Add a debian/changelog entry for version 2.2.1-1fdn2.
[l2tpns.git] / l2tpns.c
index 8230945..90ac930 100644 (file)
--- a/l2tpns.c
+++ b/l2tpns.c
@@ -14,7 +14,6 @@ char const *cvs_id_l2tpns = "$Id: l2tpns.c,v 1.176 2011/01/20 12:48:40 bodea Exp
 #define SYSLOG_NAMES
 #include <syslog.h>
 #include <malloc.h>
-#include <math.h>
 #include <net/route.h>
 #include <sys/mman.h>
 #include <netdb.h>
@@ -31,7 +30,7 @@ char const *cvs_id_l2tpns = "$Id: l2tpns.c,v 1.176 2011/01/20 12:48:40 bodea Exp
 #include <sys/time.h>
 #include <sys/resource.h>
 #include <sys/wait.h>
-#include <linux/if.h>
+#include <net/if.h>
 #include <stddef.h>
 #include <time.h>
 #include <dlfcn.h>
@@ -70,9 +69,10 @@ 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.
+char hostname[MAXHOSTNAME] = "";       // us.
 static int tunidx;             // ifr_ifindex of tun device
 int nlseqnum = 0;              // netlink sequence number
+int min_initok_nlseqnum = 0;   // minimun seq number for messages after init is ok
 static int syslog_log = 0;     // are we logging to syslog
 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
@@ -159,6 +159,10 @@ config_descriptt config_values[] = {
        CONFIG("cluster_hb_timeout", cluster_hb_timeout, INT),
        CONFIG("cluster_master_min_adv", cluster_master_min_adv, INT),
        CONFIG("ipv6_prefix", ipv6_prefix, IPv6),
+       CONFIG("cli_bind_address", cli_bind_address, IPv4),
+       CONFIG("hostname", hostname, STRING),
+       CONFIG("nexthop_address", nexthop_address, IPv4),
+       CONFIG("nexthop6_address", nexthop6_address, IPv6),
        { NULL, 0, 0, 0 },
 };
 
@@ -166,8 +170,6 @@ static char *plugin_functions[] = {
        NULL,
        "plugin_pre_auth",
        "plugin_post_auth",
-       "plugin_packet_rx",
-       "plugin_packet_tx",
        "plugin_timer",
        "plugin_new_session",
        "plugin_kill_session",
@@ -199,6 +201,8 @@ struct Tstats *_statistics = NULL;
 struct Tringbuffer *ringbuffer = NULL;
 #endif
 
+static ssize_t netlink_send(struct nlmsghdr *nh);
+static void netlink_addattr(struct nlmsghdr *nh, int type, const void *data, int alen);
 static void cache_ipmap(in_addr_t ip, sessionidt s);
 static void uncache_ipmap(in_addr_t ip);
 static void cache_ipv6map(struct in6_addr ip, int prefixlen, sessionidt s);
@@ -418,43 +422,65 @@ void random_data(uint8_t *buf, int len)
 // via BGP if enabled, and stuffs it into the
 // 'sessionbyip' cache.
 //
-// 'ip' and 'mask' must be in _host_ order.
+// 'ip' must be in _host_ order.
 //
-static void routeset(sessionidt s, in_addr_t ip, in_addr_t mask, in_addr_t gw, int add)
+static void routeset(sessionidt s, in_addr_t ip, int prefixlen, in_addr_t gw, int add)
 {
-       struct rtentry r;
+       struct {
+               struct nlmsghdr nh;
+               struct rtmsg rt;
+               char buf[32];
+       } req;
        int i;
+       in_addr_t n_ip;
 
-       if (!mask) mask = 0xffffffff;
+       if (!prefixlen) prefixlen = 32;
 
-       ip &= mask;             // Force the ip to be the first one in the route.
+       ip &= 0xffffffff << (32 - prefixlen);;  // Force the ip to be the first one in the route.
 
