = max_fd/2; else max_fd = max_fd - 20; @@ -868,12 +1007,16 @@ int main (int argc, char **argv) /* finished start-up - release original process */ if (err_pipe[1] != -1) - while (retry_send(close(err_pipe[1]))); + close(err_pipe[1]); if (daemon->port != 0) check_servers(); pid = getpid(); + + daemon->pipe_to_parent = -1; + for (i = 0; i < MAX_PROCS; i++) + daemon->tcp_pipes[i] = -1; #ifdef HAVE_INOTIFY /* Using inotify, have to select a resolv file at startup */ @@ -903,8 +1046,13 @@ int main (int argc, char **argv) #ifdef HAVE_DBUS set_dbus_listeners(); -#endif - +#endif + +#ifdef HAVE_UBUS + if (option_bool(OPT_UBUS)) + set_ubus_listeners(); +#endif + #ifdef HAVE_DHCP if (daemon->dhcp || daemon->relay4) { @@ -966,7 +1114,7 @@ int main (int argc, char **argv) #endif - /* must do this just before select(), when we know no + /* must do this just before do_poll(), when we know no more calls to my_syslog() can occur */ set_log_writer(); @@ -1034,7 +1182,20 @@ int main (int argc, char **argv) } check_dbus_listeners(); #endif - + +#ifdef HAVE_UBUS + if (option_bool(OPT_UBUS)) + { + /* if we didn't create a UBus connection, retry now. */ + if (!daemon->ubus) + { + ubus_init(); + } + + check_ubus_listeners(); + } +#endif + check_dns_listeners(now); #ifdef HAVE_TFTP @@ -1073,7 +1234,7 @@ static void sig_handler(int sig) { /* ignore anything other than TERM during startup and in helper proc. (helper ignore TERM too) */ - if (sig == SIGTERM) + if (sig == SIGTERM || sig == SIGINT) exit(EC_MISC); } else if (pid != getpid()) @@ -1099,6 +1260,15 @@ static void sig_handler(int sig) event = EVENT_DUMP; else if (sig == SIGUSR2) event = EVENT_REOPEN; + else if (sig == SIGINT) + { + /* Handle SIGINT normally in debug mode, so + ctrl-c continues to operate. */ + if (option_bool(OPT_DEBUG)) + exit(EC_MISC); + else + event = EVENT_TIME; + } else return; @@ -1181,31 +1351,40 @@ static void fatal_event(struct event_desc *ev, char *m case EVENT_FORK_ERR: die(_("cannot fork into background: %s"), NULL, EC_MISC); - + + /* fall through */ case EVENT_PIPE_ERR: die(_("failed to create helper: %s"), NULL, EC_MISC); - + + /* fall through */ case EVENT_CAP_ERR: die(_("setting capabilities failed: %s"), NULL, EC_MISC); + /* fall through */ case EVENT_USER_ERR: die(_("failed to change user-id to %s: %s"), msg, EC_MISC); + /* fall through */ case EVENT_GROUP_ERR: die(_("failed to change group-id to %s: %s"), msg, EC_MISC); - + + /* fall through */ case EVENT_PIDFILE: die(_("failed to open pidfile %s: %s"), msg, EC_FILE); + /* fall through */ case EVENT_LOG_ERR: die(_("cannot open log %s: %s"), msg, EC_FILE); - + + /* fall through */ case EVENT_LUA_ERR: die(_("failed to load Lua script: %s"), msg, EC_MISC); + /* fall through */ case EVENT_TFTP_ERR: die(_("TFTP directory %s inaccessible: %s"), msg, EC_FILE); - + + /* fall through */ case EVENT_TIME_ERR: die(_("cannot create timestamp file %s: %s" ), msg, EC_BADCONF); } @@ -1225,13 +1404,8 @@ static void async_event(int pipe, time_t now) switch (ev.event) { case EVENT_RELOAD: -#ifdef HAVE_DNSSEC - if (option_bool(OPT_DNSSEC_VALID) && option_bool(OPT_DNSSEC_TIME)) - { - my_syslog(LOG_INFO, _("now checking DNSSEC signature timestamps")); - reset_option_bool(OPT_DNSSEC_TIME); - } -#endif + daemon->soa_sn++; /* Bump zone serial, as it may have changed. */ + /* fall through */ case