X-Git-Url: http://git.sameswireless.fr/l2tpns.git/blobdiff_plain/af2489f67ad3e5c83566b661ebd67806b6b96a0e..2b05f31dc724539454f67ea09fb2bd04e5a51e40:/l2tpns.c?ds=inline diff --git a/l2tpns.c b/l2tpns.c index 488287f..384e7ca 100644 --- a/l2tpns.c +++ b/l2tpns.c @@ -1,10 +1,10 @@ // L2TP Network Server // Adrian Kennard 2002 -// Copyright (c) 2003, 2004 Optus Internet Engineering +// Copyright (c) 2003, 2004, 2005 Optus Internet Engineering // Copyright (c) 2002 FireBrick (Andrews & Arnold Ltd / Watchfront Ltd) - GPL licenced // vim: sw=8 ts=8 -char const *cvs_id_l2tpns = "$Id: l2tpns.c,v 1.61 2004-11-29 03:55:21 bodea Exp $"; +char const *cvs_id_l2tpns = "$Id: l2tpns.c,v 1.128 2005-09-02 23:59:56 bodea Exp $"; #include #include @@ -19,6 +19,7 @@ char const *cvs_id_l2tpns = "$Id: l2tpns.c,v 1.61 2004-11-29 03:55:21 bodea Exp #include #include #include +#include #include #include #include @@ -51,7 +52,7 @@ char const *cvs_id_l2tpns = "$Id: l2tpns.c,v 1.61 2004-11-29 03:55:21 bodea Exp #ifdef BGP #include "bgp.h" -#endif /* BGP */ +#endif // Globals configt *config = NULL; // all configuration @@ -59,31 +60,39 @@ int tunfd = -1; // tun interface file handle. (network device) int udpfd = -1; // UDP file handle int controlfd = -1; // Control signal handle int clifd = -1; // Socket listening for CLI connections. +int daefd = -1; // Socket listening for DAE connections. int snoopfd = -1; // UDP file handle for sending out intercept data int *radfds = NULL; // RADIUS requests file handles int ifrfd = -1; // File descriptor for routing, etc +int ifr6fd = -1; // File descriptor for IPv6 routing, etc +int rand_fd = -1; // Random data source +int cluster_sockfd = -1; // Intra-cluster communications socket. +int epollfd = -1; // event polling time_t basetime = 0; // base clock char hostname[1000] = ""; // us. -static u32 sessionid = 0; // session id for radius accounting +static int tunidx; // ifr_ifindex of tun device static int syslog_log = 0; // are we logging to syslog -static FILE *log_stream = NULL; // file handle for direct logging (i.e. direct into file, not via syslog). -extern int cluster_sockfd; // Intra-cluster communications socket. -u32 last_id = 0; // Last used PPP SID. Can I kill this?? -- mo +static FILE *log_stream = 0; // file handle for direct logging (i.e. direct into file, not via syslog). +uint32_t last_id = 0; // Unique ID for radius accounting struct cli_session_actions *cli_session_actions = NULL; // Pending session changes requested by CLI struct cli_tunnel_actions *cli_tunnel_actions = NULL; // Pending tunnel changes required by CLI static void *ip_hash[256]; // Mapping from IP address to session structures. +struct ipv6radix { + int sess; + struct ipv6radix *branch; +} ipv6_hash[256]; // Mapping from IPv6 address to session structures. // Traffic counters. -static u32 udp_rx = 0, udp_rx_pkt = 0, udp_tx = 0; -static u32 eth_rx = 0, eth_rx_pkt = 0; -u32 eth_tx = 0; +static uint32_t udp_rx = 0, udp_rx_pkt = 0, udp_tx = 0; +static uint32_t eth_rx = 0, eth_rx_pkt = 0; +uint32_t eth_tx = 0; -static u32 ip_pool_size = 1; // Size of the pool of addresses used for dynamic address allocation. +static uint32_t ip_pool_size = 1; // Size of the pool of addresses used for dynamic address allocation. time_t time_now = 0; // Current time in seconds since epoch. static char time_now_string[64] = {0}; // Current time as a string. -static char main_quit = 0; // True if we're in the process of exiting. +char main_quit = 0; // True if we're in the process of exiting. linked_list *loaded_plugins; linked_list *plugins[MAX_PLUGIN_TYPES]; @@ -94,33 +103,43 @@ config_descriptt config_values[] = { CONFIG("debug", debug, INT), CONFIG("log_file", log_filename, STRING), CONFIG("pid_file", pid_file, STRING), + CONFIG("random_device", random_device, STRING), CONFIG("l2tp_secret", l2tpsecret, STRING), - CONFIG("primary_dns", default_dns1, IP), - CONFIG("secondary_dns", default_dns2, IP), - CONFIG("save_state", save_state, BOOL), - CONFIG("primary_radius", radiusserver[0], IP), - CONFIG("secondary_radius", radiusserver[1], IP), + CONFIG("ppp_restart_time", ppp_restart_time, INT), + CONFIG("ppp_max_configure", ppp_max_configure, INT), + CONFIG("ppp_max_failure", ppp_max_failure, INT), + CONFIG("primary_dns", default_dns1, IPv4), + CONFIG("secondary_dns", default_dns2, IPv4), + CONFIG("primary_radius", radiusserver[0], IPv4), + CONFIG("secondary_radius", radiusserver[1], IPv4), CONFIG("primary_radius_port", radiusport[0], SHORT), CONFIG("secondary_radius_port", radiusport[1], SHORT), CONFIG("radius_accounting", radius_accounting, BOOL), + CONFIG("radius_interim", radius_interim, INT), CONFIG("radius_secret", radiussecret, STRING), - CONFIG("bind_address", bind_address, IP), - CONFIG("peer_address", peer_address, IP), + CONFIG("radius_authtypes", radius_authtypes_s, STRING), + CONFIG("radius_dae_port", radius_dae_port, SHORT), + CONFIG("allow_duplicate_users", allow_duplicate_users, BOOL), + CONFIG("bind_address", bind_address, IPv4), + CONFIG("peer_address", peer_address, IPv4), CONFIG("send_garp", send_garp, BOOL), CONFIG("throttle_speed", rl_rate, UNSIGNED_LONG), CONFIG("throttle_buckets", num_tbfs, INT), CONFIG("accounting_dir", accounting_dir, STRING), CONFIG("setuid", target_uid, INT), CONFIG("dump_speed", dump_speed, BOOL), - CONFIG("cleanup_interval", cleanup_interval, INT), CONFIG("multi_read_count", multi_read_count, INT), CONFIG("scheduler_fifo", scheduler_fifo, BOOL), CONFIG("lock_pages", lock_pages, BOOL), CONFIG("icmp_rate", icmp_rate, INT), - CONFIG("cluster_address", cluster_address, IP), + CONFIG("packet_limit", max_packets, INT), + CONFIG("cluster_address", cluster_address, IPv4), CONFIG("cluster_interface", cluster_interface, STRING), + CONFIG("cluster_mcast_ttl", cluster_mcast_ttl, INT), CONFIG("cluster_hb_interval", cluster_hb_interval, INT), CONFIG("cluster_hb_timeout", cluster_hb_timeout, INT), + CONFIG("cluster_master_min_adv", cluster_master_min_adv, INT), + CONFIG("ipv6_prefix", ipv6_prefix, IPv6), { NULL, 0, 0, 0 }, }; @@ -135,61 +154,72 @@ static char *plugin_functions[] = { "plugin_kill_session", "plugin_control", "plugin_radius_response", + "plugin_radius_reset", "plugin_become_master", "plugin_new_session_master", }; #define max_plugin_functions (sizeof(plugin_functions) / sizeof(char *)) +// Counters for shutdown sessions +static sessiont shut_acct[8192]; +static sessionidt shut_acct_n = 0; + tunnelt *tunnel = NULL; // Array of tunnel structures. sessiont *session = NULL; // Array of session structures. -sessioncountt *sess_count = NULL; // Array of partial per-session traffic counters. +sessionlocalt *sess_local = NULL; // Array of local per-session counters. radiust *radius = NULL; // Array of radius structures. ippoolt *ip_address_pool = NULL; // Array of dynamic IP addresses. -ip_filtert *ip_filters = NULL; // Array of named filters. +ip_filtert *ip_filters = NULL; // Array of named filters. static controlt *controlfree = 0; struct Tstats *_statistics = NULL; #ifdef RINGBUFFER struct Tringbuffer *ringbuffer = NULL; #endif -static void cache_ipmap(ipt ip, int s); -static void uncache_ipmap(ipt ip); +static void cache_ipmap(in_addr_t ip, int s); +static void uncache_ipmap(in_addr_t ip); +static void cache_ipv6map(struct in6_addr ip, int prefixlen, int s); static void free_ip_address(sessionidt s); -static void dump_acct_info(void); +static void dump_acct_info(int all); static void sighup_handler(int sig); static void sigalrm_handler(int sig); -static void sigterm_handler(int sig); -static void sigquit_handler(int sig); +static void shutdown_handler(int sig); static void sigchild_handler(int sig); -static void read_state(void); -static void dump_state(void); -static void build_chap_response(char *challenge, u8 id, u16 challenge_length, char **challenge_response); +static void build_chap_response(uint8_t *challenge, uint8_t id, uint16_t challenge_length, uint8_t **challenge_response); static void update_config(void); static void read_config_file(void); static void initplugins(void); static int add_plugin(char *plugin_name); static int remove_plugin(char *plugin_name); static void plugins_done(void); -static void processcontrol(u8 * buf, int len, struct sockaddr_in *addr, int alen); +static void processcontrol(uint8_t *buf, int len, struct sockaddr_in *addr, int alen); static tunnelidt new_tunnel(void); -static int unhide_avp(u8 *avp, tunnelidt t, sessionidt s, u16 length); +static void unhide_value(uint8_t *value, size_t len, uint16_t type, uint8_t *vector, size_t vec_len); + +// on slaves, alow BGP to withdraw cleanly before exiting +#define QUIT_DELAY 5 -// return internal time (10ths since process startup) -static clockt now(void) +// quit actions (master) +#define QUIT_FAILOVER 1 // SIGTERM: exit when all control messages have been acked (for cluster failover) +#define QUIT_SHUTDOWN 2 // SIGQUIT: shutdown sessions/tunnels, reject new connections + +// return internal time (10ths since process startup), set f if given +static clockt now(double *f) { struct timeval t; gettimeofday(&t, 0); + if (f) *f = t.tv_sec + t.tv_usec / 1000000.0; return (t.tv_sec - basetime) * 10 + t.tv_usec / 100000 + 1; } // work out a retry time based on try number // This is a straight bounded exponential backoff. // Maximum re-try time is 32 seconds. (2^5). -clockt backoff(u8 try) +clockt backoff(uint8_t try) { if (try > 5) try = 5; // max backoff - return now() + 10 * (1 << try); + return now(NULL) + 10 * (1 << try); } @@ -199,7 +229,6 @@ clockt backoff(u8 try) void _log(int level, sessionidt s, tunnelidt t, const char *format, ...) { static char message[65536] = {0}; - static char message2[65536] = {0}; va_list ap; #ifdef RINGBUFFER @@ -223,25 +252,20 @@ void _log(int level, sessionidt s, tunnelidt t, const char *format, ...) if (config->debug < level) return; va_start(ap, format); + vsnprintf(message, sizeof(message), format, ap); + if (log_stream) - { - vsnprintf(message2, 65535, format, ap); - snprintf(message, 65535, "%s %02d/%02d %s", time_now_string, t, s, message2); - fprintf(log_stream, "%s", message); - } + fprintf(log_stream, "%s %02d/%02d %s", time_now_string, t, s, message); else if (syslog_log) - { - vsnprintf(message2, 65535, format, ap); - snprintf(message, 65535, "%02d/%02d %s", t, s, message2); - syslog(level + 2, message); // We don't need LOG_EMERG or LOG_ALERT - } + syslog(level + 2, "%02d/%02d %s", t, s, message); // We don't need LOG_EMERG or LOG_ALERT + va_end(ap); } -void _log_hex(int level, const char *title, const char *data, int maxsize) +void _log_hex(int level, const char *title, const uint8_t *data, int maxsize) { int i, j; - const u8 *d = (const u8 *)data; + const uint8_t *d = data; if (config->debug < level) return; @@ -289,6 +313,76 @@ void _log_hex(int level, const char *title, const char *data, int maxsize) } } +// update a counter, accumulating 2^32 wraps +void increment_counter(uint32_t *counter, uint32_t *wrap, uint32_t delta) +{ + uint32_t new = *counter + delta; + if (new < *counter) + (*wrap)++; + + *counter = new; +} + +// initialise the random generator +static void initrandom(char *source) +{ + static char path[sizeof(config->random_device)] = "*undefined*"; + + // reinitialise only if we are forced to do so or if the config has changed + if (source && !strncmp(path, source, sizeof(path))) + return; + + // close previous source, if any + if (rand_fd >= 0) + close(rand_fd); + + rand_fd = -1; + + if (source) + { + // register changes + snprintf(path, sizeof(path), "%s", source); + + if (*path == '/') + { + rand_fd = open(path, O_RDONLY|O_NONBLOCK); + if (rand_fd < 0) + LOG(0, 0, 0, "Error opening the random device %s: %s\n", + path, strerror(errno)); + } + } +} + +// fill buffer with random data +void random_data(uint8_t *buf, int len) +{ + int n = 0; + + CSTAT(random_data); + if (rand_fd >= 0) + { + n = read(rand_fd, buf, len); + if (n >= len) return; + if (n < 0) + { + if (errno != EAGAIN) + { + LOG(0, 0, 0, "Error reading from random source: %s\n", + strerror(errno)); + + // fall back to rand() + initrandom(NULL); + } + + n = 0; + } + } + + // append missing data + while (n < len) + // not using the low order bits from the prng stream + buf[n++] = (rand() >> 4) & 0xff; +} // Add a route // @@ -298,7 +392,7 @@ void _log_hex(int level, const char *title, const char *data, int maxsize) // // 'ip' and 'mask' must be in _host_ order. // -static void routeset(sessionidt s, ipt ip, ipt mask, ipt gw, u8 add) +static void routeset(sessionidt s, in_addr_t ip, in_addr_t mask, in_addr_t gw, int add) { struct rtentry r; int i; @@ -310,11 +404,11 @@ static void routeset(sessionidt s, ipt ip, ipt mask, ipt gw, u8 add) memset(&r, 0, sizeof(r)); r.rt_dev = config->tundevice; r.rt_dst.sa_family = AF_INET; - *(u32 *) & (((struct sockaddr_in *) &r.rt_dst)->sin_addr.s_addr) = htonl(ip); + *(uint32_t *) & (((struct sockaddr_in *) &r.rt_dst)->sin_addr.s_addr) = htonl(ip); r.rt_gateway.sa_family = AF_INET; - *(u32 *) & (((struct sockaddr_in *) &r.rt_gateway)->sin_addr.s_addr) = htonl(gw); + *(uint32_t *) & (((struct sockaddr_in *) &r.rt_gateway)->sin_addr.s_addr) = htonl(gw); r.rt_genmask.sa_family = AF_INET; - *(u32 *) & (((struct sockaddr_in *) &r.rt_genmask)->sin_addr.s_addr) = htonl(mask); + *(uint32_t *) & (((struct sockaddr_in *) &r.rt_genmask)->sin_addr.s_addr) = htonl(mask); r.rt_flags = (RTF_UP | RTF_STATIC); if (gw) r.rt_flags |= RTF_GATEWAY; @@ -353,11 +447,61 @@ static void routeset(sessionidt s, ipt ip, ipt mask, ipt gw, u8 add) } } +void route6set(sessionidt s, struct in6_addr ip, int prefixlen, int add) +{ + struct in6_rtmsg rt; + char ipv6addr[INET6_ADDRSTRLEN]; + + if (ifr6fd < 0) + { + LOG(0, 0, 0, "Asked to set IPv6 route, but IPv6 not setup.