-       memset(&r, 0, sizeof(r));
-       r.rt_dev = config->tundevice;
-       r.rt_dst.sa_family = AF_INET;
-       *(uint32_t *) & (((struct sockaddr_in *) &r.rt_dst)->sin_addr.s_addr) = htonl(ip);
-       r.rt_gateway.sa_family = AF_INET;
-       *(uint32_t *) & (((struct sockaddr_in *) &r.rt_gateway)->sin_addr.s_addr) = htonl(gw);
-       r.rt_genmask.sa_family = AF_INET;
-       *(uint32_t *) & (((struct sockaddr_in *) &r.rt_genmask)->sin_addr.s_addr) = htonl(mask);
-       r.rt_flags = (RTF_UP | RTF_STATIC);
+       memset(&req, 0, sizeof(req));
+
+       if (add)
+       {
+               req.nh.nlmsg_type = RTM_NEWROUTE;
+               req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_REPLACE;
+       }
+       else
+       {
+               req.nh.nlmsg_type = RTM_DELROUTE;
+               req.nh.nlmsg_flags = NLM_F_REQUEST;
+       }
+
+       req.nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.rt));
+
+       req.rt.rtm_family = AF_INET;
+       req.rt.rtm_dst_len = prefixlen;
+       req.rt.rtm_table = RT_TABLE_MAIN;
+       req.rt.rtm_protocol = RTPROT_BOOT; // XXX
+       req.rt.rtm_scope = RT_SCOPE_LINK;
+       req.rt.rtm_type = RTN_UNICAST;
+
+       netlink_addattr(&req.nh, RTA_OIF, &tunidx, sizeof(int));
+       n_ip = htonl(ip);
+       netlink_addattr(&req.nh, RTA_DST, &n_ip, sizeof(n_ip));
        if (gw)
-               r.rt_flags |= RTF_GATEWAY;
-       else if (mask == 0xffffffff)
-               r.rt_flags |= RTF_HOST;
+       {
+               n_ip = htonl(gw);
+               netlink_addattr(&req.nh, RTA_GATEWAY, &n_ip, sizeof(n_ip));
+       }
 
-       LOG(1, s, 0, "Route %s %s/%s%s%s\n", add ? "add" : "del",
-           fmtaddr(htonl(ip), 0), fmtaddr(htonl(mask), 1),
+       LOG(1, s, 0, "Route %s %s/%d%s%s\n", add ? "add" : "del",
+           fmtaddr(htonl(ip), 0), prefixlen,
            gw ? " via" : "", gw ? fmtaddr(htonl(gw), 2) : "");
 
-       if (ioctl(ifrfd, add ? SIOCADDRT : SIOCDELRT, (void *) &r) < 0)
-               LOG(0, 0, 0, "routeset() error in ioctl: %s\n", strerror(errno));
+       if (netlink_send(&req.nh) < 0)
+               LOG(0, 0, 0, "routeset() error in sending netlink message: %s\n", strerror(errno));
 
 #ifdef BGP
        if (add)
-               bgp_add_route(htonl(ip), htonl(mask));
+               bgp_add_route(htonl(ip), prefixlen);
        else
-               bgp_del_route(htonl(ip), htonl(mask));
+               bgp_del_route(htonl(ip), prefixlen);
 #endif /* BGP */
 
                // Add/Remove the IPs to the 'sessionbyip' cache.
@@ -470,40 +496,68 @@ static void routeset(sessionidt s, in_addr_t ip, in_addr_t mask, in_addr_t gw, i
                if (!add)       // Are we deleting a route?
                        s = 0;  // Caching the session as '0' is the same as uncaching.
 
-               for (i = ip; (i&mask) == (ip&mask) ; ++i)
+               for (i = ip; i < ip+(1<<(32-prefixlen)) ; ++i)
                        cache_ipmap(i, s);
        }
 }
 
 void route6set(sessionidt s, struct in6_addr ip, int prefixlen, int add)
 {
-       struct in6_rtmsg rt;
+       struct {
+               struct nlmsghdr nh;
+               struct rtmsg rt;
+               char buf[64];
+       } req;
+       int metric;
        char ipv6addr[INET6_ADDRSTRLEN];
 
-       if (ifr6fd < 0)
+       if (!config->ipv6_prefix.s6_addr[0])
        {
                LOG(0, 0, 0, "Asked to set IPv6 route, but IPv6 not setup.\n");
                return;
        }
 
-       memset(&rt, 0, sizeof(rt));
+       memset(&req, 0, sizeof(req));
 
-       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;
+       if (add)
+       {
+               req.nh.nlmsg_type = RTM_NEWROUTE;
+               req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_REPLACE;
+       }
+       else
+       {
+               req.nh.nlmsg_type = RTM_DELROUTE;
+               req.nh.nlmsg_flags = NLM_F_REQUEST;
+       }
+
+       req.nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.rt));
+
+       req.rt.rtm_family = AF_INET6;
+       req.rt.rtm_dst_len = prefixlen;
+       req.rt.rtm_table = RT_TABLE_MAIN;
+       req.rt.rtm_protocol = RTPROT_BOOT; // XXX
+       req.rt.rtm_scope = RT_SCOPE_LINK;
+       req.rt.rtm_type = RTN_UNICAST;
+
+       netlink_addattr(&req.nh, RTA_OIF, &tunidx, sizeof(int));
+       netlink_addattr(&req.nh, RTA_DST, &ip, sizeof(ip));
+       metric = 1;
+       netlink_addattr(&req.nh, RTA_METRICS, &metric, sizeof(metric));
 