EVENT_INIT: @@ -1294,6 +1468,7 @@ static void async_event(int pipe, time_t now) daemon->tcp_pids[i] = 0; break; +#if defined(HAVE_SCRIPT) case EVENT_KILLED: my_syslog(LOG_WARNING, _("script process killed by signal %d"), ev.data); break; @@ -1307,12 +1482,19 @@ static void async_event(int pipe, time_t now) daemon->lease_change_command, strerror(ev.data)); break; + case EVENT_SCRIPT_LOG: + my_syslog(MS_SCRIPT | LOG_DEBUG, "%s", msg ? msg : ""); + free(msg); + msg = NULL; + break; + /* necessary for fatal errors in helper */ case EVENT_USER_ERR: case EVENT_DIE: case EVENT_LUA_ERR: fatal_event(&ev, msg); break; +#endif case EVENT_REOPEN: /* Note: this may leave TCP-handling processes with the old file still open. @@ -1332,6 +1514,17 @@ static void async_event(int pipe, time_t now) poll_resolv(0, 1, now); break; + case EVENT_TIME: +#ifdef HAVE_DNSSEC + if (daemon->dnssec_no_time_check && option_bool(OPT_DNSSEC_VALID) && option_bool(OPT_DNSSEC_TIME)) + { + my_syslog(LOG_INFO, _("now checking DNSSEC signature timestamps")); + daemon->dnssec_no_time_check = 0; + clear_cache_and_reload(now); + } +#endif + break; + case EVENT_TERM: /* Knock all our children on the head. */ for (i = 0; i < MAX_PROCS; i++) @@ -1348,7 +1541,7 @@ static void async_event(int pipe, time_t now) do { helper_write(); } while (!helper_buf_empty() || do_script_run(now)); - while (retry_send(close(daemon->helperfd))); + close(daemon->helperfd); } #endif @@ -1359,13 +1552,18 @@ static void async_event(int pipe, time_t now) /* update timestamp file on TERM if time is considered valid */ if (daemon->back_to_the_future) { - if (utime(daemon->timestamp_file, NULL) == -1) + if (utimes(daemon->timestamp_file, NULL) == -1) my_syslog(LOG_ERR, _("failed to update mtime on %s: %s"), daemon->timestamp_file, strerror(errno)); } #endif if (daemon->runfile) unlink(daemon->runfile); + +#ifdef HAVE_DUMPFILE + if (daemon->dumpfd != -1) + close(daemon->dumpfd); +#endif my_syslog(LOG_INFO, _("exiting on receipt of SIGTERM")); flush_log(); @@ -1456,9 +1654,6 @@ void clear_cache_and_reload(time_t now) if (option_bool(OPT_ETHERS)) dhcp_read_ethers(); reread_dhcp(); -#ifdef HAVE_INOTIFY - set_dynamic_inotify(AH_DHCP_HST | AH_DHCP_OPT, 0, NULL, 0); -#endif dhcp_update_configs(daemon->dhcp_conf); lease_update_from_configs(); lease_update_file(now); @@ -1482,16 +1677,17 @@ static int set_dns_listeners(time_t now) #ifdef HAVE_TFTP int tftp = 0; struct tftp_transfer *transfer; - for (transfer = daemon->tftp_trans; transfer; transfer = transfer->next) - { - tftp++; - poll_listen(transfer->sockfd, POLLIN); - } + if (!option_bool(OPT_SINGLE_PORT)) + for (transfer = daemon->tftp_trans; transfer; transfer = transfer->next) + { + tftp++; + poll_listen(transfer->sockfd, POLLIN); + } #endif /* will we be able to get memory? */ if (daemon->port != 0) - get_new_frec(now, &wait, 0); + get_new_frec(now, &wait, NULL); for (serverfdp = daemon->sfds; serverfdp; serverfdp = serverfdp->next) poll_listen(serverfdp->fd, POLLIN); @@ -1511,19 +1707,25 @@ static int set_dns_listeners(time_t now) we don't need to explicitly arrange to wake up here */ if (listener->tcpfd != -1) for (i = 0; i < MAX_PROCS; i++) - if (daemon->tcp_pids[i] == 0) + if (daemon->tcp_pids[i] == 0 && daemon->tcp_pipes[i] == -1) { poll_listen(listener->tcpfd, POLLIN); break; } #ifdef HAVE_TFTP + /* tftp == 0 in