\n"); + return; + } + + memset(&rt, 0, sizeof(rt)); + + memcpy(&rt.rtmsg_dst, &ip, sizeof(struct in6_addr)); + rt.rtmsg_dst_len = prefixlen; + rt.rtmsg_metric = 1; + rt.rtmsg_flags = RTF_UP; + rt.rtmsg_ifindex = tunidx; + + LOG(1, 0, 0, "Route %s %s/%d\n", + add ? "add" : "del", + inet_ntop(AF_INET6, &ip, ipv6addr, INET6_ADDRSTRLEN), + prefixlen); + + if (ioctl(ifr6fd, add ? SIOCADDRT : SIOCDELRT, (void *) &rt) < 0) + LOG(0, 0, 0, "route6set() error in ioctl: %s\n", + strerror(errno)); + + // FIXME: need to add BGP routing (RFC2858) + + if (s) + { + if (!add) // Are we deleting a route? + s = 0; // Caching the session as '0' is the same as uncaching. + + cache_ipv6map(ip, prefixlen, s); + } + + return; +} + +// defined in linux/ipv6.h, but tricky to include from user-space +// TODO: move routing to use netlink rather than ioctl +struct in6_ifreq { + struct in6_addr ifr6_addr; + __u32 ifr6_prefixlen; + unsigned int ifr6_ifindex; +}; + // // Set up TUN interface static void inittun(void) { struct ifreq ifr; + struct in6_ifreq ifr6; struct sockaddr_in sin = {0}; memset(&ifr, 0, sizeof(ifr)); ifr.ifr_flags = IFF_TUN; @@ -403,9 +547,45 @@ static void inittun(void) LOG(0, 0, 0, "Error setting tun flags: %s\n", strerror(errno)); exit(1); } + if (ioctl(ifrfd, SIOCGIFINDEX, (void *) &ifr) < 0) + { + LOG(0, 0, 0, "Error getting tun ifindex: %s\n", strerror(errno)); + exit(1); + } + tunidx = ifr.ifr_ifindex; + + // Only setup IPv6 on the tun device if we have a configured prefix + if (config->ipv6_prefix.s6_addr[0]) { + ifr6fd = socket(PF_INET6, SOCK_DGRAM, 0); + + // Link local address is FE80::1 + memset(&ifr6.ifr6_addr, 0, sizeof(ifr6.ifr6_addr)); + ifr6.ifr6_addr.s6_addr[0] = 0xFE; + ifr6.ifr6_addr.s6_addr[1] = 0x80; + ifr6.ifr6_addr.s6_addr[15] = 1; + ifr6.ifr6_prefixlen = 64; + ifr6.ifr6_ifindex = ifr.ifr_ifindex; + if (ioctl(ifr6fd, SIOCSIFADDR, (void *) &ifr6) < 0) + { + LOG(0, 0, 0, "Error setting tun IPv6 link local address:" + " %s\n", strerror(errno)); + } + + // Global address is prefix::1 + memset(&ifr6.ifr6_addr, 0, sizeof(ifr6.ifr6_addr)); + ifr6.ifr6_addr = config->ipv6_prefix; + ifr6.ifr6_addr.s6_addr[15] = 1; + ifr6.ifr6_prefixlen = 64; + ifr6.ifr6_ifindex = ifr.ifr_ifindex; + if (ioctl(ifr6fd, SIOCSIFADDR, (void *) &ifr6) < 0) + { + LOG(0, 0, 0, "Error setting tun IPv6 global address: %s\n", + strerror(errno)); + } + } } -// set up UDP port +// set up UDP ports static void initudp(void) { int on = 1; @@ -427,7 +607,6 @@ static void initudp(void) LOG(0, 0, 0, "Error in UDP bind: %s\n", strerror(errno)); exit(1); } - snoopfd = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP); // Control memset(&addr, 0, sizeof(addr)); @@ -440,6 +619,21 @@ static void initudp(void) LOG(0, 0, 0, "Error in control bind: %s\n", strerror(errno)); exit(1); } + + // Dynamic Authorization Extensions to RADIUS + memset(&addr, 0, sizeof(addr)); + addr.sin_family = AF_INET; + addr.sin_port = htons(config->radius_dae_port); + daefd = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP); + setsockopt(daefd, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on)); + if (bind(daefd, (void *) &addr, sizeof(addr)) < 0) + { + LOG(0, 0, 0, "Error in DAE bind: %s\n", strerror(errno)); + exit(1); + } + + // Intercept + snoopfd = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP); } // @@ -457,27 +651,74 @@ static void initudp(void) // IP address. // -static int lookup_ipmap(ipt ip) +static int lookup_ipmap(in_addr_t ip) +{ + uint8_t *a = (uint8_t *) &ip; + uint8_t **d = (uint8_t **) ip_hash; + + if (!(d = (uint8_t **) d[(size_t) *a++])) return 0; + if (!(d = (uint8_t **) d[(size_t) *a++])) return 0; + if (!(d = (uint8_t **) d[(size_t) *a++])) return 0; + + return (int) (intptr_t) d[(size_t) *a]; +} + +static int lookup_ipv6map(struct in6_addr ip) { - u8 *a = (u8 *)&ip; - char **d = (char **) ip_hash; + struct ipv6radix *curnode; + int i; int s; + char ipv6addr[INET6_ADDRSTRLEN]; + + curnode = &ipv6_hash[ip.s6_addr[0]]; + i = 1; + s = curnode->sess; + + while (s == 0 && i < 15 && curnode->branch != NULL) + { + curnode = &curnode->branch[ip.s6_addr[i]]; + s = curnode->sess; + i++; + } - if (!(d = (char **) d[(size_t) *a++])) return 0; - if (!(d = (char **) d[(size_t) *a++])) return 0; - if (!(d = (char **) d[(size_t) *a++])) return 0; + LOG(4, s, session[s].tunnel, "Looking up address %s and got %d\n", + inet_ntop(AF_INET6, &ip, ipv6addr, + INET6_ADDRSTRLEN), + s); - s = (ipt) d[(size_t) *a]; return s; } -sessionidt sessionbyip(ipt ip) +sessionidt sessionbyip(in_addr_t ip) { int s = lookup_ipmap(ip); - CSTAT(call_sessionbyip); + CSTAT(sessionbyip); + + if (s > 0 && s < MAXSESSION && session[s].opened) + return (sessionidt) s; + + return 0; +} + +sessionidt sessionbyipv6(struct in6_addr ip) +{ + int s; + CSTAT(sessionbyipv6); - if (s > 0 && s < MAXSESSION && session[s].tunnel) + if (!memcmp(&config->ipv6_prefix, &ip, 8) || + (ip.s6_addr[0] == 0xFE && + ip.s6_addr[1] == 0x80 && + ip.s6_addr16[1] == 0 && + ip.s6_addr16[2] == 0 && + ip.s6_addr16[3] == 0)) { + s = lookup_ipmap(*(in_addr_t *) &ip.s6_addr[8]); + } else { + s = lookup_ipv6map(ip); + } + + if (s > 0 && s < MAXSESSION && session[s].opened) return s; + return 0; } @@ -487,25 +728,25 @@ sessionidt sessionbyip(ipt ip) // // (It's actually cached in network order) // -static void cache_ipmap(ipt ip, int s) +static void cache_ipmap(in_addr_t ip, int s) { - ipt nip = htonl(ip); // MUST be in network order. I.e. MSB must in be ((char*)(&ip))[0] - u8 *a = (u8 *) &nip; - char **d = (char **) ip_hash; + in_addr_t nip = htonl(ip); // MUST be in network order. I.e. MSB must in be ((char *) (&ip))[0] + uint8_t *a = (uint8_t *) &nip; + uint8_t **d = (uint8_t **) ip_hash; int i; for (i = 0; i < 3; i++) { if (!d[(size_t) a[i]]) { - if (!(d[(size_t) a[i]] = calloc(256, sizeof (void *)))) + if (!(d[(size_t) a[i]] = calloc(256, sizeof(void *)))) return; } - d = (char **) d[(size_t) a[i]]; + d = (uint8_t **) d[(size_t) a[i]]; } - d[(size_t) a[3]] = (char *)((int)s); + d[(size_t) a[3]] = (uint8_t *) (intptr_t) s; if (s > 0) LOG(4, s, session[s].tunnel, "Caching ip address %s\n", fmtaddr(nip, 0)); @@ -515,11 +756,47 @@ static void cache_ipmap(ipt ip, int s) // else a map to an ip pool index. } -static void uncache_ipmap(ipt ip) +static void uncache_ipmap(in_addr_t ip) { cache_ipmap(ip, 0); // Assign it to the NULL session. } +static void cache_ipv6map(struct in6_addr ip, int prefixlen, int s) +{ + int i; + int bytes; + struct ipv6radix *curnode; + char ipv6addr[INET6_ADDRSTRLEN]; + + curnode = &ipv6_hash[ip.s6_addr[0]]; + + bytes = prefixlen >> 3; + i = 1; + while (i < bytes) { + if (curnode->branch == NULL) + { + if (!(curnode->branch = calloc(256, + sizeof (struct ipv6radix)))) + return; + } + curnode = &curnode->branch[ip.s6_addr[i]]; + i++; + } + + curnode->sess = s; + + if (s > 0) + LOG(4, s, session[s].tunnel, "Caching ip address %s/%d\n", + inet_ntop(AF_INET6, &ip, ipv6addr, + INET6_ADDRSTRLEN), + prefixlen); + else if (s == 0) + LOG(4, 0, 0, "Un-caching ip address %s/%d\n", + inet_ntop(AF_INET6, &ip, ipv6addr, + INET6_ADDRSTRLEN), + prefixlen); +} + // // CLI list to dump current ipcache. // @@ -538,22 +815,22 @@ int cmd_show_ipcache(struct cli_def *cli, char *command, char **argv, int argc) { if (!d[i]) continue; - e = (char**) d[i]; + e = (char **) d[i]; for (j = 0; j < 256; ++j) { if (!e[j]) continue; - f = (char**) e[j]; + f = (char **) e[j]; for (k = 0; k < 256; ++k) { if (!f[k]) continue; - g = (char**)f[k]; + g = (char **)f[k]; for (l = 0; l < 256; ++l) { if (!g[l]) continue; - cli_print(cli, "%7d %d.%d.%d.%d", (int) g[l], i, j, k, l); + cli_print(cli, "%7d %d.%d.%d.%d", (int) (intptr_t) g[l], i, j, k, l); ++count; } } @@ -573,10 +850,13 @@ int cmd_show_ipcache(struct cli_def *cli, char *command, char **argv, int argc) sessionidt sessionbyuser(char *username) { int s; - CSTAT(call_sessionbyuser); + CSTAT(sessionbyuser); - for (s = 1; s < MAXSESSION ; ++s) + for (s = 1; s <= config->cluster_highest_sessionid ; ++s) { + if (!session[s].opened) + continue; + if (session[s].walled_garden) continue; // Skip walled garden users. @@ -587,11 +867,11 @@ sessionidt sessionbyuser(char *username) return 0; // Not found. } -void send_garp(ipt ip) +void send_garp(in_addr_t ip) { int s; struct ifreq ifr; - u8 mac[6]; + uint8_t mac[6]; s = socket(PF_INET, SOCK_DGRAM, 0); if (s < 0) @@ -618,26 +898,25 @@ void send_garp(ipt ip) sendarp(ifr.ifr_ifindex, mac, ip); } -// Find session by username, 0 for not found static sessiont *sessiontbysessionidt(sessionidt s) { - if (!s || s > MAXSESSION) return NULL; + if (!s || s >= MAXSESSION) return NULL; return &session[s]; } static sessionidt sessionidtbysessiont(sessiont *s) { sessionidt val = s-session; - if (s < session || val > MAXSESSION) return 0; + if (s < session || val >= MAXSESSION) return 0; return val; } // actually send a control message for a specific tunnel -void tunnelsend(u8 * buf, u16 l, tunnelidt t) +void tunnelsend(uint8_t * buf, uint16_t l, tunnelidt t) { struct sockaddr_in addr; - CSTAT(call_tunnelsend); + CSTAT(tunnelsend); if (!t) { @@ -659,11 +938,11 @@ void tunnelsend(u8 * buf, u16 l, tunnelidt t) memset(&addr, 0, sizeof(addr)); addr.sin_family = AF_INET; - *(u32 *) & addr.sin_addr = htonl(tunnel[t].ip); + *(uint32_t *) & addr.sin_addr = htonl(tunnel[t].ip); addr.sin_port = htons(tunnel[t].port); // sequence expected, if sequence in message - if (*buf & 0x08) *(u16 *) (buf + ((*buf & 0x40) ? 10 : 8)) = htons(tunnel[t].nr); + if (*buf & 0x08) *(uint16_t *) (buf + ((*buf & 0x40) ? 10 : 8)) = htons(tunnel[t].nr); // If this is a control message, deal with retries if (*buf & 0x80) @@ -679,7 +958,7 @@ void tunnelsend(u8 * buf, u16 l, tunnelidt t) if (sendto(udpfd, buf, l, 0, (void *) &addr, sizeof(addr)) < 0) { - LOG(0, ntohs((*(u16 *) (buf + 6))), t, "Error sending data out tunnel: %s (udpfd=%d, buf=%p, len=%d, dest=%s)\n", + LOG(0, ntohs((*(uint16_t *) (buf + 6))), t, "Error sending data out tunnel: %s (udpfd=%d, buf=%p, len=%d, dest=%s)\n", strerror(errno), udpfd, buf, l, inet_ntoa(addr.sin_addr)); STAT(tunnel_tx_errors); return; @@ -694,37 +973,37 @@ void tunnelsend(u8 * buf, u16 l, tunnelidt t) // Tiny helper function to write data to // the 'tun' device. // -int tun_write(u8 * data, int size) +int tun_write(uint8_t * data, int size) { return write(tunfd, data, size); } // process outgoing (to tunnel) IP // -static void processipout(u8 * buf, int len) +static void processipout(uint8_t *buf, int len) { sessionidt s; sessiont *sp; tunnelidt t; - ipt ip; + in_addr_t ip; - char * data = buf; // Keep a copy of the originals. + uint8_t *data = buf; // Keep a copy of the originals. int size = len; - u8 b[MAXETHER + 20]; + uint8_t b[MAXETHER + 20]; - CSTAT(call_processipout); + CSTAT(processipout); if (len < MIN_IP_SIZE) { LOG(1, 0, 0, "Short IP, %d bytes\n", len); - STAT(tunnel_tx_errors); + STAT(tun_rx_errors); return; } if (len >= MAXETHER) { LOG(1, 0, 0, "Oversize IP packet %d bytes\n", len); - STAT(tunnel_tx_errors); + STAT(tun_rx_errors); return; } @@ -733,13 +1012,13 @@ static void processipout(u8 * buf, int len) len -= 4; // Got an IP header now - if (*(u8 *)(buf) >> 4 != 4) + if (*(uint8_t *)(buf) >> 4 != 4) { LOG(1, 0, 0, "IP: Don't understand anything except IPv4\n"); return; } - ip = *(u32 *)(buf + 16); + ip = *(uint32_t *)(buf + 16); if (!(s = sessionbyip(ip))) { // Is this a packet for a session that doesn't exist? @@ -754,14 +1033,54 @@ static void processipout(u8 * buf, int len) if (rate++ < config->icmp_rate) // Only send a max of icmp_rate per second. { - LOG(4, 0, 0, "IP: Sending ICMP host unreachable to %s\n", fmtaddr(*(u32 *)(buf + 12), 0)); - host_unreachable(*(u32 *)(buf + 12), *(u16 *)(buf + 4), ip, buf, (len < 64) ? 64 : len); + LOG(4, 0, 0, "IP: Sending ICMP host unreachable to %s\n", fmtaddr(*(in_addr_t *)(buf + 12), 0)); + host_unreachable(*(in_addr_t *)(buf + 12), *(uint16_t *)(buf + 4), + config->bind_address ? config->bind_address : my_address, buf, len); } return; } t = session[s].tunnel; sp = &session[s]; + // DoS prevention: enforce a maximum number of packets per 0.1s for a session + if (config->max_packets > 0) + { + if (sess_local[s].last_packet_out == TIME) + { + int max = config->max_packets; + + // All packets for throttled sessions are handled by the + // master, so further limit by using the throttle rate. + // A bit of a kludge, since throttle rate is in kbps, + // but should still be generous given our average DSL + // packet size is 200 bytes: a limit of 28kbps equates + // to around 180 packets per second. + if (!config->cluster_iam_master && sp->throttle_out && sp->throttle_out < max) + max = sp->throttle_out; + + if (++sess_local[s].packets_out > max) + { + sess_local[s].packets_dropped++; + return; + } + } + else + { + if (sess_local[s].packets_dropped) + { + INC_STAT(tun_rx_dropped, sess_local[s].packets_dropped); + LOG(3, s, t, "Dropped %u/%u packets to %s for %suser %s\n", + sess_local[s].packets_dropped, sess_local[s].packets_out, + fmtaddr(ip, 0), sp->throttle_out ? "throttled " : "", + sp->user); + } + + sess_local[s].last_packet_out = TIME; + sess_local[s].packets_out = 1; + sess_local[s].packets_dropped = 0; + } + } + // run access-list if any if (session[s].filter_out && !ip_filter(buf, len, session[s].filter_out - 1)) return; @@ -786,7 +1105,118 @@ static void processipout(u8 * buf, int len) // Add on L2TP header { - u8 *p = makeppp(b, sizeof(b), buf, len, t, s, PPPIP); + uint8_t *p = makeppp(b, sizeof(b), buf, len, s, t, PPPIP); + if (!p) return; + tunnelsend(b, len + (p-b), t); // send it... + } + + // Snooping this session, send it to intercept box + if (sp->snoop_ip && sp->snoop_port) + snoop_send_packet(buf, len, sp->snoop_ip, sp->snoop_port); + + increment_counter(&sp->cout, &sp->cout_wrap, len); // byte count + sp->cout_delta += len; + sp->pout++; + udp_tx += len; + + sess_local[s].cout += len; // To send to master.. + sess_local[s].pout++; +} + +// process outgoing (to tunnel) IPv6 +// +static void processipv6out(uint8_t * buf, int len) +{ + sessionidt s; + sessiont *sp; + tunnelidt t; + in_addr_t ip; + struct in6_addr ip6; + + uint8_t *data = buf; // Keep a copy of the originals. + int size = len; + + uint8_t b[MAXETHER + 20]; + + CSTAT(processipv6out); + + if (len < MIN_IP_SIZE) + { + LOG(1, 0, 0, "Short IPv6, %d bytes\n", len); + STAT(tunnel_tx_errors); + return; + } + if (len >= MAXETHER) + { + LOG(1, 0, 0, "Oversize IPv6 packet %d bytes\n", len); + STAT(tunnel_tx_errors); + return; + } + + // Skip the tun header + buf += 4; + len -= 4; + + // Got an IP header now + if (*(uint8_t *)(buf) >> 4 != 6) + { + LOG(1, 0, 0, "IP: Don't understand anything except IPv6\n"); + return; + } + + ip6 = *(struct in6_addr *)(buf+24); + s = sessionbyipv6(ip6); + + if (s == 0) + { + ip = *(uint32_t *)(buf + 32); + s = sessionbyip(ip); + } + + if (s == 0) + { + // Is this a packet for a session that doesn't exist? + static int rate = 0; // Number of ICMP packets we've sent this second. + static int last = 0; // Last time we reset the ICMP packet counter 'rate'. + + if (last != time_now) + { + last = time_now; + rate = 0; + } + + if (rate++ < config->icmp_rate) // Only send a max of icmp_rate per second. + { + // FIXME: Should send icmp6 host unreachable + } + return; + } + t = session[s].tunnel; + sp = &session[s]; + + // FIXME: add DoS prevention/filters? + + if (sp->tbf_out) + { + // Are we throttling this session? + if (config->cluster_iam_master) + tbf_queue_packet(sp->tbf_out, data, size); + else + master_throttle_packet(sp->tbf_out, data, size); + return; + } + else if (sp->walled_garden && !config->cluster_iam_master) + { + // We are walled-gardening this + master_garden_packet(s, data, size); + return; + } + + LOG(5, s, t, "Ethernet -> Tunnel (%d bytes)\n", len); + + // Add on L2TP header + { + uint8_t *p = makeppp(b, sizeof(b), buf, len, s, t, PPPIPV6); if (!p) return; tunnelsend(b, len + (p-b), t); // send it... } @@ -795,24 +1225,26 @@ static void processipout(u8 * buf, int len) if (sp->snoop_ip && sp->snoop_port) snoop_send_packet(buf, len, sp->snoop_ip, sp->snoop_port); - sp->cout += len; // byte count - sp->total_cout += len; // byte count + increment_counter(&sp->cout, &sp->cout_wrap, len); // byte count + sp->cout_delta += len; sp->pout++; udp_tx += len; - sess_count[s].cout += len; // To send to master.. + + sess_local[s].cout += len; // To send to master.. + sess_local[s].pout++; } // // Helper routine for the TBF filters. // Used to send queued data in to the user! // -static void send_ipout(sessionidt s, u8 *buf, int len) +static void send_ipout(sessionidt s, uint8_t *buf, int len) { sessiont *sp; tunnelidt t; - ipt ip; + in_addr_t ip; - u8 b[MAXETHER + 20]; + uint8_t b[MAXETHER + 20]; if (len < 0 || len > MAXETHER) { @@ -824,7 +1256,7 @@ static void send_ipout(sessionidt s, u8 *buf, int len) buf += 4; len -= 4; - ip = *(u32 *)(buf + 16); + ip = *(in_addr_t *)(buf + 16); if (!session[s].ip) return; @@ -836,7 +1268,7 @@ static void send_ipout(sessionidt s, u8 *buf, int len) // Add on L2TP header { - u8 *p = makeppp(b, sizeof(b), buf, len, t, s, PPPIP); + uint8_t *p = makeppp(b, sizeof(b), buf, len, s, t, PPPIP); if (!p) return; tunnelsend(b, len + (p-b), t); // send it... } @@ -845,59 +1277,61 @@ static void send_ipout(sessionidt s, u8 *buf, int len) if (sp->snoop_ip && sp->snoop_port) snoop_send_packet(buf, len, sp->snoop_ip, sp->snoop_port); - sp->cout += len; // byte count - sp->total_cout += len; // byte count + increment_counter(&sp->cout, &sp->cout_wrap, len); // byte count + sp->cout_delta += len; sp->pout++; udp_tx += len; - sess_count[s].cout += len; // To send to master.. + + sess_local[s].cout += len; // To send to master.. + sess_local[s].pout++; } // add an AVP (16 bit) -static void control16(controlt * c, u16 avp, u16 val, u8 m) +static void control16(controlt * c, uint16_t avp, uint16_t val, uint8_t m) { - u16 l = (m ? 