        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));
+       if (netlink_send(&req.nh) < 0)
+               LOG(0, 0, 0, "route6set() error in sending netlink message: %s\n", strerror(errno));
 
-       // FIXME: need to add BGP routing (RFC2858)
+#ifdef BGP
+       if (add)
+               bgp_add_route6(ip, prefixlen);
+       else
+               bgp_del_route6(ip, prefixlen);
+#endif /* BGP */
 
        if (s)
        {
@@ -540,7 +594,7 @@ static void initnetlink(void)
        }
 }
 
-static ssize_t netlink_send(struct nlmsghdr *nh, int want_ack)
+static ssize_t netlink_send(struct nlmsghdr *nh)
 {
        struct sockaddr_nl nladdr;
        struct iovec iov;
@@ -548,8 +602,6 @@ static ssize_t netlink_send(struct nlmsghdr *nh, int want_ack)
 
        nh->nlmsg_pid = getpid();
        nh->nlmsg_seq = ++nlseqnum;
-       if (want_ack)
-               nh->nlmsg_flags |= NLM_F_ACK;
 
        // set kernel address
        memset(&nladdr, 0, sizeof(nladdr));
@@ -561,7 +613,7 @@ static ssize_t netlink_send(struct nlmsghdr *nh, int want_ack)
        return sendmsg(nlfd, &msg, 0);
 }
 
-static ssize_t netlink_recv(char *buf, ssize_t len)
+static ssize_t netlink_recv(void *buf, ssize_t len)
 {
        struct sockaddr_nl nladdr;
        struct iovec iov;
@@ -577,47 +629,34 @@ static ssize_t netlink_recv(char *buf, ssize_t len)
        return recvmsg(nlfd, &msg, 0);
 }
 
-static ssize_t netlink_check_ack(struct nlmsghdr *ack_nh, int msg_count)
+/* adapted from iproute2 */
+static void netlink_addattr(struct nlmsghdr *nh, int type, const void *data, int alen)
 {
-       ssize_t len;
-       uint32_t seqnum;
-
-       // expect ack
-       for (; NLMSG_OK (ack_nh, len); ack_nh = NLMSG_NEXT (ack_nh, len))
-       {
-               if (ack_nh->nlmsg_type == NLMSG_DONE)
-                       return msg_count != 0 || seqnum != nlseqnum;
-                       // 1 => lost an ack
-
-               if (ack_nh->nlmsg_type == NLMSG_ERROR)
-               {
-                       struct nlmsgerr *errmsg = NLMSG_DATA(ack_nh);
-                       if (errmsg->error)
-                               return errmsg->error; // got an error back
-                       else
-                       {
-                               // ack received
-                               --msg_count;
-                               if (errmsg->msg.nlmsg_seq > seqnum);
-                                       seqnum = errmsg->msg.nlmsg_seq;
-                       }
-               }
-               else
-                       // unknown message
-                       LOG(3, 0, 0, "Got an unknown netlink message: type %d\n", ack_nh->nlmsg_type);
-       }
-
-       return 1; // malformed message?!
+       int len = RTA_LENGTH(alen);
+       struct rtattr *rta;
+
+       rta = (struct rtattr *)(((void *)nh) + NLMSG_ALIGN(nh->nlmsg_len));
+       rta->rta_type = type;
+       rta->rta_len = len;
+       memcpy(RTA_DATA(rta), data, alen);
+       nh->nlmsg_len = NLMSG_ALIGN(nh->nlmsg_len) + RTA_ALIGN(len);
 }
 
+// messages corresponding to different phases seq number
+static char *tun_nl_phase_msg[] = {
+       "initialized",
+       "getting tun interface index",
+       "setting tun interface parameters",
+       "setting tun IPv4 address",
+       "setting tun LL IPv6 address",
+       "setting tun global IPv6 address",
+};
+
 //
 // Set up TUN interface
 static void inittun(void)
 {
-       struct ifinfomsg ifinfo;
        struct ifreq ifr;
-       char buf[4096];
-       struct nlmsghdr resp_nh;
 
        memset(&ifr, 0, sizeof(ifr));
        ifr.ifr_flags = IFF_TUN;
@@ -637,188 +676,136 @@ static void inittun(void)
                LOG(0, 0, 0, "Can't set tun interface: %s\n", strerror(errno));
                exit(1);
        }
-       assert(strlen(ifr.ifr_name) < sizeof(config->tundevice));
-       strncpy(config->tundevice, ifr.ifr_name, sizeof(config->tundevice) - 1);
+       assert(strlen(ifr.ifr_name) < sizeof(config->tundevice) - 1);
+       strncpy(config->tundevice, ifr.ifr_name, sizeof(config->tundevice));
 