single-port mode. */ if (tftp <= daemon->tftp_max && listener->tftpfd != -1) poll_listen(listener->tftpfd, POLLIN); #endif } + if (!option_bool(OPT_DEBUG)) + for (i = 0; i < MAX_PROCS; i++) + if (daemon->tcp_pipes[i] != -1) + poll_listen(daemon->tcp_pipes[i], POLLIN); + return wait; } @@ -1532,7 +1734,8 @@ static void check_dns_listeners(time_t now) struct serverfd *serverfdp; struct listener *listener; int i; - + int pipefd[2]; + for (serverfdp = daemon->sfds; serverfdp; serverfdp = serverfdp->next) if (poll_check(serverfdp->fd, POLLIN)) reply_query(serverfdp->fd, serverfdp->source_addr.sa.sa_family, now); @@ -1542,7 +1745,24 @@ static void check_dns_listeners(time_t now) if (daemon->randomsocks[i].refcount != 0 && poll_check(daemon->randomsocks[i].fd, POLLIN)) reply_query(daemon->randomsocks[i].fd, daemon->randomsocks[i].family, now); - + + /* Races. The child process can die before we read all of the data from the + pipe, or vice versa. Therefore send tcp_pids to zero when we wait() the + process, and tcp_pipes to -1 and close the FD when we read the last + of the data - indicated by cache_recv_insert returning zero. + The order of these events is indeterminate, and both are needed + to free the process slot. Once the child process has gone, poll() + returns POLLHUP, not POLLIN, so have to check for both here. */ + if (!option_bool(OPT_DEBUG)) + for (i = 0; i < MAX_PROCS; i++) + if (daemon->tcp_pipes[i] != -1 && + poll_check(daemon->tcp_pipes[i], POLLIN | POLLHUP) && + !cache_recv_insert(now, daemon->tcp_pipes[i])) + { + close(daemon->tcp_pipes[i]); + daemon->tcp_pipes[i] = -1; + } + for (listener = daemon->listeners; listener; listener = listener->next) { if (listener->fd != -1 && poll_check(listener->fd, POLLIN)) @@ -1568,7 +1788,7 @@ static void check_dns_listeners(time_t now) if (getsockname(confd, (struct sockaddr *)&tcp_addr, &tcp_len) == -1) { - while (retry_send(close(confd))); + close(confd); continue; } @@ -1596,15 +1816,16 @@ static void check_dns_listeners(time_t now) if ((if_index = tcp_interface(confd, tcp_addr.sa.sa_family)) != 0 && indextoname(listener->tcpfd, if_index, intr_name)) { - struct all_addr addr; - addr.addr.addr4 = tcp_addr.in.sin_addr; -#ifdef HAVE_IPV6 + union all_addr addr; + if (tcp_addr.sa.sa_family == AF_INET6) - addr.addr.addr6 = tcp_addr.in6.sin6_addr; -#endif + addr.addr6 = tcp_addr.in6.sin6_addr; + else + addr.addr4 = tcp_addr.in.sin_addr; for (iface = daemon->interfaces; iface; iface = iface->next) - if (iface->index == if_index) + if (iface->index == if_index && + iface->addr.sa.sa_family == tcp_addr.sa.sa_family) break; if (!iface && !loopback_exception(listener->tcpfd, tcp_addr.sa.sa_family, &addr, intr_name)) @@ -1633,27 +1854,48 @@ static void check_dns_listeners(time_t now) if (!client_ok) { shutdown(confd, SHUT_RDWR); - while (retry_send(close(confd))); + close(confd); } -#ifndef NO_FORK - else if (!option_bool(OPT_DEBUG) && (p = fork()) != 0) + else if (!option_bool(OPT_DEBUG) && pipe(pipefd) == 0 && (p = fork()) != 0) { - if (p != -1) + close(pipefd[1]); /* parent needs read pipe end. */ + if (p == -1) + close(pipefd[0]); + else { int i; +#ifdef HAVE_LINUX_NETWORK + /* The child process inherits the netlink socket, + which it never uses, but when the parent (us) + uses it in the future, the answer may go to the + child, resulting in the parent