0x8008 : 0x0008); - *(u16 *) (c->buf + c->length + 0) = htons(l); - *(u16 *) (c->buf + c->length + 2) = htons(0); - *(u16 *) (c->buf + c->length + 4) = htons(avp); - *(u16 *) (c->buf + c->length + 6) = htons(val); + uint16_t l = (m ? 0x8008 : 0x0008); + *(uint16_t *) (c->buf + c->length + 0) = htons(l); + *(uint16_t *) (c->buf + c->length + 2) = htons(0); + *(uint16_t *) (c->buf + c->length + 4) = htons(avp); + *(uint16_t *) (c->buf + c->length + 6) = htons(val); c->length += 8; } // add an AVP (32 bit) -static void control32(controlt * c, u16 avp, u32 val, u8 m) +static void control32(controlt * c, uint16_t avp, uint32_t val, uint8_t m) { - u16 l = (m ? 0x800A : 0x000A); - *(u16 *) (c->buf + c->length + 0) = htons(l); - *(u16 *) (c->buf + c->length + 2) = htons(0); - *(u16 *) (c->buf + c->length + 4) = htons(avp); - *(u32 *) (c->buf + c->length + 6) = htonl(val); + uint16_t l = (m ? 0x800A : 0x000A); + *(uint16_t *) (c->buf + c->length + 0) = htons(l); + *(uint16_t *) (c->buf + c->length + 2) = htons(0); + *(uint16_t *) (c->buf + c->length + 4) = htons(avp); + *(uint32_t *) (c->buf + c->length + 6) = htonl(val); c->length += 10; } -// add an AVP (32 bit) -static void controls(controlt * c, u16 avp, char *val, u8 m) +// add an AVP (string) +static void controls(controlt * c, uint16_t avp, char *val, uint8_t m) { - u16 l = ((m ? 0x8000 : 0) + strlen(val) + 6); - *(u16 *) (c->buf + c->length + 0) = htons(l); - *(u16 *) (c->buf + c->length + 2) = htons(0); - *(u16 *) (c->buf + c->length + 4) = htons(avp); + uint16_t l = ((m ? 0x8000 : 0) + strlen(val) + 6); + *(uint16_t *) (c->buf + c->length + 0) = htons(l); + *(uint16_t *) (c->buf + c->length + 2) = htons(0); + *(uint16_t *) (c->buf + c->length + 4) = htons(avp); memcpy(c->buf + c->length + 6, val, strlen(val)); c->length += 6 + strlen(val); } // add a binary AVP -static void controlb(controlt * c, u16 avp, char *val, unsigned int len, u8 m) +static void controlb(controlt * c, uint16_t avp, uint8_t *val, unsigned int len, uint8_t m) { - u16 l = ((m ? 0x8000 : 0) + len + 6); - *(u16 *) (c->buf + c->length + 0) = htons(l); - *(u16 *) (c->buf + c->length + 2) = htons(0); - *(u16 *) (c->buf + c->length + 4) = htons(avp); + uint16_t l = ((m ? 0x8000 : 0) + len + 6); + *(uint16_t *) (c->buf + c->length + 0) = htons(l); + *(uint16_t *) (c->buf + c->length + 2) = htons(0); + *(uint16_t *) (c->buf + c->length + 4) = htons(avp); memcpy(c->buf + c->length + 6, val, len); c->length += 6 + len; } // new control connection -static controlt *controlnew(u16 mtype) +static controlt *controlnew(uint16_t mtype) { controlt *c; if (!controlfree) @@ -909,7 +1343,7 @@ static controlt *controlnew(u16 mtype) } assert(c); c->next = 0; - *(u16 *) (c->buf + 0) = htons(0xC802); // flags/ver + *(uint16_t *) (c->buf + 0) = htons(0xC802); // flags/ver c->length = 12; control16(c, 0, mtype, 1); return c; @@ -919,26 +1353,26 @@ static controlt *controlnew(u16 mtype) // (ZLB send). static void controlnull(tunnelidt t) { - u8 buf[12]; + uint8_t buf[12]; if (tunnel[t].controlc) // Messages queued; They will carry the ack. return; - *(u16 *) (buf + 0) = htons(0xC802); // flags/ver - *(u16 *) (buf + 2) = htons(12); // length - *(u16 *) (buf + 4) = htons(tunnel[t].far); // tunnel - *(u16 *) (buf + 6) = htons(0); // session - *(u16 *) (buf + 8) = htons(tunnel[t].ns); // sequence - *(u16 *) (buf + 10) = htons(tunnel[t].nr); // sequence + *(uint16_t *) (buf + 0) = htons(0xC802); // flags/ver + *(uint16_t *) (buf + 2) = htons(12); // length + *(uint16_t *) (buf + 4) = htons(tunnel[t].far); // tunnel + *(uint16_t *) (buf + 6) = htons(0); // session + *(uint16_t *) (buf + 8) = htons(tunnel[t].ns); // sequence + *(uint16_t *) (buf + 10) = htons(tunnel[t].nr); // sequence tunnelsend(buf, 12, t); } // add a control message to a tunnel, and send if within window -static void controladd(controlt * c, tunnelidt t, sessionidt s) +static void controladd(controlt *c, sessionidt far, tunnelidt t) { - *(u16 *) (c->buf + 2) = htons(c->length); // length - *(u16 *) (c->buf + 4) = htons(tunnel[t].far); // tunnel - *(u16 *) (c->buf + 6) = htons(s ? session[s].far : 0); // session - *(u16 *) (c->buf + 8) = htons(tunnel[t].ns); // sequence + *(uint16_t *) (c->buf + 2) = htons(c->length); // length + *(uint16_t *) (c->buf + 4) = htons(tunnel[t].far); // tunnel + *(uint16_t *) (c->buf + 6) = htons(far); // session + *(uint16_t *) (c->buf + 8) = htons(tunnel[t].ns); // sequence tunnel[t].ns++; // advance sequence // link in message in to queue if (tunnel[t].controlc) @@ -969,7 +1403,7 @@ static void controladd(controlt * c, tunnelidt t, sessionidt s) // void throttle_session(sessionidt s, int rate_in, int rate_out) { - if (!session[s].tunnel) + if (!session[s].opened) return; // No-one home. if (!*session[s].user) @@ -1007,7 +1441,7 @@ void throttle_session(sessionidt s, int rate_in, int rate_out) // add/remove filters from session (-1 = no change) void filter_session(sessionidt s, int filter_in, int filter_out) { - if (!session[s].tunnel) + if (!session[s].opened) return; // No-one home. if (!*session[s].user) @@ -1043,16 +1477,16 @@ void filter_session(sessionidt s, int filter_in, int filter_out) } // start tidy shutdown of session -void sessionshutdown(sessionidt s, char *reason) +void sessionshutdown(sessionidt s, char *reason, int result, int error) { int walled_garden = session[s].walled_garden; - CSTAT(call_sessionshutdown); + CSTAT(sessionshutdown); - if (!session[s].tunnel) + if (!session[s].opened) { - LOG(3, s, session[s].tunnel, "Called sessionshutdown on a session with no tunnel.\n"); + LOG(3, s, session[s].tunnel, "Called sessionshutdown on an unopened session.\n"); return; // not a live session } @@ -1063,26 +1497,22 @@ void sessionshutdown(sessionidt s, char *reason) run_plugins(PLUGIN_KILL_SESSION, &data); } - // RADIUS Stop message - if (session[s].opened && !walled_garden && !session[s].die) + if (session[s].ip && !walled_garden && !session[s].die) { - u16 r = session[s].radius; - if (!r) + // RADIUS Stop message + uint16_t r = radiusnew(s); + if (r) { - if (!(r = radiusnew(s))) - { - LOG(1, s, session[s].tunnel, "No free RADIUS sessions for Stop message\n"); - STAT(radius_overflow); - } - else - { - int n; - for (n = 0; n < 15; n++) - radius[r].auth[n] = rand(); - } + // stop, if not already trying + if (radius[r].state != RADIUSSTOP) + radiussend(r, RADIUSSTOP); } - if (r && radius[r].state != RADIUSSTOP) - radiussend(r, RADIUSSTOP); // stop, if not already trying + else + LOG(1, s, session[s].tunnel, "No free RADIUS sessions for Stop message\n"); + + // Save counters to dump to accounting file + if (*config->accounting_dir && shut_acct_n < sizeof(shut_acct) / sizeof(*shut_acct)) + memcpy(&shut_acct[shut_acct_n++], &session[s], sizeof(session[s])); } if (session[s].ip) @@ -1106,88 +1536,137 @@ void sessionshutdown(sessionidt s, char *reason) } else free_ip_address(s); + + // unroute IPv6, if setup + if (session[s].ppp.ipv6cp == Opened && session[s].ipv6prefixlen) + route6set(s, session[s].ipv6route, session[s].ipv6prefixlen, 0); } if (session[s].throttle_in || session[s].throttle_out) // Unthrottle if throttled. throttle_session(s, 0, 0); + if (result) { // Send CDN controlt *c = controlnew(14); // sending CDN - control16(c, 1, 3, 1); // result code (admin reasons - TBA make error, general error, add message + if (error) + { + uint8_t buf[4]; + *(uint16_t *) buf = htons(result); + *(uint16_t *) (buf+2) = htons(error); + controlb(c, 1, buf, 4, 1); + } + else + control16(c, 1, result, 1); + control16(c, 14, s, 1); // assigned session (our end) - controladd(c, session[s].tunnel, s); // send the message + controladd(c, session[s].far, session[s].tunnel); // send the message } if (!session[s].die) - session[s].die = now() + 150; // Clean up in 15 seconds + session[s].die = TIME + 150; // Clean up in 15 seconds // update filter refcounts if (session[s].filter_in) ip_filters[session[s].filter_in - 1].used--; if (session[s].filter_out) ip_filters[session[s].filter_out - 1].used--; + // clear PPP state + memset(&session[s].ppp, 0, sizeof(session[s].ppp)); + sess_local[s].lcp.restart = 0; + sess_local[s].ipcp.restart = 0; + sess_local[s].ipv6cp.restart = 0; + sess_local[s].ccp.restart = 0; + cluster_send_session(s); } -void sendipcp(tunnelidt t, sessionidt s) +void sendipcp(sessionidt s, tunnelidt t) { - u8 buf[MAXCONTROL]; - u16 r = session[s].radius; - u8 *q; - - CSTAT(call_sendipcp); - - if (!r) - r = radiusnew(s); + uint8_t buf[MAXETHER]; + uint8_t *q; - if (radius[r].state != RADIUSIPCP) - { - radius[r].state = RADIUSIPCP; - radius[r].try = 0; - } + CSTAT(sendipcp); + LOG(3, s, t, "IPCP: send ConfigReq\n"); - radius[r].retry = backoff(radius[r].try++); - if (radius[r].try > 10) + if (!session[s].unique_id) { - radiusclear(r, s); // Clear radius session. - sessionshutdown(s, "No reply on IPCP"); - return; + if (!++last_id) ++last_id; // skip zero + session[s].unique_id = last_id; } - q = makeppp(buf,sizeof(buf), 0, 0, t, s, PPPIPCP); + q = makeppp(buf, sizeof(buf), 0, 0, s, t, PPPIPCP); if (!q) return; *q = ConfigReq; - q[1] = r << RADIUS_SHIFT; // ID, dont care, we only send one type of request - *(u16 *) (q + 2) = htons(10); - q[4] = 3; - q[5] = 6; - *(u32 *) (q + 6) = config->peer_address ? config->peer_address : - config->bind_address ? config->bind_address : - my_address; // send my IP + q[1] = session[s].unique_id & 0xf; // ID, dont care, we only send one type of request + *(uint16_t *) (q + 2) = htons(10); // packet length + q[4] = 3; // ip address option + q[5] = 6; // option length + *(in_addr_t *) (q + 6) = config->peer_address ? config->peer_address : + config->bind_address ? config->bind_address : + my_address; // send my IP tunnelsend(buf, 10 + (q - buf), t); // send it - session[s].flags &= ~SF_IPCP_ACKED; // Clear flag. } -// kill a session now -static void sessionkill(sessionidt s, char *reason) +void sendipv6cp(sessionidt s, tunnelidt t) { + uint8_t buf[MAXETHER]; + uint8_t *q; - CSTAT(call_sessionkill); + CSTAT(sendipv6cp); + LOG(3, s, t, "IPV6CP: send ConfigReq\n"); - session[s].die = now(); - sessionshutdown(s, reason); // close radius/routes, etc. - if (session[s].radius) - radiusclear(session[s].radius, s); // cant send clean accounting data, session is killed + q = makeppp(buf, sizeof(buf), 0, 0, s, t, PPPIPV6CP); + if (!q) return; - LOG(2, s, session[s].tunnel, "Kill session %d (%s): %s\n", s, session[s].user, reason); + *q = ConfigReq; + q[1] = session[s].unique_id & 0xf; // ID, don't care, we + // only send one type + // of request + *(uint16_t *) (q + 2) = htons(14); + q[4] = 1; // interface identifier option + q[5] = 10; // option length + *(uint32_t *) (q + 6) = 0; // We'll be prefix::1 + *(uint32_t *) (q + 10) = 0; + q[13] = 1; + + tunnelsend(buf, 14 + (q - buf), t); // send it +} +static void sessionclear(sessionidt s) +{ memset(&session[s], 0, sizeof(session[s])); + memset(&sess_local[s], 0, sizeof(sess_local[s])); + memset(&cli_session_actions[s], 0, sizeof(cli_session_actions[s])); + session[s].tunnel = T_FREE; // Mark it as free. session[s].next = sessionfree; sessionfree = s; - cli_session_actions[s].action = 0; - cluster_send_session(s); +} + +// kill a session now +void sessionkill(sessionidt s, char *reason) +{ + + CSTAT(sessionkill); + + if (!session[s].opened) // not alive + return; + + if (session[s].next) + { + LOG(0, s, session[s].tunnel, "Tried to kill a session with next pointer set (%d)\n", session[s].next); + return; + } + + session[s].die = TIME; + sessionshutdown(s, reason, 3, 0); // close radius/routes, etc. + if (sess_local[s].radius) + radiusclear(sess_local[s].radius, s); // cant send clean accounting data, session is killed + + LOG(2, s, session[s].tunnel, "Kill session %d (%s): %s\n", s, session[s].user, reason); + sessionclear(s); + cluster_send_session(s); } static void tunnelclear(tunnelidt t) @@ -1203,7 +1682,7 @@ static void tunnelkill(tunnelidt t, char *reason) sessionidt s; controlt *c; - CSTAT(call_tunnelkill); + CSTAT(tunnelkill); tunnel[t].state = TUNNELDIE; @@ -1217,23 +1696,23 @@ static void tunnelkill(tunnelidt t, char *reason) controlfree = c; } // kill sessions - for (s = 1; s < MAXSESSION; s++) + for (s = 1; s <= config->cluster_highest_sessionid ; ++s) if (session[s].tunnel == t) sessionkill(s, reason); // free tunnel tunnelclear(t); LOG(1, 0, t, "Kill tunnel %d: %s\n", t, reason); - cli_tunnel_actions[s].action = 0; + cli_tunnel_actions[t].action = 0; cluster_send_tunnel(t); } // shut down a tunnel cleanly -static void tunnelshutdown(tunnelidt t, char *reason) +static void tunnelshutdown(tunnelidt t, char *reason, int result, int error, char *msg) { sessionidt s; - CSTAT(call_tunnelshutdown); + CSTAT(tunnelshutdown); if (!tunnel[t].last || !tunnel[t].far || tunnel[t].state == TUNNELFREE) { @@ -1244,31 +1723,52 @@ static void tunnelshutdown(tunnelidt t, char *reason) LOG(1, 0, t, "Shutting down tunnel %d (%s)\n", t, reason); // close session - for (s = 1; s < MAXSESSION; s++) + for (s = 1; s <= config->cluster_highest_sessionid ; ++s) if (session[s].tunnel == t) - sessionshutdown(s, reason); + sessionshutdown(s, reason, 0, 0); tunnel[t].state = TUNNELDIE; - tunnel[t].die = now() + 700; // Clean up in 70 seconds + tunnel[t].die = TIME + 700; // Clean up in 70 seconds cluster_send_tunnel(t); // TBA - should we wait for sessions to stop? - { // Send StopCCN - controlt *c = controlnew(4); // sending StopCCN - control16(c, 1, 1, 1); // result code (admin reasons - TBA make error, general error, add message - control16(c, 9, t, 1); // assigned tunnel (our end) - controladd(c, t, 0); // send the message + if (result) + { + controlt *c = controlnew(4); // sending StopCCN + if (error) + { + uint8_t buf[64]; + int l = 4; + *(uint16_t *) buf = htons(result); + *(uint16_t *) (buf+2) = htons(error); + if (msg) + { + int m = strlen(msg); + if (m + 4 > sizeof(buf)) + m = sizeof(buf) - 4; + + memcpy(buf+4, msg, m); + l += m; + } + + controlb(c, 1, buf, l, 1); + } + else + control16(c, 1, result, 