+       tunidx = if_nametoindex(config->tundevice);
+       if (tunidx == 0)
        {
-               // get the interface index
-               ssize_t len;
-               struct {
-                       struct nlmsghdr nh;
-                       struct ifinfomsg ifinfo __attribute__ ((aligned(NLMSG_ALIGNTO)));
-                       struct rtattr ifname_rta __attribute__ ((aligned(RTA_ALIGNTO)));
-                       char ifname[IFNAMSIZ];
-               } req;
-
-               req.nh.nlmsg_type = RTM_GETLINK;
-               req.nh.nlmsg_flags = NLM_F_REQUEST;
-
-               req.ifinfo.ifi_family = AF_UNSPEC; // as the man says
-
-               req.ifname_rta.rta_len = RTA_LENGTH(strlen(config->tundevice)+1);
-               req.ifname_rta.rta_type = IFLA_IFNAME;
-               strncpy(req.ifname, config->tundevice, IFNAMSIZ-1);
-
-               if(netlink_send(&req.nh, 0) < 0 || (len = netlink_recv(buf, sizeof(buf))) < 0)
-               {
-                       LOG(0, 0, 0, "Error getting tun ifindex: %s\n", strerror(errno));
-                       exit(1);
-               }
-
-               resp_nh = (struct nlmsghdr *)buf;
-               if (NLMSG_OK (resp_nh, len) && resp_nh.nlmsg_type == RTM_GETLINK)
-               {
-                       memcpy(&ifinfo, NLMSG_DATA(resp_nh), sizeof(ifinfo));
-                       // got index
-                       tunidx = ifinfo.ifi_index;
-               }
-               else
-               {
-                       LOG(0, 0, 0, "Malformed answer getting tun ifindex\n");
-                       exit(1);
-               }
+               LOG(0, 0, 0, "Can't get tun interface index\n");
+               exit(1);
        }
+
        {
-               ssize_t err;
                struct {
                        // interface setting
                        struct nlmsghdr nh;
-                       struct ifinfomsg ifinfo __attribute__ ((aligned(NLMSG_ALIGNTO)));
-                       struct rtattr txqlen_rta __attribute__ ((aligned(RTA_ALIGNTO)));
-                       uint32_t txqlen;
-                       struct rtattr mtu_rta __attribute__ ((aligned(RTA_ALIGNTO)));
-                       uint32_t mtu;
-
-                       // IPv4 address
-                       struct nlmsghdr addr_nh __attribute __ ((aligned(NLMSG_ALIGNTO)));
-                       struct ifaddrmsg ifaddr __attribute __ ((aligned(NLMSG_ALIGNTO)));
-                       struct rtattr addr_rta __attribute __ ((aligned(RTA_ALIGNTO)));
-                       struct in_addr addr;
-
-                       // IPv6 LL address
-                       struct nlmsghdr lladdr6_nh __attribute __ ((aligned(NLMSG_ALIGNTO)));
-                       struct ifaddrmsg llifaddr6 __attribute __ ((aligned(NLMSG_ALIGNTO)));
-                       struct rtattr lladdr6_rta __attribute __ ((aligned(RTA_ALIGNTO)));
-                       struct in6_addr lladdr6;
-
-                       // IPv6 global address
-                       struct nlmsghdr addr6_nh __attribute __ ((aligned(NLMSG_ALIGNTO)));
-                       struct ifaddrmsg ifaddr6 __attribute __ ((aligned(NLMSG_ALIGNTO)));
-                       struct rtattr addr6_rta __attribute __ ((aligned(RTA_ALIGNTO)));
-                       struct in6_addr addr6;
-
-                       // end header
-                       struct nlmsghdr end_nh __attribute __ ((aligned(NLMSG_ALIGNTO)));
+                       union {
+                               struct ifinfomsg ifinfo;
+                               struct ifaddrmsg ifaddr;
+                       } ifmsg;
+                       char rtdata[32]; // 32 should be enough
                } req;
-               char buf[4096];
+               uint32_t txqlen, mtu;
+               in_addr_t ip;
 
                memset(&req, 0, sizeof(req));
 
                req.nh.nlmsg_type = RTM_NEWLINK;
-               req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_MULTI;
+               req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_MULTI;
+               req.nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.ifmsg.ifinfo));
 