blocking + forever awaiting the result. To avoid this + the child closes the netlink socket, but there's + a nasty race, since the parent may use netlink + before the child has done the close. + + To avoid this, the parent blocks here until a + single byte comes back up the pipe, which + is sent by the child after it has closed the + netlink socket. */ + + unsigned char a; + read_write(pipefd[0], &a, 1, 1); +#endif + for (i = 0; i < MAX_PROCS; i++) - if (daemon->tcp_pids[i] == 0) + if (daemon->tcp_pids[i] == 0 && daemon->tcp_pipes[i] == -1) { daemon->tcp_pids[i] = p; + daemon->tcp_pipes[i] = pipefd[0]; break; } } - while (retry_send(close(confd))); + close(confd); /* The child can use up to TCP_MAX_QUERIES ids, so skip that many. */ daemon->log_id += TCP_MAX_QUERIES; } -#endif else { unsigned char *buff; @@ -1661,7 +1903,7 @@ static void check_dns_listeners(time_t now) int flags; struct in_addr netmask; int auth_dns; - + if (iface) { netmask = iface->netmask; @@ -1673,13 +1915,22 @@ static void check_dns_listeners(time_t now) auth_dns = 0; } -#ifndef NO_FORK - /* Arrange for SIGALARM after CHILD_LIFETIME seconds to + /* Arrange for SIGALRM after CHILD_LIFETIME seconds to terminate the process. */ if (!option_bool(OPT_DEBUG)) - alarm(CHILD_LIFETIME); -#endif + { +#ifdef HAVE_LINUX_NETWORK + /* See comment above re: netlink socket. */ + unsigned char a = 0; + close(daemon->netlinkfd); + read_write(pipefd[1], &a, 1, 0); +#endif + alarm(CHILD_LIFETIME); + close(pipefd[0]); /* close read end in child. */ + daemon->pipe_to_parent = pipefd[1]; + } + /* start with no upstream connections. */ for (s = daemon->servers; s; s = s->next) s->tcpfd = -1; @@ -1693,7 +1944,7 @@ static void check_dns_listeners(time_t now) buff = tcp_request(confd, now, &tcp_addr, netmask, auth_dns); shutdown(confd, SHUT_RDWR); - while (retry_send(close(confd))); + close(confd); if (buff) free(buff); @@ -1702,15 +1953,15 @@ static void check_dns_listeners(time_t now) if (s->tcpfd != -1) { shutdown(s->tcpfd, SHUT_RDWR); - while (retry_send(close(s->tcpfd))); + close(s->tcpfd); } -#ifndef NO_FORK + if (!option_bool(OPT_DEBUG)) { + close(daemon->pipe_to_parent); flush_log(); _exit(0); } -#endif } } } @@ -1739,29 +1990,15 @@ int icmp_ping(struct in_addr addr) { /* Try and get an ICMP echo from a machine. */ - /* Note that whilst in the three second wait, we check for - (and service) events on the DNS and TFTP sockets, (so doing that - better not use any resources our caller has in use...) - but we remain deaf to signals or further DHCP packets. */ - - /* There can be a problem using dnsmasq_time() to end the loop, since - it's not monotonic, and can go backwards if the system clock is - tweaked, leading to the code getting stuck in this loop and - ignoring DHCP requests. To fix this, we check to see if select returned - as a result of a timeout rather than a socket becoming available. We - only allow this to happen as many times as it takes to get to the wait time - in quarter-second chunks. This provides a fallback way to end loop. */ - - int fd, rc; + int fd; struct sockaddr_in saddr; struct { struct ip ip; struct icmp icmp; } packet; unsigned short id = rand16(); - unsigned int i, j, timeout_count; + unsigned int i, j; int gotreply = 0; - time_t start, now; #if defined(HAVE_LINUX_NETWORK) || defined (HAVE_SOLARIS_NETWORK) if ((fd = make_icmp_sock()) == -1) @@ -1791,14 +2028,46 @@ int icmp_ping(struct in_addr addr) while (retry_send(sendto(fd, (char *)&packet.icmp, sizeof(struct icmp), 0, (struct sockaddr *)&saddr, sizeof(saddr)))); - for (now = start = dnsmasq_time(), timeout_count = 0; - (difftime(now, start) < (float)PING_WAIT) && (timeout_count < PING_WAIT * 4);) + gotreply = delay_dhcp(dnsmasq_time(), PING_WAIT, fd, addr.s_addr, id); + +#if defined(HAVE_LINUX_NETWORK) || defined(HAVE_SOLARIS_NETWORK) + close(fd); +#else + opt = 1; + setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &opt, sizeof(opt)); +#endif + + return gotreply; +} + +int delay_dhcp(time_t start, int sec, int fd, uint32_t addr, unsigned short id) +{ + /* Delay processing DHCP packets for "sec" seconds counting from "start". + If "fd" is not -1 it will stop waiting if an ICMP echo reply is received + from "addr" with ICMP ID "id" and return 1 */ + + /* Note that whilst waiting, we check for + (and service) events on the DNS and TFTP sockets, (so doing that + better not use any resources our caller has in use...) + but we remain deaf to signals or further DHCP packets. */ + + /* There can be a problem using dnsmasq_time() to end the loop, since + it's not monotonic, and can go backwards if the system clock is + tweaked, leading to the code getting stuck in this loop and + ignoring DHCP requests. To fix this, we check to see if select returned + as a result of a timeout rather than a socket becoming available. We + only allow this to happen as many times as it takes to get to the wait time + in quarter-second chunks. This provides a fallback way to end loop. */ + + int rc, timeout_count; + time_t now; + + for (now = dnsmasq_time(), timeout_count = 0; + (difftime(now, start) <= (float)sec) && (timeout_count < sec * 4);) { - struct sockaddr_in faddr; - socklen_t len = sizeof(faddr); - poll_reset(); - poll_listen(fd, POLLIN); + if (fd != -1) + poll_listen(fd, POLLIN); set_dns_listeners(now); set_log_writer(); @@ -1815,10 +2084,10 @@ int icmp_ping(struct in_addr addr) timeout_count++; now = dnsmasq_time(); - + check_log_writer(0); check_dns_listeners(now); - + #ifdef HAVE_DHCP6 if (daemon->doing_ra && poll_check(daemon->icmp6fd, POLLIN)) icmp6_packet(now); @@ -1828,28 +2097,25 @@ int icmp_ping(struct in_addr addr) check_tftp_listeners(now); #endif - if (poll_check(fd, POLLIN) && - recvfrom(fd, &packet, sizeof(packet), 0, - (struct sockaddr *)&faddr, &len) == sizeof(packet) && - saddr.sin_addr.s_addr == faddr.sin_addr.s_addr && - packet.icmp.icmp_type == ICMP_ECHOREPLY && - packet.icmp.icmp_seq == 0 && - packet.icmp.icmp_id == id) - { - gotreply = 1; - break; + if (fd != -1) + { + struct { + struct ip ip; + struct icmp icmp; + } packet; + struct sockaddr_in faddr; + socklen_t len = sizeof(faddr); + + if (poll_check(fd, POLLIN) && + recvfrom(fd, &packet, sizeof(packet), 0, (struct sockaddr *)&faddr, &len) == sizeof(packet) && + addr == faddr.sin_addr.s_addr && + packet.icmp.icmp_type == ICMP_ECHOREPLY && + packet.icmp.icmp_seq == 0 && + packet.icmp.icmp_id == id) + return 1; } } - -#if defined(HAVE_LINUX_NETWORK) || defined(HAVE_SOLARIS_NETWORK) - while (retry_send(close(fd))); -#else - opt = 1; - setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &opt, sizeof(opt)); -#endif - return gotreply; + return 0; } -#endif - - +#endif /* HAVE_DHCP */