1); + + control16(c, 9, t, 1); // assigned tunnel (our end) + controladd(c, 0, t); // send the message } } // read and process packet on tunnel (UDP) -void processudp(u8 * buf, int len, struct sockaddr_in *addr) +void processudp(uint8_t *buf, int len, struct sockaddr_in *addr) { - char *chapresponse = NULL; - u16 l = len, t = 0, s = 0, ns = 0, nr = 0; - u8 *p = buf + 2; + uint8_t *chapresponse = NULL; + uint16_t l = len, t = 0, s = 0, ns = 0, nr = 0; + uint8_t *p = buf + 2; - CSTAT(call_processudp); + CSTAT(processudp); udp_rx += len; udp_rx_pkt++; @@ -1289,12 +1789,12 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) } if (*buf & 0x40) { // length - l = ntohs(*(u16 *) p); + l = ntohs(*(uint16_t *) p); p += 2; } - t = ntohs(*(u16 *) p); + t = ntohs(*(uint16_t *) p); p += 2; - s = ntohs(*(u16 *) p); + s = ntohs(*(uint16_t *) p); p += 2; if (s >= MAXSESSION) { @@ -1310,36 +1810,34 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) } if (*buf & 0x08) { // ns/nr - ns = ntohs(*(u16 *) p); + ns = ntohs(*(uint16_t *) p); p += 2; - nr = ntohs(*(u16 *) p); + nr = ntohs(*(uint16_t *) p); p += 2; } if (*buf & 0x02) { // offset - u16 o = ntohs(*(u16 *) p); + uint16_t o = ntohs(*(uint16_t *) p); p += o + 2; } if ((p - buf) > l) { - LOG(1, s, t, "Bad length %d>%d\n", (p - buf), l); + LOG(1, s, t, "Bad length %d>%d\n", (int) (p - buf), l); STAT(tunnel_rx_errors); return; } l -= (p - buf); if (*buf & 0x80) { // control - u16 message = 0xFFFF; // message type - u8 fatal = 0; - u8 mandatorymessage = 0; - u8 chap = 0; // if CHAP being used - u16 asession = 0; // assigned session - u32 amagic = 0; // magic number - u8 aflags = 0; // flags from last LCF - u16 version = 0x0100; // protocol version (we handle 0.0 as well and send that back just in case) - int requestchap = 0; // do we request PAP instead of original CHAP request? - char called[MAXTEL] = ""; // called number - char calling[MAXTEL] = ""; // calling number + uint16_t message = 0xFFFF; // message type + uint8_t fatal = 0; + uint8_t mandatory = 0; + uint16_t asession = 0; // assigned session + uint32_t amagic = 0; // magic number + uint8_t aflags = 0; // flags from last LCF + uint16_t version = 0x0100; // protocol version (we handle 0.0 as well and send that back just in case) + char called[MAXTEL] = ""; // called number + char calling[MAXTEL] = ""; // calling number if (!config->cluster_iam_master) { @@ -1347,7 +1845,10 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) return; } - if ((*buf & 0xCA) != 0xC8) + // control messages must have bits 0x80|0x40|0x08 + // (type, length and sequence) set, and bits 0x02|0x01 + // (offset and priority) clear + if ((*buf & 0xCB) != 0xC8) { LOG(1, s, t, "Bad control header %02X\n", *buf); STAT(tunnel_rx_errors); @@ -1365,7 +1866,7 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) for (i = 1; i <= config->cluster_highest_tunnelid ; ++i) { if (tunnel[i].state != TUNNELOPENING || - tunnel[i].ip != ntohl(*(ipt *) & addr->sin_addr) || + tunnel[i].ip != ntohl(*(in_addr_t *) & addr->sin_addr) || tunnel[i].port != ntohs(addr->sin_port) ) continue; t = i; @@ -1387,7 +1888,7 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) return; } tunnelclear(t); - tunnel[t].ip = ntohl(*(ipt *) & addr->sin_addr); + tunnel[t].ip = ntohl(*(in_addr_t *) & addr->sin_addr); tunnel[t].port = ntohs(addr->sin_port); tunnel[t].window = 4; // default window STAT(tunnel_created); @@ -1457,95 +1958,130 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) } if (l) { // if not a null message + int result = 0; + int error = 0; + char *msg = 0; + // process AVPs - while (l && !(fatal & 0x80)) + while (l && !(fatal & 0x80)) // 0x80 = mandatory AVP { - u16 n = (ntohs(*(u16 *) p) & 0x3FF); - u8 *b = p; - u8 flags = *p; - u16 mtype; - p += n; // next - if (l < n) + uint16_t n = (ntohs(*(uint16_t *) p) & 0x3FF); + uint8_t *b = p; + uint8_t flags = *p; + uint16_t mtype; + if (n > l) { LOG(1, s, t, "Invalid length in AVP\n"); STAT(tunnel_rx_errors); - fatal = flags; return; } + p += n; // next l -= n; + if (flags & 0x3C) // reserved bits, should be clear + { + LOG(1, s, t, "Unrecognised AVP flags %02X\n", *b); + fatal = flags; + result = 2; // general error + error = 3; // reserved field non-zero + msg = 0; + continue; // next + } + b += 2; + if (*(uint16_t *) (b)) + { + LOG(2, s, t, "Unknown AVP vendor %d\n", ntohs(*(uint16_t *) (b))); + fatal = flags; + result = 2; // general error + error = 6; // generic vendor-specific error + msg = "unsupported vendor-specific"; + continue; // next + } + b += 2; + mtype = ntohs(*(uint16_t *) (b)); + b += 2; + n -= 6; + if (flags & 0x40) { + uint16_t orig_len; + // handle hidden AVPs if (!*config->l2tpsecret) { LOG(1, s, t, "Hidden AVP requested, but no L2TP secret.\n"); fatal = flags; + result = 2; // general error + error = 6; // generic vendor-specific error + msg = "secret not specified"; continue; } if (!session[s].random_vector_length) { LOG(1, s, t, "Hidden AVP requested, but no random vector.\n"); fatal = flags; + result = 2; // general error + error = 6; // generic + msg = "no random vector"; + continue; + } + if (n < 8) + { + LOG(2, s, t, "Short hidden AVP.\n"); + fatal = flags; + result = 2; // general error + error = 2; // length is wrong + msg = 0; continue; } - LOG(4, s, t, "Hidden AVP\n"); + // Unhide the AVP - n = unhide_avp(b, t, s, n); - if (n == 0) + unhide_value(b, n, mtype, session[s].random_vector, session[s].random_vector_length); + + orig_len = ntohs(*(uint16_t *) b); + if (orig_len > n + 2) { + LOG(1, s, t, "Original length %d too long in hidden AVP of length %d; wrong secret?\n", + orig_len, n); + fatal = flags; + result = 2; // general error + error = 2; // length is wrong + msg = 0; continue; } + + b += 2; + n = orig_len; } - if (*b & 0x3C) - { - LOG(1, s, t, "Unrecognised AVP flags %02X\n", *b); - fatal = flags; - continue; // next - } - b += 2; - if (*(u16 *) (b)) - { - LOG(2, s, t, "Unknown AVP vendor %d\n", ntohs(*(u16 *) (b))); - fatal = flags; - continue; // next - } - b += 2; - mtype = ntohs(*(u16 *) (b)); - b += 2; - n -= 6; - LOG(4, s, t, " AVP %d (%s) len %d\n", mtype, avpnames[mtype], n); + LOG(4, s, t, " AVP %d (%s) len %d%s%s\n", mtype, l2tp_avp_name(mtype), n, + flags & 0x40 ? ", hidden" : "", flags & 0x80 ? ", mandatory" : ""); + switch (mtype) { case 0: // message type - message = ntohs(*(u16 *) b); - LOG(4, s, t, " Message type = %d (%s)\n", *b, l2tp_message_types[message]); - mandatorymessage = flags; + message = ntohs(*(uint16_t *) b); + mandatory = flags & 0x80; + LOG(4, s, t, " Message type = %d (%s)\n", *b, l2tp_code(message)); break; case 1: // result code { - u16 rescode = ntohs(*(u16 *)(b)); + uint16_t rescode = ntohs(*(uint16_t *) b); const char* resdesc = "(unknown)"; if (message == 4) { /* StopCCN */ - if (rescode <= MAX_STOPCCN_RESULT_CODE) - resdesc = stopccn_result_codes[rescode]; + resdesc = l2tp_stopccn_result_code(rescode); } else if (message == 14) { /* CDN */ - if (rescode <= MAX_CDN_RESULT_CODE) - resdesc = cdn_result_codes[rescode]; + resdesc = l2tp_cdn_result_code(rescode); } LOG(4, s, t, " Result Code %d: %s\n", rescode, resdesc); if (n >= 4) { - u16 errcode = ntohs(*(u16 *)(b + 2)); - const char* errdesc = "(unknown)"; - if (errcode <= MAX_ERROR_CODE) - errdesc = error_codes[errcode]; - LOG(4, s, t, " Error Code %d: %s\n", errcode, errdesc); + uint16_t errcode = ntohs(*(uint16_t *)(b + 2)); + LOG(4, s, t, " Error Code %d: %s\n", errcode, l2tp_error_code(errcode)); } if (n > 4) LOG(4, s, t, " Error String: %.*s\n", n-4, b+4); @@ -1555,12 +2091,15 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) break; case 2: // protocol version { - version = ntohs(*(u16 *) (b)); + version = ntohs(*(uint16_t *) (b)); LOG(4, s, t, " Protocol version = %d\n", version); if (version && version != 0x0100) { // allow 0.0 and 1.0 LOG(1, s, t, " Bad protocol version %04X\n", version); fatal = flags; + result = 5; // unspported protocol version + error = 0x0100; // supported version + msg = 0; continue; // next } } @@ -1579,22 +2118,22 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) // LOG(4, s, t, "Firmware revision\n"); break; case 7: // host name - memset(tunnel[t].hostname, 0, 128); - memcpy(tunnel[t].hostname, b, (n >= 127) ? 127 : n); + memset(tunnel[t].hostname, 0, sizeof(tunnel[t].hostname)); + memcpy(tunnel[t].hostname, b, (n < sizeof(tunnel[t].hostname)) ? n : sizeof(tunnel[t].hostname) - 1); LOG(4, s, t, " Tunnel hostname = \"%s\"\n", tunnel[t].hostname); // TBA - to send to RADIUS break; case 8: // vendor name memset(tunnel[t].vendor, 0, sizeof(tunnel[t].vendor)); - memcpy(tunnel[t].vendor, b, (n >= sizeof(tunnel[t].vendor) - 1) ? sizeof(tunnel[t].vendor) - 1 : n); + memcpy(tunnel[t].vendor, b, (n < sizeof(tunnel[t].vendor)) ? n : sizeof(tunnel[t].vendor) - 1); LOG(4, s, t, " Vendor name = \"%s\"\n", tunnel[t].vendor); break; case 9: // assigned tunnel - tunnel[t].far = ntohs(*(u16 *) (b)); + tunnel[t].far = ntohs(*(uint16_t *) (b)); LOG(4, s, t, " Remote tunnel id = %d\n", tunnel[t].far); break; case 10: // rx window - tunnel[t].window = ntohs(*(u16 *) (b)); + tunnel[t].window = ntohs(*(uint16_t *) (b)); if (!tunnel[t].window) tunnel[t].window = 1; // window of 0 is silly LOG(4, s, t, " rx window = %d\n", tunnel[t].window); @@ -1611,28 +2150,28 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) break; case 14: // assigned session - asession = session[s].far = ntohs(*(u16 *) (b)); + asession = session[s].far = ntohs(*(uint16_t *) (b)); LOG(4, s, t, " assigned session = %d\n", asession); break; case 15: // call serial number - LOG(4, s, t, " call serial number = %d\n", ntohl(*(u32 *)b)); + LOG(4, s, t, " call serial number = %d\n", ntohl(*(uint32_t *)b)); break; case 18: // bearer type - LOG(4, s, t, " bearer type = %d\n", ntohl(*(u32 *)b)); + LOG(4, s, t, " bearer type = %d\n", ntohl(*(uint32_t *)b)); // TBA - for RADIUS break; case 19: // framing type - LOG(4, s, t, " framing type = %d\n", ntohl(*(u32 *)b)); + LOG(4, s, t, " framing type = %d\n", ntohl(*(uint32_t *)b)); // TBA break; case 21: // called number - memset(called, 0, MAXTEL); - memcpy(called, b, (n >= MAXTEL) ? (MAXTEL-1) : n); + memset(called, 0, sizeof(called)); + memcpy(called, b, (n < sizeof(called)) ? n : sizeof(called) - 1); LOG(4, s, t, " Called <%s>\n", called); break; case 22: // calling number - memset(calling, 0, MAXTEL); - memcpy(calling, b, (n >= MAXTEL) ? (MAXTEL-1) : n); + memset(calling, 0, sizeof(calling)); + memcpy(calling, b, (n < sizeof(calling)) ? n : sizeof(calling) - 1); LOG(4, s, t, " Calling <%s>\n", calling); break; case 23: // subtype @@ -1640,13 +2179,14 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) case 24: // tx connect speed if (n == 4) { - session[s].tx_connect_speed = ntohl(*(u32 *)b); + session[s].tx_connect_speed = ntohl(*(uint32_t *)b); } else { // AS5300s send connect speed as a string - char tmp[30] = {0}; - memcpy(tmp, b, (n >= 30) ? 30 : n); + char tmp[30]; + memset(tmp, 0, sizeof(tmp)); + memcpy(tmp, b, (n < sizeof(tmp)) ? n : sizeof(tmp) - 1); session[s].tx_connect_speed = atol(tmp); } LOG(4, s, t, " TX connect speed <%u>\n", session[s].tx_connect_speed); @@ -1654,68 +2194,60 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) case 38: // rx connect speed if (n == 4) { - session[s].rx_connect_speed = ntohl(*(u32 *)b); + session[s].rx_connect_speed = ntohl(*(uint32_t *)b); } else { // AS5300s send connect speed as a string - char tmp[30] = {0}; - memcpy(tmp, b, (n >= 30) ? 30 : n); + char tmp[30]; + memset(tmp, 0, sizeof(tmp)); + memcpy(tmp, b, (n < sizeof(tmp)) ? n : sizeof(tmp) - 1); session[s].rx_connect_speed = atol(tmp); } LOG(4, s, t, " RX connect speed <%u>\n", session[s].rx_connect_speed); break; case 25: // Physical Channel ID { - u32 tmp = ntohl(*(u32 *)b); + uint32_t tmp = ntohl(*(uint32_t *) b); LOG(4, s, t, " Physical Channel ID <%X>\n", tmp); break; } case 29: // Proxy Authentication Type { - u16 authtype = ntohs(*(u16 *)b); - LOG(4, s, t, " Proxy Auth Type %d (%s)\n", authtype, authtypes[authtype]); - requestchap = (authtype == 2); + uint16_t atype = ntohs(*(uint16_t *)b); + LOG(4, s, t, " Proxy Auth Type %d (%s)\n", atype, ppp_auth_type(atype)); break; } case 30: // Proxy Authentication Name { - char authname[64] = {0}; - memcpy(authname, b, (n > 63) ? 63 : n); + char authname[64]; + memset(authname, 0, sizeof(authname)); + memcpy(authname, b, (n < sizeof(authname)) ? n : sizeof(authname) - 1); LOG(4, s, t, " Proxy Auth Name (%s)\n", authname); break; } case 31: // Proxy Authentication Challenge { - memcpy(radius[session[s].radius].auth, b, 16); LOG(4, s, t, " Proxy Auth Challenge\n"); break; } case 32: // Proxy Authentication ID { - u16 authid = ntohs(*(u16 *)(b)); + uint16_t authid = ntohs(*(uint16_t *)(b)); LOG(4, s, t, " Proxy Auth ID (%d)\n", authid); - if (session[s].radius) - radius[session[s].radius].id = authid; break; } case 33: // Proxy Authentication Response - { - char authresp[64] = {0}; - memcpy(authresp, b, (n > 63) ? 63 : n); - LOG(4, s, t, " Proxy Auth Response\n"); - break; - } - case 27: // last send lcp + LOG(4, s, t, " Proxy Auth Response\n"); + break; + case 27: // last sent lcp { // find magic number - u8 *p = b, *e = p + n; + uint8_t *p = b, *e = p + n; while (p + 1 < e && p[1] && p + p[1] <= e) { if (*p == 5 && p[1] == 6) // Magic-Number - amagic = ntohl(*(u32 *) (p + 2)); - else if (*p == 3 && p[1] == 5 && *(u16 *) (p + 2) == htons(PPPCHAP) && p[4] == 5) // Authentication-Protocol - chap = 1; + amagic = ntohl(*(uint32_t *) (p + 2)); else if (*p == 7) // Protocol-Field-Compression aflags |= SESSIONPFC; else if (*p == 8) // Address-and-Control-Field-Compression @@ -1731,34 +2263,47 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) case 39: // seq required - we control it as an LNS anyway... break; case 36: // Random Vector - LOG(4, s, t, " Random Vector received. Enabled AVP Hiding.\n"); + LOG(4, s, t, " Random Vector received. Enabled AVP Hiding.\n"); memset(session[s].random_vector, 0, sizeof(session[s].random_vector)); + if (n > sizeof(session[s].random_vector)) + n = sizeof(session[s].random_vector); memcpy(session[s].random_vector, b, n); session[s].random_vector_length = n; break; default: - LOG(2, s, t, " Unknown AVP type %d\n", mtype); - fatal = flags; - continue; // next + { + static char e[] = "unknown AVP 0xXXXX"; + LOG(2, s, t, " Unknown AVP type %d\n", mtype); + fatal = flags; + result = 2; // general error + error = 8; // unknown mandatory AVP + sprintf((msg = e) + 14, "%04x", mtype); + continue; // next + } } } // process message if (fatal & 0x80) - tunnelshutdown(t, "Unknown Mandatory AVP"); + tunnelshutdown(t, "Invalid mandatory AVP", result, error, msg); else switch (message) { case 1: // SCCRQ - Start Control Connection Request + tunnel[t].state = TUNNELOPENING; + if (main_quit != QUIT_SHUTDOWN) { controlt *c = controlnew(2); // sending SCCRP control16(c, 2, version, 1); // protocol version control32(c, 3, 3, 1); // framing - controls(c, 7, tunnel[t].hostname, 1); // host name (TBA) + controls(c, 7, hostname, 1); // host name if (chapresponse) controlb(c, 13, chapresponse, 16, 1); // Challenge response control16(c, 9, t, 1); // assigned tunnel - controladd(c, t, s); // send the resply + controladd(c, 0, t); // send the resply + } + else + { + tunnelshutdown(t, "Shutting down", 6, 0, 0); } - tunnel[t].state = TUNNELOPENING; break; case 2: // SCCRP tunnel[t].state = TUNNELOPEN; @@ -1769,8 +2314,7 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) break; case 4: // StopCCN controlnull(t); // ack - tunnelshutdown(t, "Stopped"); // Shut down cleanly - tunnelkill(t, "Stopped"); // Immediately force everything dead + tunnelshutdown(t, "Stopped", 0, 0, 0); // Shut down cleanly break; case 6: // HELLO controlnull(t); // simply ACK @@ -1785,15 +2329,9 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) // TBA break; case 10: // ICRQ - if (!sessionfree) + if (sessionfree && main_quit != QUIT_SHUTDOWN) { - STAT(session_overflow); - tunnelshutdown(t, "No free sessions"); - } - else - { - u16 r; - controlt *c; + controlt *c = controlnew(11); // ICRP s = sessionfree; sessionfree = session[s].next; @@ -1802,35 +2340,38 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) if (s > config->cluster_highest_sessionid) config->cluster_highest_sessionid = s; - // make a RADIUS session - if (!