-               req.ifinfo = ifinfo;
-               req.ifinfo.ifi_flags |= IFF_UP; // set interface up
-               req.ifinfo.ifi_change = IFF_UP; // only change this flag
+               req.ifmsg.ifinfo.ifi_family = AF_UNSPEC;
+               req.ifmsg.ifinfo.ifi_index = tunidx;
+               req.ifmsg.ifinfo.ifi_flags |= IFF_UP; // set interface up
+               req.ifmsg.ifinfo.ifi_change = IFF_UP; // only change this flag
 
-               req.txqlen_rta.rta_len = RTA_LENGTH(sizeof(req.txqlen));
-               req.txqlen_rta.rta_type = IFLA_TXQLEN;
                /* Bump up the qlen to deal with bursts from the network */
-               req.txqlen = 1000;
-
-               req.mtu_rta.rta_len = RTA_LENGTH(sizeof(req.txqlen));
-               req.mtu_rta.rta_type = IFLA_MTU;
+               txqlen = 1000;
+               netlink_addattr(&req.nh, IFLA_TXQLEN, &txqlen, sizeof(txqlen));
                /* set MTU to modem MRU */
-               req.mtu = MRU;
+               mtu = MRU;
+               netlink_addattr(&req.nh, IFLA_MTU, &mtu, sizeof(mtu));
 
-               req.nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.ifinfo)
-                               + req.txqlen_rla.rta_len
-                               + req.mtu_rta.rta_len);
+               if (netlink_send(&req.nh) < 0)
+                       goto senderror;
+
+               memset(&req, 0, sizeof(req));
 
-               req.addr_nh.nlmsg_type = RTM_NEWADDR;
-               req.addr_nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_MULTI;
+               req.nh.nlmsg_type = RTM_NEWADDR;
+               req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_REPLACE | NLM_F_MULTI;
+               req.nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.ifmsg.ifaddr));
 
-               req.ifaddr.ifa_family = AF_INET;
-               req.ifaddr.ifa_prefixlen = 32;
-               req.ifaddr.ifa_index = ifinfo.ifi_index;
+               req.ifmsg.ifaddr.ifa_family = AF_INET;
+               req.ifmsg.ifaddr.ifa_prefixlen = 32;
+               req.ifmsg.ifaddr.ifa_scope = RT_SCOPE_UNIVERSE;
+               req.ifmsg.ifaddr.ifa_index = tunidx;
 
-               req.addr_rta.rta_len = RTA_LENGTH(sizeof(req.addr));
-               req.addr_rta.rta_type = IFA_ADDRESS;
-               req.addr.s_addr = config->bind_address ?
-                       config->bind_address : 0x01010101; // 1.1.1.1
+               if (config->bind_address)
+                       ip = config->bind_address;
+               else
+                       ip = 0x01010101; // 1.1.1.1
+               netlink_addattr(&req.nh, IFA_LOCAL, &ip, sizeof(ip));
 
-               req.addr_nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.ifaddr)
-                               + req.addr_rta.rta_len);
+               if (netlink_send(&req.nh) < 0)
+                       goto senderror;
 
                // Only setup IPv6 on the tun device if we have a configured prefix
                if (config->ipv6_prefix.s6_addr[0]) {
-                       req.lladdr6_nh.nlmsg_type = RTM_NEWADDR;
-                       req.lladdr6_nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_MULTI;
+                       struct in6_addr ip6;
+
+                       memset(&req, 0, sizeof(req));
 
-                       req.llifaddr6.ifa_family = AF_INET6;
-                       req.llifaddr6.ifa_prefixlen = 64;
-                       req.llifaddr6.ifa_scope = RT_SCOPE_LINK;
-                       req.llifaddr6.ifa_index = ifinfo.ifi_index;
+                       req.nh.nlmsg_type = RTM_NEWADDR;
+                       req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_REPLACE | NLM_F_MULTI;
+                       req.nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.ifmsg.ifaddr));
+
+                       req.ifmsg.ifaddr.ifa_family = AF_INET6;
+                       req.ifmsg.ifaddr.ifa_prefixlen = 64;
+                       req.ifmsg.ifaddr.ifa_scope = RT_SCOPE_LINK;
+                       req.ifmsg.ifaddr.ifa_index = tunidx;
 