(r = radiusnew(s))) - { - LOG(1, s, t, "No free RADIUS sessions for ICRQ\n"); - sessionkill(s, "no free RADIUS sesions"); - return; - } - - c = controlnew(11); // sending ICRP - session[s].id = sessionid++; - session[s].opened = time(NULL); + session[s].opened = time_now; session[s].tunnel = t; session[s].far = asession; session[s].last_packet = time_now; LOG(3, s, t, "New session (%d/%d)\n", tunnel[t].far, session[s].far); control16(c, 14, s, 1); // assigned session - controladd(c, t, s); // send the reply - { - // Generate a random challenge - int n; - for (n = 0; n < 15; n++) - radius[r].auth[n] = rand(); - } - strncpy(radius[r].calling, calling, sizeof(radius[r].calling) - 1); + controladd(c, asession, t); // send the reply + strncpy(session[s].called, called, sizeof(session[s].called) - 1); strncpy(session[s].calling, calling, sizeof(session[s].calling) - 1); + + session[s].ppp.phase = Establish; + session[s].ppp.lcp = Starting; + STAT(session_created); + break; } - break; + + { + controlt *c = controlnew(14); // CDN + if (!sessionfree) + { + STAT(session_overflow); + LOG(1, 0, t, "No free sessions\n"); + control16(c, 1, 4, 0); // temporary lack of resources + } + else + control16(c, 1, 2, 7); // shutting down, try another + + controladd(c, asession, t); // send the message + } + return; case 11: // ICRP // TBA break; @@ -1838,23 +2379,29 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) if (amagic == 0) amagic = time_now; session[s].magic = amagic; // set magic number session[s].l2tp_flags = aflags; // set flags received - LOG(3, s, t, "Magic %X Flags %X\n", amagic, aflags); + session[s].mru = DEFAULT_MRU; controlnull(t); // ack - // In CHAP state, request PAP instead - if (requestchap) - initlcp(t, s); + + // start LCP + sess_local[s].lcp.restart = time_now + config->ppp_restart_time; + sess_local[s].lcp.conf_sent = 1; + sess_local[s].lcp.nak_sent = 0; + sess_local[s].lcp_authtype = config->radius_authprefer; + session[s].ppp.lcp = RequestSent; + sendlcp(s, t); + break; case 14: // CDN controlnull(t); // ack - sessionshutdown(s, "Closed (Received CDN)"); + sessionshutdown(s, "Closed (Received CDN).", 0, 0); break; case 0xFFFF: LOG(1, s, t, "Missing message type\n"); break; default: STAT(tunnel_rx_errors); - if (mandatorymessage & 0x80) - tunnelshutdown(t, "Unknown message"); + if (mandatory) + tunnelshutdown(t, "Unknown message type", 2, 6, "unknown message type"); else LOG(1, s, t, "Unknown message type %d\n", message); break; @@ -1869,7 +2416,7 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) } else { // data - u16 prot; + uint16_t prot; LOG_HEX(5, "Receive Tunnel Data", p, l); if (l > 2 && p[0] == 0xFF && p[1] == 0x03) @@ -1890,12 +2437,12 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) } else { - prot = ntohs(*(u16 *) p); + prot = ntohs(*(uint16_t *) p); p += 2; l -= 2; } - if (s && !session[s].tunnel) // Is something wrong?? + if (s && !session[s].opened) // Is something wrong?? { if (!config->cluster_iam_master) { @@ -1905,58 +2452,85 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) } - LOG(1, s, t, "UDP packet contains session %d but no session[%d].tunnel " - "exists (LAC said tunnel = %d). Dropping packet.\n", s, s, t); - + LOG(1, s, t, "UDP packet contains session which is not opened. Dropping packet.\n"); STAT(tunnel_rx_errors); return; } - if (session[s].die) - { - LOG(3, s, t, "Session %d is closing. Don't process PPP packets\n", s); -// I'm pretty sure this isn't right -- mo. -// return; // closing session, PPP not processed - } if (prot == PPPPAP) { session[s].last_packet = time_now; if (!config->cluster_iam_master) { master_forward_packet(buf, len, addr->sin_addr.s_addr, addr->sin_port); return; } - processpap(t, s, p, l); + processpap(s, t, p, l); } else if (prot == PPPCHAP) { session[s].last_packet = time_now; if (!config->cluster_iam_master) { master_forward_packet(buf, len, addr->sin_addr.s_addr, addr->sin_port); return; } - processchap(t, s, p, l); + processchap(s, t, p, l); } else if (prot == PPPLCP) { session[s].last_packet = time_now; if (!config->cluster_iam_master) { master_forward_packet(buf, len, addr->sin_addr.s_addr, addr->sin_port); return; } - processlcp(t, s, p, l); + processlcp(s, t, p, l); } else if (prot == PPPIPCP) { session[s].last_packet = time_now; if (!config->cluster_iam_master) { master_forward_packet(buf, len, addr->sin_addr.s_addr, addr->sin_port); return; } - processipcp(t, s, p, l); + processipcp(s, t, p, l); + } + else if (prot == PPPIPV6CP) + { + session[s].last_packet = time_now; + if (!config->cluster_iam_master) { master_forward_packet(buf, len, addr->sin_addr.s_addr, addr->sin_port); return; } + processipv6cp(s, t, p, l); } else if (prot == PPPCCP) { session[s].last_packet = time_now; if (!config->cluster_iam_master) { master_forward_packet(buf, len, addr->sin_addr.s_addr, addr->sin_port); return; } - processccp(t, s, p, l); + processccp(s, t, p, l); } else if (prot == PPPIP) { + if (session[s].die) + { + LOG(4, s, t, "Session %d is closing. Don't process PPP packets\n", s); + return; // closing session, PPP not processed + } + + session[s].last_packet = time_now; + if (session[s].walled_garden && !config->cluster_iam_master) + { + master_forward_packet(buf, len, addr->sin_addr.s_addr, addr->sin_port); + return; + } + + processipin(s, t, p, l); + } + else if (prot == PPPIPV6) + { + if (!config->ipv6_prefix.s6_addr[0]) + { + LOG(1, s, t, "IPv6 not configured; yet received IPv6 packet. Ignoring.\n"); + return; + } + if (session[s].die) + { + LOG(4, s, t, "Session %d is closing. Don't process PPP packets\n", s); + return; // closing session, PPP not processed + } + session[s].last_packet = time_now; if (session[s].walled_garden && !config->cluster_iam_master) { master_forward_packet(buf, len, addr->sin_addr.s_addr, addr->sin_port); return; } - processipin(t, s, p, l); + + processipv6in(s, t, p, l); } else { @@ -1967,13 +2541,13 @@ void processudp(u8 * buf, int len, struct sockaddr_in *addr) } // read and process packet on tun -static void processtun(u8 * buf, int len) +static void processtun(uint8_t * buf, int len) { LOG_HEX(5, "Receive TUN Data", buf, len); STAT(tun_rx_packets); INC_STAT(tun_rx_bytes, len); - CSTAT(call_processtun); + CSTAT(processtun); eth_rx_pkt++; eth_rx += len; @@ -1984,46 +2558,70 @@ static void processtun(u8 * buf, int len) return; } - if (*(u16 *) (buf + 2) == htons(PKTIP)) // IP + if (*(uint16_t *) (buf + 2) == htons(PKTIP)) // IPv4 processipout(buf, len); + else if (*(uint16_t *) (buf + 2) == htons(PKTIPV6) // IPV6 + && config->ipv6_prefix.s6_addr[0]) + processipv6out(buf, len); + // Else discard. } -// -// Maximum number of actions to complete. -// This is to avoid sending out too many packets -// at once. -#define MAX_ACTIONS 500 - -static int regular_cleanups(void) +// Handle retries, timeouts. Runs every 1/10th sec, want to ensure +// that we look at the whole of the tunnel, radius and session tables +// every second +static void regular_cleanups(double period) { - static sessionidt s = 0; // Next session to check for actions on. - tunnelidt t; - int count=0,i; - u16 r; - static clockt next_acct = 0; + // Next tunnel, radius and session to check for actions on. + static tunnelidt t = 0; + static int r = 0; + static sessionidt s = 0; + + int t_actions = 0; + int r_actions = 0; + int s_actions = 0; + + int t_slice; + int r_slice; + int s_slice; + + int i; int a; - LOG(3, 0, 0, "Begin regular cleanup\n"); + // divide up tables into slices based on the last run + t_slice = config->cluster_highest_tunnelid * period; + r_slice = (MAXRADIUS - 1) * period; + s_slice = config->cluster_highest_sessionid * period; - for (r = 1; r < MAXRADIUS; r++) - { - if (!radius[r].state) - continue; - if (radius[r].retry) - { - if (radius[r].retry <= TIME) - radiusretry(r); - } else - radius[r].retry = backoff(radius[r].try+1); // Is this really needed? --mo - } - for (t = 1; t <= config->cluster_highest_tunnelid; t++) + if (t_slice < 1) + t_slice = 1; + else if (t_slice > config->cluster_highest_tunnelid) + t_slice = config->cluster_highest_tunnelid; + + if (r_slice < 1) + r_slice = 1; + else if (r_slice > (MAXRADIUS - 1)) + r_slice = MAXRADIUS - 1; + + if (s_slice < 1) + s_slice = 1; + else if (s_slice > config->cluster_highest_sessionid) + s_slice = config->cluster_highest_sessionid; + + LOG(4, 0, 0, "Begin regular cleanup (last %f seconds ago)\n", period); + + for (i = 0; i < t_slice; i++) { + t++; + if (t > config->cluster_highest_tunnelid) + t = 1; + // check for expired tunnels if (tunnel[t].die && tunnel[t].die <= TIME) { STAT(tunnel_timeout); tunnelkill(t, "Expired"); + t_actions++; continue; } // check for message resend @@ -2033,7 +2631,7 @@ static int regular_cleanups(void) if (tunnel[t].retry <= TIME) { controlt *c = tunnel[t].controls; - u8 w = tunnel[t].window; + uint8_t w = tunnel[t].window; tunnel[t].try++; // another try if (tunnel[t].try > 5) tunnelkill(t, "Timeout on control message"); // game over @@ -2043,14 +2641,17 @@ static int regular_cleanups(void) tunnelsend(c->buf, c->length, t); c = c->next; } + + t_actions++; } } // Send hello - if (tunnel[t].state == TUNNELOPEN && tunnel[t].lastrec < TIME + 600) + if (tunnel[t].state == TUNNELOPEN && (time_now - tunnel[t].lastrec) > 60) { controlt *c = controlnew(6); // sending HELLO - controladd(c, t, 0); // send the message + controladd(c, 0, t); // send the message LOG(3, 0, t, "Sending HELLO message\n"); + t_actions++; } // Check for tunnel changes requested from the CLI @@ -2060,63 +2661,193 @@ static int regular_cleanups(void) if (a & CLI_TUN_KILL) { LOG(2, 0, t, "Dropping tunnel by CLI\n"); - tunnelshutdown(t, "Requested by administrator"); + tunnelshutdown(t, "Requested by administrator", 1, 0, 0); + t_actions++; } } + } + + for (i = 0; i < r_slice; i++) + { + r++; + if (r >= MAXRADIUS) + r = 1; + + if (!radius[r].state) + continue; + if (radius[r].retry <= TIME) + { + radiusretry(r); + r_actions++; + } } - count = 0; - for (i = 1; i <= config->cluster_highest_sessionid; i++) + for (i = 0; i < s_slice; i++) { s++; if (s > config->cluster_highest_sessionid) s = 1; - if (!session[s].tunnel) // Session isn't in use + if (!session[s].opened) // Session isn't in use continue; - if (!session[s].die && session[s].ip && !(session[s].flags & SF_IPCP_ACKED)) + // check for expired sessions + if (session[s].die) { - // IPCP has not completed yet. Resend - LOG(3, s, session[s].tunnel, "No ACK for initial IPCP ConfigReq... resending\n"); - sendipcp(session[s].tunnel, s); + if (session[s].die <= TIME) + { + sessionkill(s, "Expired"); + s_actions++; + } + continue; } - // check for expired sessions - if (session[s].die && session[s].die <= TIME) + // PPP timeouts + if (sess_local[s].lcp.restart <= time_now) { - sessionkill(s, "Expired"); - if (++count >= MAX_ACTIONS) break; - continue; + int next_state = session[s].ppp.lcp; + switch (session[s].ppp.lcp) + { + case RequestSent: + case AckReceived: + next_state = RequestSent; + + case AckSent: + if (sess_local[s].lcp.conf_sent < config->ppp_max_configure) + { + LOG(3, s, session[s].tunnel, "No ACK for LCP ConfigReq... resending\n"); + sess_local[s].lcp.restart = time_now + config->ppp_restart_time; + sess_local[s].lcp.conf_sent++; + sendlcp(s, session[s].tunnel); + change_state(s, lcp, next_state); + } + else + { + sessionshutdown(s, "No response to LCP ConfigReq.", 3, 0); + STAT(session_timeout); + } + + s_actions++; + } + + if (session[s].die) + continue; + } + + if (sess_local[s].ipcp.restart <= time_now) + { + int next_state = session[s].ppp.ipcp; + switch (session[s].ppp.ipcp) + { + case RequestSent: + case AckReceived: + next_state = RequestSent; + + case AckSent: + if (sess_local[s].ipcp.conf_sent < config->ppp_max_configure) + { + LOG(3, s, session[s].tunnel, "No ACK for IPCP ConfigReq... resending\n"); + sess_local[s].ipcp.restart = time_now + config->ppp_restart_time; + sess_local[s].ipcp.conf_sent++; + sendipcp(s, session[s].tunnel); + change_state(s, ipcp, next_state); + } + else + { + sessionshutdown(s, "No response to IPCP ConfigReq.", 3, 0); + STAT(session_timeout); + } + + s_actions++; + } + + if (session[s].die) + continue; + } + + if (sess_local[s].ipv6cp.restart <= time_now) + { + int next_state = session[s].ppp.ipv6cp; + switch (session[s].ppp.ipv6cp) + { + case RequestSent: + case AckReceived: + next_state = RequestSent; + + case AckSent: + if (sess_local[s].ipv6cp.conf_sent < config->ppp_max_configure) + { + LOG(3, s, session[s].tunnel, "No ACK for IPV6CP ConfigReq... resending\n"); + sess_local[s].ipv6cp.restart = time_now + config->ppp_restart_time; + sess_local[s].ipv6cp.conf_sent++; + sendipv6cp(s, session[s].tunnel); + change_state(s, ipv6cp, next_state); + } + else + { + LOG(3, s, session[s].tunnel, "No ACK for IPV6CP ConfigReq\n"); + change_state(s, ipv6cp, Stopped); + } + + s_actions++; + } + } + + if (sess_local[s].ccp.restart <= time_now) + { + int next_state = session[s].ppp.ccp; + switch (session[s].ppp.ccp) + { + case RequestSent: + case AckReceived: + next_state = RequestSent; + + case AckSent: + if (sess_local[s].ccp.conf_sent < config->ppp_max_configure) + { + LOG(3, s, session[s].tunnel, "No ACK for CCP ConfigReq... resending\n"); + sess_local[s].ccp.restart = time_now + config->ppp_restart_time; + sess_local[s].ccp.conf_sent++; + sendccp(s, session[s].tunnel); + change_state(s, ccp, next_state); + } + else + { + LOG(3, s, session[s].tunnel, "No ACK for CCP ConfigReq\n"); + change_state(s, ccp, Stopped); + } + + s_actions++; + } } // Drop sessions who have not responded within IDLE_TIMEOUT seconds if (session[s].last_packet && (time_now - session[s].last_packet >= IDLE_TIMEOUT)) { - sessionkill(s, "No response to LCP ECHO requests"); + sessionshutdown(s, "No response to LCP ECHO requests.", 3, 0); STAT(session_timeout); - if (++count >= MAX_ACTIONS) break; + s_actions++; continue; } - // No data in IDLE_TIMEOUT seconds, send LCP ECHO - if (session[s].user[0] && (time_now - session[s].last_packet >= ECHO_TIMEOUT)) + // No data in ECHO_TIMEOUT seconds, send LCP ECHO + if (session[s].ppp.phase >= Establish && (time_now - session[s].last_packet >= ECHO_TIMEOUT)) { - u8 b[MAXCONTROL] = {0}; + uint8_t b[MAXETHER]; - u8 *q = makeppp(b, sizeof(b), 0, 0, session[s].tunnel, s, PPPLCP); + uint8_t *q = makeppp(b, sizeof(b), 0, 0, s, session[s].tunnel, PPPLCP); if (!q) continue; *q = EchoReq; - *(u8 *)(q + 1) = (time_now % 255); // ID - *(u16 *)(q + 2) = htons(8); // Length - *(u32 *)(q + 4) = 0; // Magic Number (not supported) + *(uint8_t *)(q + 1) = (time_now % 255); // ID + *(uint16_t *)(q + 2) = htons(8); // Length + *(uint32_t *)(q + 4) = 0; // Magic Number (not supported) LOG(4, s, session[s].tunnel, "No data in %d seconds, sending LCP ECHO\n", (int)(time_now - session[s].last_packet)); tunnelsend(b, 24, session[s].tunnel); // send it - if (++count >= MAX_ACTIONS) break; + s_actions++; } // Check for actions requested from the CLI @@ -2128,8 +2859,9 @@ static int regular_cleanups(void) if (a & CLI_SESS_KILL) { LOG(2, s, session[s].tunnel, "Dropping session by CLI\n"); - sessionshutdown(s, "Requested by administrator"); + sessionshutdown(s, "Requested by administrator.", 3, 0); a = 0; // dead, no need to check for other actions + s_actions++; } if (a & CLI_SESS_NOSNOOP) @@ -2137,6 +2869,7 @@ static int regular_cleanups(void) LOG(2, s, session[s].tunnel, "Unsnooping session by CLI\n"); session[s].snoop_ip = 0; session[s].snoop_port = 0; + s_actions++; send++; } else if (a & CLI_SESS_SNOOP) @@ -2147,6 +2880,7 @@ static int regular_cleanups(void) session[s].snoop_ip = cli_session_actions[s].snoop_ip; session[s].snoop_port = cli_session_actions[s].snoop_port; + s_actions++; send++; } @@ -2154,6 +2888,7 @@ static int regular_cleanups(void) { LOG(2, s, session[s].tunnel, "Un-throttling session by CLI\n"); throttle_session(s, 0, 0); + s_actions++; send++; } else if (a & CLI_SESS_THROTTLE) @@ -2163,6 +2898,7 @@ static int regular_cleanups(void) cli_session_actions[s].throttle_out); throttle_session(s, cli_session_actions[s].throttle_in, cli_session_actions[s].throttle_out); + s_actions++; send++; } @@ -2170,6 +2906,7 @@ static int regular_cleanups(void) { LOG(2, s, session[s].tunnel, "Un-filtering session by CLI\n"); filter_session(s, 0, 0); + s_actions++; send++; } else if (a & CLI_SESS_FILTER) @@ -2179,30 +2916,40 @@ static int regular_cleanups(void) cli_session_actions[s].filter_out); filter_session(s, cli_session_actions[s].filter_in, cli_session_actions[s].filter_out); + s_actions++; send++; } if (send) cluster_send_session(s); - - if (++count >= MAX_ACTIONS) break; } - } - if (*config->accounting_dir && next_acct <= TIME) - { - // Dump accounting data - next_acct = TIME + ACCT_TIME; - dump_acct_info(); - } + // RADIUS interim accounting + if (config->radius_accounting && config->radius_interim > 0 + && session[s].ip && !session[s].walled_garden + && !sess_local[s].radius // RADIUS already in progress + && time_now - sess_local[s].last_interim >= config->radius_interim) + { + int rad = radiusnew(s); + if (!rad) + { + LOG(1, s, session[s].tunnel, "No free RADIUS sessions for Interim message\n"); + STAT(radius_overflow); + continue; + } - if (count >= MAX_ACTIONS) - return 1; // Didn't finish! + LOG(3, s, session[s].tunnel, "Sending RADIUS Interim for %s (%u)\n", + session[s].user, session[s].unique_id); - LOG(3, 0, 0, "End regular cleanup (%d actions), next in %d seconds\n", count, config->cleanup_interval); - return 0; -} + radiussend(rad, RADIUSINTERIM); + sess_local[s].last_interim = time_now; + s_actions++; + } + } + LOG(4, 0, 0, "End regular cleanup: checked %d/%d/%d tunnels/radius/sessions; %d/%d/%d actions\n", + t_slice, r_slice, s_slice, t_actions, r_actions, s_actions); +} // // Are we in the middle of a tunnel update, or radius @@ -2213,6 +2960,53 @@ static int still_busy(void) int i; static clockt last_talked = 0; static clockt start_busy_wait = 0; + + if (!config->cluster_iam_master) + { +#ifdef BGP + static time_t stopped_bgp = 0; + if (bgp_configured) + { + if (!stopped_bgp) + { + LOG(1, 0, 0, "Shutting down in %d seconds, stopping BGP...\n", QUIT_DELAY); + + for (i = 0; i < BGP_NUM_PEERS; i++) + if (bgp_peers[i].state == Established) + bgp_stop(&bgp_peers[i]); + + stopped_bgp = time_now; + + // we don't want to become master + cluster_send_ping(0); + + return 1; + } + + if (time_now < (stopped_bgp + QUIT_DELAY)) + return 1; + } +#endif /* BGP */ + + return 0; + } + + if (main_quit == QUIT_SHUTDOWN) + { + static int dropped = 0; + if (!dropped) + { + int i; + + LOG(1, 0, 0, "Dropping sessions and tunnels\n"); + for (i = 1; i < MAXTUNNEL; i++) + if (tunnel[i].ip || tunnel[i].state) + tunnelshutdown(i, "L2TPNS Closing", 6, 0, 0); + + dropped = 1; + } + } + if (start_busy_wait == 0) start_busy_wait = TIME; @@ -2232,7 +3026,7 @@ static int still_busy(void) // We stop waiting for radius after BUSY_WAIT_TIME 1/10th seconds if (abs(TIME - start_busy_wait) > BUSY_WAIT_TIME) { - LOG(1, 0, 0, "Giving up waiting for RADIUS to be empty. Shutting down anyway.\n"); + LOG(1, 0, 0, "Giving up waiting for RADIUS to be empty. Shutting down anyway.\n"); return 0; } @@ -2254,78 +3048,108 @@ static int still_busy(void) return 0; } -static fd_set readset; -static int readset_n = 0; +#ifdef HAVE_EPOLL +# include +#else +# define FAKE_EPOLL_IMPLEMENTATION /* include the functions */ +# include "fake_epoll.h" +#endif + +// the base set of fds polled: cli, cluster, tun, udp, control, dae +#define BASE_FDS 6 + +// additional polled fds +#ifdef BGP +# define EXTRA_FDS BGP_NUM_PEERS +#else +# define EXTRA_FDS 0 +#endif // main loop - gets packets on tun or udp and processes them static void mainloop(void) { int i; - u8 buf[65536]; - struct timeval to; + uint8_t buf[65536]; clockt next_cluster_ping = 0; // send initial ping immediately - time_t next_clean = time_now + config->cleanup_interval; - - LOG(4, 0, 0, "Beginning of main loop. udpfd=%d, tunfd=%d, cluster_sockfd=%d, controlfd=%d\n", - udpfd, tunfd, cluster_sockfd, controlfd); - - FD_ZERO(&readset); - FD_SET(udpfd, &readset); - FD_SET(tunfd, &readset); - FD_SET(controlfd, &readset); - FD_SET(clifd, &readset); - if (cluster_sockfd) FD_SET(cluster_sockfd, &readset); - readset_n = udpfd; - if (tunfd > readset_n) readset_n = tunfd; - if (controlfd > readset_n) readset_n = controlfd; - if (clifd > readset_n) readset_n = clifd; - if (cluster_sockfd > readset_n) readset_n = cluster_sockfd; + struct epoll_event events[BASE_FDS + RADIUS_FDS + EXTRA_FDS]; + int maxevent = sizeof(events)/sizeof(*events); - while (!main_quit || still_busy()) + if ((epollfd = epoll_create(maxevent)) < 0) { - fd_set r; - int n = readset_n; + LOG(0, 0, 0, "epoll_create failed: %s\n", strerror(errno)); + exit(1); + } + + LOG(4, 0, 0, "Beginning of main loop. clifd=%d, cluster_sockfd=%d, tunfd=%d, udpfd=%d, controlfd=%d, daefd=%d\n", + clifd, cluster_sockfd, tunfd, udpfd, controlfd, daefd); + + /* setup our fds to poll for input */ + { + static struct event_data d[BASE_FDS]; + struct epoll_event e; + + e.events = EPOLLIN; + i = 0; + + d[i].type = FD_TYPE_CLI; + e.data.ptr = &d[i++]; + epoll_ctl(epollfd, EPOLL_CTL_ADD, clifd, &e); + + d[i].type = FD_TYPE_CLUSTER; + e.data.ptr = &d[i++]; + epoll_ctl(epollfd, EPOLL_CTL_ADD, cluster_sockfd, &e); + + d[i].type = FD_TYPE_TUN; + e.data.ptr = &d[i++]; + epoll_ctl(epollfd, EPOLL_CTL_ADD, tunfd, &e); + + d[i].type = FD_TYPE_UDP; + e.data.ptr = &d[i++]; + epoll_ctl(epollfd, EPOLL_CTL_ADD, udpfd, &e); + + d[i].type = FD_TYPE_CONTROL; + e.data.ptr = &d[i++]; + epoll_ctl(epollfd, EPOLL_CTL_ADD, controlfd, &e); + + d[i].type = FD_TYPE_DAE; + e.data.ptr = &d[i++]; + epoll_ctl(epollfd, EPOLL_CTL_ADD, daefd, &e); + } + #ifdef BGP - fd_set w; - int bgp_set[BGP_NUM_PEERS]; + signal(SIGPIPE, SIG_IGN); + bgp_setup(config->as_number); + if (config->bind_address) + bgp_add_route(config->bind_address, 0xffffffff); + + for (i = 0; i < BGP_NUM_PEERS; i++) + { + if (config->neighbour[i].name[0]) + bgp_start(&bgp_peers[i], config->neighbour[i].name, + config->neighbour[i].as, config->neighbour[i].keepalive, + config->neighbour[i].hold, 0); /* 0 = routing disabled */ + } #endif /* BGP */ + while (!main_quit || still_busy()) + { + int more = 0; + int n; + if (config->reload_config) { // Update the config state based on config settings update_config(); } - memcpy(&r, &readset, sizeof(fd_set)); - to.tv_sec = 0; - to.tv_usec = 100000; // 1/10th of a second. - #ifdef BGP - FD_ZERO(&w); - for (i = 0; i < BGP_NUM_PEERS; i++) - { - bgp_set[i] = bgp_select_state(&bgp_peers[i]); - if (bgp_set[i] & 1) - { - FD_SET(bgp_peers[i].sock, &r); - if (bgp_peers[i].sock > n) - n = bgp_peers[i].sock; - } - - if (bgp_set[i] & 2) - { - FD_SET(bgp_peers[i].sock, &w); - if (bgp_peers[i].sock > n) - n = bgp_peers[i].sock; - } - } - - n = select(n + 1, &r, &w, 0, &to); -#else /* BGP */ - n = select(n + 1, &r, 0, 0, &to); + bgp_set_poll(); #endif /* BGP */ - TIME = now(); + n = epoll_wait(epollfd, events, maxevent, 100); // timeout 100ms (1/10th sec) + STAT(select_called); + + TIME = now(NULL); if (n < 0) { if (errno == EINTR || @@ -2333,72 +3157,153 @@ static void mainloop(void) continue; LOG(0, 0, 0, "Error returned from select(): %s\n", strerror(errno)); - main_quit++; - break; + break; // exit } - else if (n) + + if (n) { struct sockaddr_in addr; - int alen = sizeof(addr); - if (FD_ISSET(udpfd, &r)) + socklen_t alen; + int c, s; + int udp_ready = 0; + int tun_ready = 0; + int cluster_ready = 0; + int udp_pkts = 0; + int tun_pkts = 0; + int cluster_pkts = 0; +#ifdef BGP + uint32_t bgp_events[BGP_NUM_PEERS]; + memset(bgp_events, 0, sizeof(bgp_events)); +#endif /* BGP */ + + for (c = n, i = 0; i < c; i++) { - int c, n; - for (c = 0; c < config->multi_read_count; c++) + struct event_data *d = events[i].data.ptr; + switch (d->type) { - if ((n = recvfrom(udpfd, buf, sizeof(buf), 0, (void *) &addr, &alen)) > 0) - processudp(buf, n, &addr); - else - break; - } - } - if (FD_ISSET(tunfd, &r)) - { - int c, n; - for (c = 0; c < config->multi_read_count; c++) + case FD_TYPE_CLI: // CLI connections { - if ((n = read(tunfd, buf, sizeof(buf))) > 0) - processtun(buf, n); + int cli; + + alen = sizeof(addr); + if ((cli = accept(clifd, (struct sockaddr *)&addr, &alen)) >= 0) + { + cli_do(cli); + close(cli); + } else - break; + LOG(0, 0, 0, "accept error: %s\n", strerror(errno)); + + n--; + break; } - } - if (config->cluster_iam_master) - for (i = 0; i < config->num_radfds; i++) - if (FD_ISSET(radfds[i], &r)) - processrad(buf, recv(radfds[i], buf, sizeof(buf), 0), i); + // these are handled below, with multiple interleaved reads + case FD_TYPE_CLUSTER: cluster_ready++; break; + case FD_TYPE_TUN: tun_ready++; break; + case FD_TYPE_UDP: udp_ready++; break; - if (FD_ISSET(cluster_sockfd, &r)) - { - int size; - size = recvfrom(cluster_sockfd, buf, sizeof(buf), MSG_WAITALL, (void *) &addr, &alen); - processcluster(buf, size, addr.sin_addr.s_addr); + case FD_TYPE_CONTROL: // nsctl commands + alen = sizeof(addr); + processcontrol(buf, recvfrom(controlfd, buf, sizeof(buf), MSG_WAITALL, (void *) &addr, &alen), &addr, alen); + n--; + break; + + case FD_TYPE_DAE: // DAE requests + alen = sizeof(addr); + processdae(buf, recvfrom(daefd, buf, sizeof(buf), MSG_WAITALL, (void *) &addr, &alen), &addr, alen); + n--; + break; + + case FD_TYPE_RADIUS: // RADIUS response + s = recv(radfds[d->index], buf, sizeof(buf), 0); + if (s >= 0 && config->cluster_iam_master) + processrad(buf, s, d->index); + + n--; + break; + +#ifdef BGP + case FD_TYPE_BGP: + bgp_events[d->index] = events[i].events; + n--; + break; +#endif /* BGP */ + + default: + LOG(0, 0, 0, "Unexpected fd type returned from epoll_wait: %d\n", d->type); + } } - if (FD_ISSET(controlfd, &r)) - processcontrol(buf, recvfrom(controlfd, buf, sizeof(buf), MSG_WAITALL, (void *) &addr, &alen), &addr, alen); +#ifdef BGP + bgp_process(bgp_events); +#endif /* BGP */ - if (FD_ISSET(clifd, &r)) + for (c = 0; n && c < config->multi_read_count; c++) { - struct sockaddr_in addr; - int sockfd; - int len = sizeof(addr); + // L2TP + if (udp_ready) + { + alen = sizeof(addr); + if ((s = recvfrom(udpfd, buf, sizeof(buf), 0, (void *) &addr, &alen)) > 0) + { + processudp(buf, s, &addr); + udp_pkts++; + } + else + { + udp_ready = 0; + n--; + } + } - if ((sockfd = accept(clifd, (struct sockaddr *)&addr, &len)) <= 0) + // incoming IP + if (tun_ready) { - LOG(0, 0, 0, "accept error: %s\n", strerror(errno)); - continue; + if ((s = read(tunfd, buf, sizeof(buf))) > 0) + { + processtun(buf, s); + tun_pkts++; + } + else + { + tun_ready = 0; + n--; + } } - else + + // cluster + if (cluster_ready) { - cli_do(sockfd); - close(sockfd); + alen = sizeof(addr); + if ((s = recvfrom(cluster_sockfd, buf, sizeof(buf), MSG_WAITALL, (void *) &addr, &alen)) > 0) + { + processcluster(buf, s, addr.sin_addr.s_addr); + cluster_pkts++; + } + else + { + cluster_ready = 0; + n--; + } } } + + if (udp_pkts > 1 || tun_pkts > 1 || cluster_pkts > 1) + STAT(multi_read_used); + + if (c >= config->multi_read_count) + { + LOG(3, 0, 0, "Reached multi_read_count (%d); processed %d udp, %d tun and %d cluster packets\n", + config->multi_read_count, udp_pkts, tun_pkts, cluster_pkts); + + STAT(multi_read_exceeded); + more++; + } } // Runs on every machine (master and slaves). - if (cluster_sockfd && next_cluster_ping <= TIME) + if (next_cluster_ping <= TIME) { // Check to see which of the cluster is still alive.. @@ -2416,9 +3321,11 @@ static void mainloop(void) next_cluster_ping = TIME + config->cluster_hb_interval; } + if (!config->cluster_iam_master) + continue; + // Run token bucket filtering queue.. // Only run it every 1/10th of a second. - // Runs on all machines both master and slave. { static clockt last_run = 0; if (last_run != TIME) @@ -2428,31 +3335,44 @@ static void mainloop(void) } } - /* Handle timeouts. Make sure that this gets run anyway, even if there was - * something to read, else under load this will never actually run.... - * - */ - if (config->cluster_iam_master && next_clean <= time_now) + // Handle timeouts, retries etc. { - if (regular_cleanups()) - { - // Did it finish? - next_clean = time_now + 1 ; // Didn't finish. Check quickly. - } - else + static double last_clean = 0; + double this_clean; + double diff; + + TIME = now(&this_clean); + diff = this_clean - last_clean; + + // Run during idle time (after we've handled + // all incoming packets) or every 1/10th sec + if (!more || diff > 0.1) { - next_clean = time_now + config->cleanup_interval; // Did. Move to next interval. + regular_cleanups(diff); + last_clean = this_clean; } } -#ifdef BGP - for (i = 0; i < BGP_NUM_PEERS; i++) + if (*config->accounting_dir) { - bgp_process(&bgp_peers[i], - bgp_set[i] ? FD_ISSET(bgp_peers[i].sock, &r) : 0, - bgp_set[i] ? FD_ISSET(bgp_peers[i].sock, &w) : 0); + static clockt next_acct = 0; + static clockt next_shut_acct = 0; + + if (next_acct <= TIME) + { + // Dump accounting data + next_acct = TIME + ACCT_TIME; + next_shut_acct = TIME + ACCT_SHUT_TIME; + dump_acct_info(1); + } + else if (next_shut_acct <= TIME) + { + // Dump accounting data for shutdown sessions + next_shut_acct = TIME + ACCT_SHUT_TIME; + if (shut_acct_n) + dump_acct_info(0); + } } -#endif /* BGP */ } // Are we the master and shutting down?? @@ -2465,6 +3385,7 @@ static void mainloop(void) // // Important!!! We MUST not process any packets past this point! + LOG(1, 0, 0, "Shutdown complete\n"); } static void stripdomain(char *host) @@ -2537,23 +3458,41 @@ static void initdata(int optdebug, char *optconfig) { int i; - if (!(_statistics = shared_malloc(sizeof(struct Tstats)))) - { - LOG(0, 0, 0, "Error doing malloc for _statistics: %s\n", strerror(errno)); - exit(1); - } if (!(config = shared_malloc(sizeof(configt)))) { - LOG(0, 0, 0, "Error doing malloc for configuration: %s\n", strerror(errno)); + fprintf(stderr, "Error doing malloc for configuration: %s\n", strerror(errno)); exit(1); } + memset(config, 0, sizeof(configt)); time(&config->start_time); strncpy(config->config_file, optconfig, strlen(optconfig)); config->debug = optdebug; config->num_tbfs = MAXTBFS; config->rl_rate = 28; // 28kbps + config->cluster_mcast_ttl = 1; + config->cluster_master_min_adv = 1; + config->ppp_restart_time = 3; + config->ppp_max_configure = 10; + config->ppp_max_failure = 5; + strcpy(config->random_device, RANDOMDEVICE); + + log_stream = stderr; + +#ifdef RINGBUFFER + if (!(ringbuffer = shared_malloc(sizeof(struct Tringbuffer)))) + { + LOG(0, 0, 0, "Error doing malloc for ringbuffer: %s\n", strerror(errno)); + exit(1); + } + memset(ringbuffer, 0, sizeof(struct Tringbuffer)); +#endif + if (!(_statistics = shared_malloc(sizeof(struct Tstats)))) + { + LOG(0, 0, 0, "Error doing malloc for _statistics: %s\n", strerror(errno)); + exit(1); + } if (!(tunnel = shared_malloc(sizeof(tunnelt) * MAXTUNNEL))) { LOG(0, 0, 0, "Error doing malloc for tunnels: %s\n", strerror(errno)); @@ -2565,9 +3504,9 @@ static void initdata(int optdebug, char *optconfig) exit(1); } - if (!(sess_count = shared_malloc(sizeof(sessioncountt) * MAXSESSION))) + if (!(sess_local = shared_malloc(sizeof(sessionlocalt) * MAXSESSION))) { - LOG(0, 0, 0, "Error doing malloc for sessions_count: %s\n", strerror(errno)); + LOG(0, 0, 0, "Error doing malloc for sess_local: %s\n", strerror(errno)); exit(1); } @@ -2583,21 +3522,12 @@ static void initdata(int optdebug, char *optconfig) exit(1); } -if (!(ip_filters = shared_malloc(sizeof(ip_filtert) * MAXFILTER))) -{ - LOG(0, 0, 0, "Error doing malloc for ip_filters: %s\n", strerror(errno)); - exit(1); -} -memset(ip_filters, 0, sizeof(ip_filtert) * MAXFILTER); - -#ifdef RINGBUFFER - if (!(ringbuffer = shared_malloc(sizeof(struct Tringbuffer)))) + if (!(ip_filters = shared_malloc(sizeof(ip_filtert) * MAXFILTER))) { - LOG(0, 0, 0, "Error doing malloc for ringbuffer: %s\n", strerror(errno)); + LOG(0, 0, 0, "Error doing malloc for ip_filters: %s\n", strerror(errno)); exit(1); } - memset(ringbuffer, 0, sizeof(struct Tringbuffer)); -#endif + memset(ip_filters, 0, sizeof(ip_filtert) * MAXFILTER); if (!(cli_session_actions = shared_malloc(sizeof(struct cli_session_actions) * MAXSESSION))) { @@ -2619,7 +3549,7 @@ memset(ip_filters, 0, sizeof(ip_filtert) * MAXFILTER); memset(ip_address_pool, 0, sizeof(ippoolt) * MAXIPPOOL); // Put all the sessions on the free list marked as undefined. - for (i = 1; i < MAXSESSION - 1; i++) + for (i = 1; i < MAXSESSION; i++) { session[i].next = i + 1; session[i].tunnel = T_UNDEF; // mark it as not filled in. @@ -2628,7 +3558,7 @@ memset(ip_filters, 0, sizeof(ip_filtert) * MAXFILTER); sessionfree = 1; // Mark all the tunnels as undefined (waiting to be filled in by a download). - for (i = 1; i < MAXTUNNEL- 1; i++) + for (i = 1; i < MAXTUNNEL; i++) tunnel[i].state = TUNNELUNDEF; // mark it as not filled in. if (!*hostname) @@ -2651,14 +3581,14 @@ memset(ip_filters, 0, sizeof(ip_filtert) * MAXFILTER); static int assign_ip_address(sessionidt s) { - u32 i; + uint32_t i; int best = -1; time_t best_time = time_now; char *u = session[s].user; char reuse = 0; - CSTAT(call_assign_ip_address); + CSTAT(assign_ip_address); for (i = 1; i < ip_pool_size; i++) { @@ -2711,7 +3641,7 @@ static void free_ip_address(sessionidt s) int i = session[s].ip_pool_index; - CSTAT(call_free_ip_address); + CSTAT(free_ip_address); if (!session[s].ip) return; // what the? @@ -2754,8 +3684,9 @@ void rebuild_address_pool(void) for (i = 0; i < MAXSESSION; ++i) { int ipid; - if (!session[i].ip || !session[i].tunnel) + if (!(session[i].opened && session[i].ip)) continue; + ipid = - lookup_ipmap(htonl(session[i].ip)); if (session[i].ip_pool_index < 0) @@ -2812,7 +3743,7 @@ static void fix_address_pool(int sid) // // Add a block of addresses to the IP pool to hand out. // -static void add_to_ip_pool(u32 addr, u32 mask) +static void add_to_ip_pool(in_addr_t addr, in_addr_t mask) { int i; if (mask == 0) @@ -2863,7 +3794,7 @@ static void initippool() if ((p = (char *)strrchr(buf, '\n'))) *p = 0; if ((p = (char *)strchr(buf, ':'))) { - ipt src; + in_addr_t src; *p = '\0'; src = inet_addr(buf); if (src == INADDR_NONE) @@ -2881,7 +3812,7 @@ static void initippool() { // It's a range int numbits = 0; - u32 start = 0, mask = 0; + in_addr_t start = 0, mask = 0; LOG(2, 0, 0, "Adding IP address range %s\n", buf); *p++ = 0; @@ -2891,7 +3822,7 @@ static void initippool() continue; } start = ntohl(inet_addr(pool)); - mask = (u32)(pow(2, numbits) - 1) << (32 - numbits); + mask = (in_addr_t) (pow(2, numbits) - 1) << (32 - numbits); // Add a static route for this pool LOG(5, 0, 0, "Adding route for address pool %s/%u\n", @@ -2911,7 +3842,7 @@ static void initippool() LOG(1, 0, 0, "IP address pool is %d addresses\n", ip_pool_size - 1); } -void snoop_send_packet(char *packet, u16 size, ipt destination, u16 port) +void snoop_send_packet(uint8_t *packet, uint16_t size, in_addr_t destination, uint16_t port) { struct sockaddr_in snoop_addr = {0}; if (!destination || !port || snoopfd <= 0 || size <= 0 || !packet) @@ -2921,8 +3852,8 @@ void snoop_send_packet(char *packet, u16 size, ipt destination, u16 port) snoop_addr.sin_addr.s_addr = destination; snoop_addr.sin_port = ntohs(port); - LOG(5, 0, 0, "Snooping packet at %p (%d bytes) to %s:%d\n", - packet, size, fmtaddr(snoop_addr.sin_addr.s_addr, 0), + LOG(5, 0, 0, "Snooping %d byte packet to %s:%d\n", size, + fmtaddr(snoop_addr.sin_addr.s_addr, 0), htons(snoop_addr.sin_port)); if (sendto(snoopfd, packet, size, MSG_DONTWAIT | MSG_NOSIGNAL, (void *) &snoop_addr, sizeof(snoop_addr)) < 0) @@ -2931,55 +3862,72 @@ void snoop_send_packet(char *packet, u16 size, ipt destination, u16 port) STAT(packets_snooped); } -static void dump_acct_info() +static int dump_session(FILE **f, sessiont *s) { - char filename[1024]; - char timestr[64]; - time_t t = time(NULL); - int i; - FILE *f = NULL; - + if (!s->opened || !s->ip || !(s->cin_delta || s->cout_delta) || !*s->user || s->walled_garden) + return 1; - CSTAT(call_dump_acct_info); + if (!*f) + { + char filename[1024]; + char timestr[64]; + time_t now = time(NULL); - strftime(timestr, 64, "%Y%m%d%H%M%S", localtime(&t)); - snprintf(filename, 1024, "%s/%s", config->accounting_dir, timestr); + strftime(timestr, sizeof(timestr), "%Y%m%d%H%M%S", localtime(&now)); + snprintf(filename, sizeof(filename), "%s/%s", config->accounting_dir, timestr); - for (i = 0; i < MAXSESSION; i++) - { - if (!session[i].opened || !session[i].ip || !(session[i].cin || session[i].cout) || !*session[i].user || session[i].walled_garden) - continue; - if (!f) + if (!(*f = fopen(filename, "w"))) { - time_t now = time(NULL); - if (!(f = fopen(filename, "w"))) - { - LOG(0, 0, 0, "Can't write accounting info to %s: %s\n", filename, strerror(errno)); - return ; - } - LOG(3, 0, 0, "Dumping accounting information to %s\n", filename); - fprintf(f, "# dslwatch.pl dump file V1.01\n" - "# host: %s\n" - "# time: %ld\n" - "# uptime: %ld\n" - "# format: username ip qos uptxoctets downrxoctets\n", - hostname, - now, - now - basetime); + LOG(0, 0, 0, "Can't write accounting info to %s: %s\n", filename, strerror(errno)); + return 0; } - LOG(4, 0, 0, "Dumping accounting information for %s\n", session[i].user); - fprintf(f, "%s %s %d %u %u\n", - session[i].user, // username - fmtaddr(htonl(session[i].ip), 0), // ip - (session[i].throttle_in || session[i].throttle_out) ? 2 : 1, // qos - (u32)session[i].cin, // uptxoctets - (u32)session[i].cout); // downrxoctets + LOG(3, 0, 0, "Dumping accounting information to %s\n", filename); + fprintf(*f, "# dslwatch.pl dump file V1.01\n" + "# host: %s\n" + "# endpoint: %s\n" + "# time: %ld\n" + "# uptime: %ld\n" + "# format: username ip qos uptxoctets downrxoctets\n", + hostname, + fmtaddr(config->bind_address ? config->bind_address : my_address, 0), + now, + now - basetime); + } + + LOG(4, 0, 0, "Dumping accounting information for %s\n", s->user); + fprintf(*f, "%s %s %d %u %u\n", + s->user, // username + fmtaddr(htonl(s->ip), 0), // ip + (s->throttle_in || s->throttle_out) ? 2 : 1, // qos + (uint32_t) s->cin_delta, // uptxoctets + (uint32_t) s->cout_delta); // downrxoctets + + s->cin_delta = s->cout_delta = 0; + + return 1; +} + +static void dump_acct_info(int all) +{ + int i; + FILE *f = NULL; + + + CSTAT(dump_acct_info); + + if (shut_acct_n) + { + for (i = 0; i < shut_acct_n; i++) + dump_session(&f, &shut_acct[i]); - session[i].pin = session[i].cin = 0; - session[i].pout = session[i].cout = 0; + shut_acct_n = 0; } + if (all) + for (i = 1; i <= config->cluster_highest_sessionid; i++) + dump_session(&f, &session[i]); + if (f) fclose(f); } @@ -3040,7 +3988,7 @@ int main(int argc, char *argv[]) init_tbf(config->num_tbfs); LOG(0, 0, 0, "L2TPNS version " VERSION "\n"); - LOG(0, 0, 0, "Copyright (c) 2003, 2004 Optus Internet Engineering\n"); + LOG(0, 0, 0, "Copyright (c) 2003, 2004, 2005 Optus Internet Engineering\n"); LOG(0, 0, 0, "Copyright (c) 2002 FireBrick (Andrews & Arnold Ltd / Watchfront Ltd) - GPL licenced\n"); { struct rlimit rlim; @@ -3069,7 +4017,7 @@ int main(int argc, char *argv[]) { if ((ret = sched_setscheduler(0, SCHED_FIFO, ¶ms)) == 0) { - LOG(1, 0, 0, "Using FIFO scheduler. Say goodbye to any other processes running\n"); + LOG(1, 0, 0, "Using FIFO scheduler. Say goodbye to any other processes running\n"); } else { @@ -3083,19 +4031,6 @@ int main(int argc, char *argv[]) if (cluster_init() < 0) exit(1); -#ifdef BGP - signal(SIGPIPE, SIG_IGN); - bgp_setup(config->as_number); - bgp_add_route(config->bind_address, 0xffffffff); - for (i = 0; i < BGP_NUM_PEERS; i++) - { - if (config->neighbour[i].name[0]) - bgp_start(&bgp_peers[i], config->neighbour[i].name, - config->neighbour[i].as, config->neighbour[i].keepalive, - config->neighbour[i].hold, 0); /* 0 = routing disabled */ - } -#endif /* BGP */ - inittun(); LOG(1, 0, 0, "Set up on interface %s\n", config->tundevice); @@ -3103,13 +4038,18 @@ int main(int argc, char *argv[]) initrad(); initippool(); - read_state(); + // seed prng + { + unsigned seed = time_now ^ getpid(); + LOG(4, 0, 0, "Seeding the pseudo random generator: %u\n", seed); + srand(seed); + } - signal(SIGHUP, sighup_handler); - signal(SIGTERM, sigterm_handler); - signal(SIGINT, sigterm_handler); - signal(SIGQUIT, sigquit_handler); + signal(SIGHUP, sighup_handler); signal(SIGCHLD, sigchild_handler); + signal(SIGTERM, shutdown_handler); + signal(SIGINT, shutdown_handler); + signal(SIGQUIT, shutdown_handler); // Prevent us from getting paged out if (config->lock_pages) @@ -3128,14 +4068,6 @@ int main(int argc, char *argv[]) mainloop(); -#ifdef BGP - /* try to shut BGP down cleanly; with luck the sockets will be - writable since we're out of the select */ - for (i = 0; i < BGP_NUM_PEERS; i++) - if (bgp_peers[i].state == Established) - bgp_stop(&bgp_peers[i]); -#endif /* BGP */ - /* remove plugins (so cleanup code gets run) */ plugins_done(); @@ -3151,9 +4083,11 @@ int main(int argc, char *argv[]) static void sighup_handler(int sig) { - if (log_stream && log_stream != stderr) + if (log_stream) { - fclose(log_stream); + if (log_stream != stderr) + fclose(log_stream); + log_stream = NULL; } @@ -3193,32 +4127,10 @@ static void sigalrm_handler(int sig) } -static void sigterm_handler(int sig) -{ - LOG(1, 0, 0, "Shutting down cleanly\n"); - if (config->save_state) - dump_state(); - - main_quit++; -} - -static void sigquit_handler(int sig) +static void shutdown_handler(int sig) { - int i; - - LOG(1, 0, 0, "Shutting down without saving sessions\n"); - for (i = 1; i < MAXSESSION; i++) - { - if (session[i].opened) - sessionkill(i, "L2TPNS Closing"); - } - for (i = 1; i < MAXTUNNEL; i++) - { - if (tunnel[i].ip || tunnel[i].state) - tunnelshutdown(i, "L2TPNS Closing"); - } - - main_quit++; + LOG(1, 0, 0, "Shutting down\n"); + main_quit = (sig == SIGQUIT) ? QUIT_SHUTDOWN : QUIT_FAILOVER; } static void sigchild_handler(int sig) @@ -3227,183 +4139,7 @@ static void sigchild_handler(int sig) ; } -static void read_state() -{ - struct stat sb; - int i; - ippoolt itmp; - FILE *f; - char magic[sizeof(DUMP_MAGIC) - 1]; - u32 buf[2]; - - if (!config->save_state) - { - unlink(STATEFILE); - return ; - } - - if (stat(STATEFILE, &sb) < 0) - { - unlink(STATEFILE); - return ; - } - - if (sb.st_mtime < (time(NULL) - 60)) - { - LOG(0, 0, 0, "State file is too old to read, ignoring\n"); - unlink(STATEFILE); - return ; - } - - f = fopen(STATEFILE, "r"); - unlink(STATEFILE); - - if (!f) - { - LOG(0, 0, 0, "Can't read state file: %s\n", strerror(errno)); - exit(1); - } - - if (fread(magic, sizeof(magic), 1, f) != 1 || strncmp(magic, DUMP_MAGIC, sizeof(magic))) - { - LOG(0, 0, 0, "Bad state file magic\n"); - exit(1); - } - - LOG(1, 0, 0, "Reading state information\n"); - if (fread(buf, sizeof(buf), 1, f) != 1 || buf[0] > MAXIPPOOL || buf[1] != sizeof(ippoolt)) - { - LOG(0, 0, 0, "Error/mismatch reading ip pool header from state file\n"); - exit(1); - } - - if (buf[0] > ip_pool_size) - { - LOG(0, 0, 0, "ip pool has shrunk! state = %d, current = %d\n", buf[0], ip_pool_size); - exit(1); - } - - LOG(2, 0, 0, "Loading %u ip addresses\n", buf[0]); - for (i = 0; i < buf[0]; i++) - { - if (fread(&itmp, sizeof(itmp), 1, f) != 1) - { - LOG(0, 0, 0, "Error reading ip %d from state file: %s\n", i, strerror(errno)); - exit(1); - } - - if (itmp.address != ip_address_pool[i].address) - { - LOG(0, 0, 0, "Mismatched ip %d from state file: pool may only be extended\n", i); - exit(1); - } - - memcpy(&ip_address_pool[i], &itmp, sizeof(itmp)); - } - - if (fread(buf, sizeof(buf), 1, f) != 1 || buf[0] != MAXTUNNEL || buf[1] != sizeof(tunnelt)) - { - LOG(0, 0, 0, "Error/mismatch reading tunnel header from state file\n"); - exit(1); - } - - LOG(2, 0, 0, "Loading %u tunnels\n", MAXTUNNEL); - if (fread(tunnel, sizeof(tunnelt), MAXTUNNEL, f) != MAXTUNNEL) - { - LOG(0, 0, 0, "Error reading tunnel data from state file\n"); - exit(1); - } - - for (i = 0; i < MAXTUNNEL; i++) - { - tunnel[i].controlc = 0; - tunnel[i].controls = NULL; - tunnel[i].controle = NULL; - if (*tunnel[i].hostname) - LOG(3, 0, 0, "Created tunnel for %s\n", tunnel[i].hostname); - } - - if (fread(buf, sizeof(buf), 1, f) != 1 || buf[0] != MAXSESSION || buf[1] != sizeof(sessiont)) - { - LOG(0, 0, 0, "Error/mismatch reading session header from state file\n"); - exit(1); - } - - LOG(2, 0, 0, "Loading %u sessions\n", MAXSESSION); - if (fread(session, sizeof(sessiont), MAXSESSION, f) != MAXSESSION) - { - LOG(0, 0, 0, "Error reading session data from state file\n"); - exit(1); - } - - for (i = 0; i < MAXSESSION; i++) - { - session[i].tbf_in = 0; - session[i].tbf_out = 0; - if (session[i].opened) - { - LOG(2, i, 0, "Loaded active session for user %s\n", session[i].user); - if (session[i].ip) - sessionsetup(session[i].tunnel, i); - } - } - - fclose(f); - LOG(0, 0, 0, "Loaded saved state information\n"); -} - -static void dump_state() -{ - FILE *f; - u32 buf[2]; - - if (!config->save_state) - return; - - do - { - if (!