-                       req.lladdr6_rta.rta_len = RTA_LENGTH(sizeof(req.lladdr6));
-                       req.lladdr6_rta.rta_type = IFA_LOCAL;
                        // Link local address is FE80::1
-                       memset(&req.lladdr6, 0, sizeof(req.lladdr6));
-                       req.lladdr6.s6_addr[0] = 0xFE;
-                       req.lladdr6.s6_addr[1] = 0x80;
-                       req.lladdr6.s6_addr[15] = 1;
+                       memset(&ip6, 0, sizeof(ip6));
+                       ip6.s6_addr[0] = 0xFE;
+                       ip6.s6_addr[1] = 0x80;
+                       ip6.s6_addr[15] = 1;
+                       netlink_addattr(&req.nh, IFA_LOCAL, &ip6, sizeof(ip6));
 
-                       req.lladdr6_nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.ifaddr)
-                                       + req.lladdr6_rta.rta_len);
+                       if (netlink_send(&req.nh) < 0)
+                               goto senderror;
 
-                       req.addr6_nh.nlmsg_type = RTM_NEWADDR;
-                       req.addr6_nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_MULTI;
+                       memset(&req, 0, sizeof(req));
 
-                       req.ifaddr6.ifa_family = AF_INET6;
-                       req.ifaddr6.ifa_prefixlen = 64;
-                       req.ifaddr6.ifa_scope = RT_SCOPE_UNIVERSE;
-                       req.ifaddr6.ifa_index = ifinfo.ifi_index;
+                       req.nh.nlmsg_type = RTM_NEWADDR;
+                       req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_REPLACE | NLM_F_MULTI;
+                       req.nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.ifmsg.ifaddr));
 
-                       req.addr6_rta.rta_len = RTA_LENGTH(sizeof(req.addr6));
-                       req.addr6_rta.rta_type = IFA_LOCAL;
-                       // Global address is prefix::1
-                       req.addr6 = config->ipv6_prefix;
-                       req.addr6.s6_addr[15] = 1;
+                       req.ifmsg.ifaddr.ifa_family = AF_INET6;
+                       req.ifmsg.ifaddr.ifa_prefixlen = 64;
+                       req.ifmsg.ifaddr.ifa_scope = RT_SCOPE_UNIVERSE;
+                       req.ifmsg.ifaddr.ifa_index = tunidx;
 
-                       req.lladdr6_nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.ifaddr)
-                                       + req.lladdr6_rta.rta_len);
+                       // Global address is prefix::1
+                       ip6 = config->ipv6_prefix;
+                       ip6.s6_addr[15] = 1;
+                       netlink_addattr(&req.nh, IFA_LOCAL, &ip6, sizeof(ip6));
 
-                       req.end_nh.nlmsg_type = NLMSG_DONE;
-                       req.end_nh.nlmsg_len = NLMSG_LENGTH(0);
-               }
-               else
-               {
-                       // we terminate at the beginning of IPv6 stuff
-                       req.lladdr6_nh.nlmsg_type = NLMSG_DONE;
-                       req.lladdr6_nh.nlmsg_len = NLMSG_LENGTH(0);
+                       if (netlink_send(&req.nh) < 0)
+                               goto senderror;
                }
 
-               if (netlink_send(&req.nh, 1) < 0)
-                       LOG(0, 0, 0, "Error setting up tun device: %s\n", strerror(errno));
-                       exit(1);
-               }
+               memset(&req, 0, sizeof(req));
 
-               if (netlink_recv(buf, sizeof(buf)) < 0)
-               {
-                       LOG(0, 0, 0, "Error waiting ack for setting up tun device: %s\n", strerror(errno));
-                       exit(1);
-               }
+               req.nh.nlmsg_type = NLMSG_DONE;
+               req.nh.nlmsg_len = NLMSG_LENGTH(0);
 
-               resp_nh = (struct nlmsghdr *)buf;
-               if ((err = netlink_check_ack(resp_nh, 4)))
-               {
-                       if (err < 0)
-                               LOG(0, 0, 0, "Error while receiving tun device ack: %s\n", strerror(errno));
-                       else
-                               LOG(0, 0, 0, "Error while receiving tun device ack\n", strerror(errno));
-                       exit(1);
-               }
+               if (netlink_send(&req.nh) < 0)
+                       goto senderror;
+
+               // if we get an error for seqnum < min_initok_nlseqnum,
+               // we must exit as initialization went wrong
+               if (config->ipv6_prefix.s6_addr[0])
+                       min_initok_nlseqnum = 5 + 1; // idx + if + addr + 2*addr6
+               else
+                       min_initok_nlseqnum = 3 + 1; // idx + if + addr
        }
+
+       return;
+
+senderror:
+       LOG(0, 0, 0, "Error while setting up tun device: %s\n", strerror(errno));
+       exit(1);
 }
 