(f = fopen(STATEFILE, "w"))) - break; - - LOG(1, 0, 0, "Dumping state information\n"); - - if (fwrite(DUMP_MAGIC, sizeof(DUMP_MAGIC) - 1, 1, f) != 1) - break; - - LOG(2, 0, 0, "Dumping %u ip addresses\n", ip_pool_size); - buf[0] = ip_pool_size; - buf[1] = sizeof(ippoolt); - if (fwrite(buf, sizeof(buf), 1, f) != 1) - break; - if (fwrite(ip_address_pool, sizeof(ippoolt), ip_pool_size, f) != ip_pool_size) - break; - - LOG(2, 0, 0, "Dumping %u tunnels\n", MAXTUNNEL); - buf[0] = MAXTUNNEL; - buf[1] = sizeof(tunnelt); - if (fwrite(buf, sizeof(buf), 1, f) != 1) - break; - if (fwrite(tunnel, sizeof(tunnelt), MAXTUNNEL, f) != MAXTUNNEL) - break; - - LOG(2, 0, 0, "Dumping %u sessions\n", MAXSESSION); - buf[0] = MAXSESSION; - buf[1] = sizeof(sessiont); - if (fwrite(buf, sizeof(buf), 1, f) != 1) - break; - if (fwrite(session, sizeof(sessiont), MAXSESSION, f) != MAXSESSION) - break; - - if (fclose(f) == 0) - return ; // OK - } - while (0); - - LOG(0, 0, 0, "Can't write state information: %s\n", strerror(errno)); - unlink(STATEFILE); -} - -static void build_chap_response(char *challenge, u8 id, u16 challenge_length, char **challenge_response) +static void build_chap_response(uint8_t *challenge, uint8_t id, uint16_t challenge_length, uint8_t **challenge_response) { MD5_CTX ctx; *challenge_response = NULL; @@ -3416,13 +4152,13 @@ static void build_chap_response(char *challenge, u8 id, u16 challenge_length, ch LOG(4, 0, 0, " Building challenge response for CHAP request\n"); - *challenge_response = (char *)calloc(17, 1); + *challenge_response = calloc(17, 1); - MD5Init(&ctx); - MD5Update(&ctx, &id, 1); - MD5Update(&ctx, config->l2tpsecret, strlen(config->l2tpsecret)); - MD5Update(&ctx, challenge, challenge_length); - MD5Final(*challenge_response, &ctx); + MD5_Init(&ctx); + MD5_Update(&ctx, &id, 1); + MD5_Update(&ctx, config->l2tpsecret, strlen(config->l2tpsecret)); + MD5_Update(&ctx, challenge, challenge_length); + MD5_Final(*challenge_response, &ctx); return; } @@ -3441,6 +4177,7 @@ static int facility_value(char *name) static void update_config() { int i; + char *p; static int timeout = 0; static int interval = 0; @@ -3449,9 +4186,12 @@ static void update_config() syslog_log = 0; if (log_stream) { - fclose(log_stream); + if (log_stream != stderr) + fclose(log_stream); + log_stream = NULL; } + if (*config->log_filename) { if (strstr(config->log_filename, "syslog:") == config->log_filename) @@ -3483,7 +4223,6 @@ static void update_config() setbuf(log_stream, NULL); } - // Update radius config->numradiusservers = 0; for (i = 0; i < MAXRADSERVER; i++) @@ -3497,7 +4236,7 @@ static void update_config() // test twice, In case someone works with // a secondary radius server without defining // a primary one, this will work even then. - if (i>0 && !config->radiusport[i]) + if (i > 0 && !config->radiusport[i]) config->radiusport[i] = config->radiusport[i-1]; if (!config->radiusport[i]) config->radiusport[i] = RADPORT; @@ -3506,7 +4245,63 @@ static void update_config() if (!config->numradiusservers) LOG(0, 0, 0, "No RADIUS servers defined!\n"); - config->num_radfds = 2 << RADIUS_SHIFT; + // parse radius_authtypes_s + config->radius_authtypes = config->radius_authprefer = 0; + p = config->radius_authtypes_s; + while (p && *p) + { + char *s = strpbrk(p, " \t,"); + int type = 0; + + if (s) + { + *s++ = 0; + while (*s == ' ' || *s == '\t') + s++; + + if (!*s) + s = 0; + } + + if (!strncasecmp("chap", p, strlen(p))) + type = AUTHCHAP; + else if (!strncasecmp("pap", p, strlen(p))) + type = AUTHPAP; + else + LOG(0, 0, 0, "Invalid RADIUS authentication type \"%s\"\n", p); + + config->radius_authtypes |= type; + if (!config->radius_authprefer) + config->radius_authprefer = type; + + p = s; + } + + if (!config->radius_authtypes) + { + LOG(0, 0, 0, "Defaulting to PAP authentication\n"); + config->radius_authtypes = config->radius_authprefer = AUTHPAP; + } + + // normalise radius_authtypes_s + if (config->radius_authprefer == AUTHPAP) + { + strcpy(config->radius_authtypes_s, "pap"); + if (config->radius_authtypes & AUTHCHAP) + strcat(config->radius_authtypes_s, ", chap"); + } + else + { + strcpy(config->radius_authtypes_s, "chap"); + if (config->radius_authtypes & AUTHPAP) + strcat(config->radius_authtypes_s, ", pap"); + } + + if (!config->radius_dae_port) + config->radius_dae_port = DAEPORT; + + // re-initialise the random number source + initrandom(config->random_device); // Update plugins for (i = 0; i < MAXPLUGINS; i++) @@ -3525,8 +4320,8 @@ static void update_config() remove_plugin(config->old_plugins[i]); } } + memcpy(config->old_plugins, config->plugins, sizeof(config->plugins)); - if (!config->cleanup_interval) config->cleanup_interval = 10; if (!config->multi_read_count) config->multi_read_count = 10; if (!config->cluster_address) config->cluster_address = inet_addr(DEFAULT_MCAST_ADDR); if (!*config->cluster_interface) @@ -3595,28 +4390,28 @@ static void read_config_file() update_config(); } -int sessionsetup(tunnelidt t, sessionidt s) +int sessionsetup(sessionidt s, tunnelidt t) { // A session now exists, set it up - ipt ip; + in_addr_t ip; char *user; sessionidt i; int r; - CSTAT(call_sessionsetup); + CSTAT(sessionsetup); LOG(3, s, t, "Doing session setup for session\n"); - if (!session[s].ip || session[s].ip == 0xFFFFFFFE) + if (!session[s].ip) { assign_ip_address(s); if (!session[s].ip) { - LOG(0, s, t, " No IP allocated. The IP address pool is FULL!\n"); - sessionshutdown(s, "No IP addresses available"); + LOG(0, s, t, " No IP allocated. The IP address pool is FULL!\n"); + sessionshutdown(s, "No IP addresses available.", 2, 7); // try another return 0; } - LOG(3, s, t, " No IP allocated. Assigned %s from pool\n", + LOG(3, s, t, " No IP allocated. Assigned %s from pool\n", fmtaddr(htonl(session[s].ip), 0)); } @@ -3633,8 +4428,16 @@ int sessionsetup(tunnelidt t, sessionidt s) for (i = 1; i <= config->cluster_highest_sessionid; i++) { if (i == s) continue; - if (ip == session[i].ip) sessionkill(i, "Duplicate IP address"); - if (!session[s].walled_garden && !session[i].walled_garden && strcasecmp(user, session[i].user) == 0) + if (!session[s].opened) continue; + if (ip == session[i].ip) + { + sessionkill(i, "Duplicate IP address"); + continue; + } + + if (config->allow_duplicate_users) continue; + if (session[s].walled_garden || session[i].walled_garden) continue; + if (!strcasecmp(user, session[i].user)) sessionkill(i, "Duplicate session for users"); } } @@ -3656,6 +4459,7 @@ int sessionsetup(tunnelidt t, sessionidt s) // convered by a Framed-Route. Anything else is part // of the IP address pool and is already routed, it // just needs to be added to the IP cache. + // IPv6 route setup is done in ppp.c, when IPV6CP is acked. if (session[s].ip_pool_index == -1) // static ip { if (!routed) routeset(s, session[s].ip, 0, 0, 1); @@ -3664,13 +4468,8 @@ int sessionsetup(tunnelidt t, sessionidt s) cache_ipmap(session[s].ip, s); } - if (!session[s].unique_id) - { - // did this session just finish radius? - LOG(3, s, t, "Sending initial IPCP to client\n"); - sendipcp(t, s); - session[s].unique_id = ++last_id; - } + sess_local[s].lcp_authtype = 0; // RADIUS authentication complete + lcp_open(s, t); // transition to Network phase and send initial IPCP // Run the plugin's against this new session. { @@ -3778,6 +4577,10 @@ int load_session(sessionidt s, sessiont *new) } } + // check v6 routing + if (new->ipv6prefixlen && new->ppp.ipv6cp == Opened && session[s].ppp.ipv6cp != Opened) + route6set(s, new->ipv6route, new->ipv6prefixlen, 1); + // check filters if (new->filter_in && (new->filter_in > MAXFILTER || !ip_filters[new->filter_in - 1].name[0])) { @@ -3807,12 +4610,6 @@ int load_session(sessionidt s, sessiont *new) // for walking the sessions to forward byte counts to the master. config->cluster_highest_sessionid = s; - // TEMP: old session struct used a u32 to define the throttle - // speed for both up/down, new uses a u16 for each. Deal with - // sessions from an old master for migration. - if (new->throttle_out == 0 && new->tbf_out) - new->throttle_out = new->throttle_in; - memcpy(&session[s], new, sizeof(session[s])); // Copy over.. // Do fixups into address pool. @@ -3876,6 +4673,7 @@ static int add_plugin(char *plugin_name) radiusnew, radiussend, getconfig, + sessionshutdown, sessionkill, throttle_session, cluster_send_session, @@ -3997,7 +4795,7 @@ static void plugins_done() run_plugin_done(p); } -static void processcontrol(u8 * buf, int len, struct sockaddr_in *addr, int alen) +static void processcontrol(uint8_t *buf, int len, struct sockaddr_in *addr, int alen) { struct nsctl request; struct nsctl response; @@ -4196,6 +4994,9 @@ static tunnelidt new_tunnel() void become_master(void) { int s, i; + static struct event_data d[RADIUS_FDS]; + struct epoll_event e; + run_plugins(PLUGIN_BECOME_MASTER, NULL); // running a bunch of iptables commands is slow and can cause @@ -4205,7 +5006,7 @@ void become_master(void) { for (s = 1; s <= config->cluster_highest_sessionid ; ++s) { - if (!session[s].tunnel) // Not an in-use session. + if (!session[s].opened) // Not an in-use session. continue; run_plugins(PLUGIN_NEW_SESSION_MASTER, &session[s]); @@ -4214,11 +5015,14 @@ void become_master(void) } // add radius fds - for (i = 0; i < config->num_radfds; i++) + e.events = EPOLLIN; + for (i = 0; i < RADIUS_FDS; i++) { - FD_SET(radfds[i], &readset); - if (radfds[i] > readset_n) - readset_n = radfds[i]; + d[i].type = FD_TYPE_RADIUS; + d[i].index = i; + e.data.ptr = &d[i]; + + epoll_ctl(epollfd, EPOLL_CTL_ADD, radfds[i], &e); } } @@ -4237,7 +5041,7 @@ int cmd_show_hist_idle(struct cli_def *cli, char *command, char **argv, int argc for (s = 1; s <= config->cluster_highest_sessionid ; ++s) { int idle; - if (!session[s].tunnel) + if (!session[s].opened) continue; idle = time_now - session[s].last_packet; @@ -4275,7 +5079,7 @@ int cmd_show_hist_open(struct cli_def *cli, char *command, char **argv, int argc for (s = 1; s <= config->cluster_highest_sessionid ; ++s) { int open = 0, d; - if (!session[s].tunnel) + if (!session[s].opened) continue; d = time_now - session[s].opened; @@ -4302,78 +5106,73 @@ int cmd_show_hist_open(struct cli_def *cli, char *command, char **argv, int argc /* Unhide an avp. * - * This unencodes the AVP using the L2TP CHAP secret and the - * previously stored random vector. It replaces the hidden data with - * the cleartext data and returns the length of the cleartext data - * (including the AVP "header" of 6 bytes). - * - * Based on code from rp-l2tpd by Roaring Penguin Software Inc. + * This unencodes the AVP using the L2TP secret and the previously + * stored random vector. It overwrites the hidden data with the + * unhidden AVP subformat. */ -static int unhide_avp(u8 *avp, tunnelidt t, sessionidt s, u16 length) +static void unhide_value(uint8_t *value, size_t len, uint16_t type, uint8_t *vector, size_t vec_len) { MD5_CTX ctx; - u8 *cursor; - u8 digest[16]; - u8 working_vector[16]; - uint16_t hidden_length; - u8 type[2]; - size_t done, todo; - u8 *output; - - // Find the AVP type. - type[0] = *(avp + 4); - type[1] = *(avp + 5); - - // Line up with the hidden data - cursor = output = avp + 6; + uint8_t digest[16]; + uint8_t *last; + size_t d = 0; + uint16_t m = htons(type); // Compute initial pad - MD5Init(&ctx); - MD5Update(&ctx, type, 2); - MD5Update(&ctx, config->l2tpsecret, strlen(config->l2tpsecret)); - MD5Update(&ctx, session[s].random_vector, session[s].random_vector_length); - MD5Final(digest, &ctx); + MD5_Init(&ctx); + MD5_Update(&ctx, (unsigned char *) &m, 2); + MD5_Update(&ctx, config->l2tpsecret, strlen(config->l2tpsecret)); + MD5_Update(&ctx, vector, vec_len); + MD5_Final(digest, &ctx); - // Get hidden length - hidden_length = ((uint16_t) (digest[0] ^ cursor[0])) * 256 + (uint16_t) (digest[1] ^ cursor[1]); + // pointer to last decoded 16 octets + last = value; - // Keep these for later use - working_vector[0] = *cursor; - working_vector[1] = *(cursor + 1); - cursor += 2; - - if (hidden_length > length - 8) + while (len > 0) { - LOG(1, s, t, "Hidden length %d too long in AVP of length %d\n", (int) hidden_length, (int) length); - return 0; + // calculate a new pad based on the last decoded block + if (d >= sizeof(digest)) + { + MD5_Init(&ctx); + MD5_Update(&ctx, config->l2tpsecret, strlen(config->l2tpsecret)); + MD5_Update(&ctx, last, sizeof(digest)); + MD5_Final(digest, &ctx); + + d = 0; + last = value; + } + + *value++ ^= digest[d++]; + len--; } +} + +int find_filter(char const *name, size_t len) +{ + int free = -1; + int i; - /* Decrypt remainder */ - done = 2; - todo = hidden_length; - while (todo) + for (i = 0; i < MAXFILTER; i++) { - working_vector[done] = *cursor; - *output = digest[done] ^ *cursor; - ++output; - ++cursor; - --todo; - ++done; - if (done == 16 && todo) + if (!*ip_filters[i].name) { - // Compute new digest - done = 0; - MD5Init(&ctx); - MD5Update(&ctx, config->l2tpsecret, strlen(config->l2tpsecret)); - MD5Update(&ctx, &working_vector, 16); - MD5Final(digest, &ctx); + if (free < 0) + free = i; + + continue; } - } - return hidden_length + 6; + if (strlen(ip_filters[i].name) != len) + continue; + + if (!strncmp(ip_filters[i].name, name, len)) + return i; + } + + return free; } -static int ip_filter_port(ip_filter_portt *p, portt port) +static int ip_filter_port(ip_filter_portt *p, uint16_t port) { switch (p->op) { @@ -4387,7 +5186,7 @@ static int ip_filter_port(ip_filter_portt *p, portt port) return 0; } -static int ip_filter_flag(u8 op, u8 sflags, u8 cflags, u8 flags) +static int ip_filter_flag(uint8_t op, uint8_t sflags, uint8_t cflags, uint8_t flags) { switch (op) { @@ -4404,15 +5203,15 @@ static int ip_filter_flag(u8 op, u8 sflags, u8 cflags, u8 flags) return 0; } -int ip_filter(u8 *buf, int len, u8 filter) +int ip_filter(uint8_t *buf, int len, uint8_t filter) { - u16 frag_offset; - u8 proto; - ipt src_ip; - ipt dst_ip; - portt src_port = 0; - portt dst_port = 0; - u8 flags = 0; + uint16_t frag_offset; + uint8_t proto; + in_addr_t src_ip; + in_addr_t dst_ip; + uint16_t src_port = 0; + uint16_t dst_port = 0; + uint8_t flags = 0; ip_filter_rulet *rule; if (len < 20) // up to end of destination address @@ -4421,10 +5220,10 @@ int ip_filter(u8 *buf, int len, u8 filter) if ((*buf >> 4) != 4) // IPv4 return 0; - frag_offset = ntohs(*(u16 *) (buf + 6)) & 0x1fff; + frag_offset = ntohs(*(uint16_t *) (buf + 6)) & 0x1fff; proto = buf[9]; - src_ip = *(u32 *) (buf + 12); - dst_ip = *(u32 *) (buf + 16); + src_ip = *(in_addr_t *) (buf + 12); + dst_ip = *(in_addr_t *) (buf + 16); if (frag_offset == 0 && (proto == IPPROTO_TCP || proto == IPPROTO_UDP)) { @@ -4432,8 +5231,8 @@ int ip_filter(u8 *buf, int len, u8 filter) if (len < l + 4) // ports return 0; - src_port = ntohs(*(u16 *) (buf + l)); - dst_port = ntohs(*(u16 *) (buf + l + 2)); + src_port = ntohs(*(uint16_t *) (buf + l)); + dst_port = ntohs(*(uint16_t *) (buf + l + 2)); if (proto == IPPROTO_TCP) { if (len < l + 14) // flags