 // set up UDP ports
@@ -1964,11 +1951,11 @@ void sessionshutdown(sessionidt s, char const *reason, int cdn_result, int cdn_e
                int routed = 0;
                for (r = 0; r < MAXROUTE && session[s].route[r].ip; r++)
                {
-                       if ((session[s].ip & session[s].route[r].mask) ==
-                           (session[s].route[r].ip & session[s].route[r].mask))
+                       if ((session[s].ip >> (32-session[s].route[r].prefixlen)) ==
+                           (session[s].route[r].ip >> (32-session[s].route[r].prefixlen)))
                                routed++;
 
-                       if (del_routes) routeset(s, session[s].route[r].ip, session[s].route[r].mask, 0, 0);
+                       if (del_routes) routeset(s, session[s].route[r].ip, session[s].route[r].prefixlen, 0, 0);
                        session[s].route[r].ip = 0;
                }
 
@@ -2972,7 +2959,7 @@ void processudp(uint8_t *buf, int len, struct sockaddr_in *addr)
 
                                        // Set multilink options before sending initial LCP packet
                                        sess_local[s].mp_mrru = 1614;
-                                       sess_local[s].mp_epdis = config->bind_address ? config->bind_address : my_address;
+                                       sess_local[s].mp_epdis = ntohl(config->bind_address ? config->bind_address : my_address);
 
                                        sendlcp(s, t);
                                        change_state(s, lcp, RequestSent);
@@ -3600,35 +3587,36 @@ static int still_busy(void)
        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)
+       static time_t stopped_bgp = 0;
+       if (bgp_configured)
+       {
+               if (!stopped_bgp)
                {
-                       if (!stopped_bgp)
-                       {
-                               LOG(1, 0, 0, "Shutting down in %d seconds, stopping BGP...\n", QUIT_DELAY);
+                       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]);
+                       for (i = 0; i < BGP_NUM_PEERS; i++)
+                               if (bgp_peers[i].state == Established)
+                                       bgp_stop(&bgp_peers[i]);
 
-                               stopped_bgp = time_now;
+                       stopped_bgp = time_now;
 
+                       if (!config->cluster_iam_master)
+                       {
                                // we don't want to become master
                                cluster_send_ping(0);
 
                                return 1;
                        }
-
-                       if (time_now < (stopped_bgp + QUIT_DELAY))
-                               return 1;
                }
+
+               if (!config->cluster_iam_master && time_now < (stopped_bgp + QUIT_DELAY))
+                       return 1;
+       }
 #endif /* BGP */
 
+       if (!config->cluster_iam_master)
                return 0;
-       }
 
        if (main_quit == QUIT_SHUTDOWN)
        {
@@ -3694,8 +3682,8 @@ static int still_busy(void)
 # include "fake_epoll.h"
 #endif
 
-// the base set of fds polled: cli, cluster, tun, udp, control, dae
-#define BASE_FDS       6
+// the base set of fds polled: cli, cluster, tun, udp, control, dae, netlink
+#define BASE_FDS       7
 
 // additional polled fds
 #ifdef BGP
@@ -3719,8 +3707,8 @@ static void mainloop(void)
                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);
+       LOG(4, 0, 0, "Beginning of main loop.  clifd=%d, cluster_sockfd=%d, tunfd=%d, udpfd=%d, controlfd=%d, daefd=%d, nlfd=%d\n",
+               clifd, cluster_sockfd, tunfd, udpfd, controlfd, daefd, nlfd);
 
        /* setup our fds to poll for input */
        {
@@ -3756,6 +3744,10 @@ static void mainloop(void)
                d[i].type = FD_TYPE_DAE;
                e.data.ptr = &d[i++];
                epoll_ctl(epollfd, EPOLL_CTL_ADD, daefd, &e);
+
+               d[i].type = FD_TYPE_NETLINK;
+               e.data.ptr = &d[i++];
+               epoll_ctl(epollfd, EPOLL_CTL_ADD, nlfd, &e);
        }
 
 #ifdef BGP
@@ -3769,7 +3761,8 @@ static void mainloop(void)
                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 */
+                               config->neighbour[i].hold, config->neighbour[i].update_source,
+                               0); /* 0 = routing disabled */
        }
 #endif /* BGP */
 
@@ -3892,6 +3885,32 @@ static void mainloop(void)
                                        break;
 #endif /* BGP */
 
+                               case FD_TYPE_NETLINK:
+                               {
+                                       struct nlmsghdr *nh = (struct nlmsghdr *)buf;
+                                       s = netlink_recv(buf, sizeof(buf));
+                                       if (nh->nlmsg_type == NLMSG_ERROR)
+                                       {
+                                               struct nlmsgerr *errmsg = NLMSG_DATA(nh);
+                                               if (errmsg->error)
+                                               {
+                                                       if (errmsg->msg.nlmsg_seq < min_initok_nlseqnum)
+                                                       {
+                                                               LOG(0, 0, 0, "Got a fatal netlink error (while %s): %s\n", tun_nl_phase_msg[nh->nlmsg_seq], strerror(-errmsg->error));
+                                                               exit(1);
+                                                       }
+                                                       else
+
+                                                               LOG(0, 0, 0, "Got a netlink error: %s\n", strerror(-errmsg->error));
+                                               }
+                                               // else it's a ack
+                                       }
+                                       else
+                                               LOG(1, 0, 0, "Got a unknown netlink message: type %d seq %d flags %d\n", nh->nlmsg_type, nh->nlmsg_seq, nh->nlmsg_flags);
+                                       n--;
+                                       break;
+                               }
+
                                default:
                                        LOG(0, 0, 0, "Unexpected fd type returned from epoll_wait: %d\n", d->type);
                                }
@@ -3963,6 +3982,11 @@ static void mainloop(void)
                                more++;
                        }
                }
+#ifdef BGP
+               else
+                       /* no event received, but timers could still have expired */
+                       bgp_process_peers_timers();
+#endif /* BGP */
 
                if (time_changed)
                {
@@ -4270,9 +4294,14 @@ static void initdata(int optdebug, char *optconfig)
 
        if (!*hostname)
        {
-               // Grab my hostname unless it's been specified
-               gethostname(hostname, sizeof(hostname));
-               stripdomain(hostname);
+               if (!*config->hostname)
+               {
+                       // Grab my hostname unless it's been specified
+                       gethostname(hostname, sizeof(hostname));
+                       stripdomain(hostname);
+               }
+               else
+                       strcpy(hostname, config->hostname);
        }
 
        _statistics->start_time = _statistics->last_reset = time(NULL);
@@ -4450,18 +4479,18 @@ 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(in_addr_t addr, in_addr_t mask)
+static void add_to_ip_pool(in_addr_t addr, int prefixlen)
 {
        int i;
-       if (mask == 0)
-               mask = 0xffffffff;      // Host route only.
+       if (prefixlen == 0)
+               prefixlen = 32;         // Host route only.
 
-       addr &= mask;
+       addr &= 0xffffffff << (32 - prefixlen);
 
        if (ip_pool_size >= MAXIPPOOL)  // Pool is full!
                return ;
 
-       for (i = addr ;(i & mask) == addr; ++i)
+       for (i = addr ; i < addr+(1<<(32-prefixlen)); ++i)
        {
                if ((i & 0xff) == 0 || (i&0xff) == 255)
                        continue;       // Skip 0 and broadcast addresses.
@@ -4519,7 +4548,7 @@ static void initippool()
                {
                        // It's a range
                        int numbits = 0;
-                       in_addr_t start = 0, mask = 0;
+                       in_addr_t start = 0;
 
                        LOG(2, 0, 0, "Adding IP address range %s\n", buf);
                        *p++ = 0;
@@ -4529,15 +4558,14 @@ static void initippool()
                                continue;
                        }
                        start = ntohl(inet_addr(pool));
-                       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",
-                               fmtaddr(htonl(start), 0), 32 + mask);
+                       LOG(5, 0, 0, "Adding route for address pool %s/%d\n",
+                               fmtaddr(htonl(start), 0), numbits);
 
-                       routeset(0, start, mask, 0, 1);
+                       routeset(0, start, numbits, 0, 1);
 
-                       add_to_ip_pool(start, mask);
+                       add_to_ip_pool(start, numbits);
                }
                else
                {
@@ -4733,6 +4761,8 @@ int main(int argc, char *argv[])
                }
        }
 
+       initnetlink();
+
        /* Set up the cluster communications port. */
        if (cluster_init() < 0)
                exit(1);
@@ -5184,11 +5214,11 @@ int sessionsetup(sessionidt s, tunnelidt t)
                // Add the route for this session.
                for (r = 0; r < MAXROUTE && session[s].route[r].ip; r++)
                {
-                       if ((session[s].ip & session[s].route[r].mask) ==
-                           (session[s].route[r].ip & session[s].route[r].mask))
+                       if ((session[s].ip >> (32-session[s].route[r].prefixlen)) ==
+                           (session[s].route[r].ip >> (32-session[s].route[r].prefixlen)))
                                routed++;
 
-                       routeset(s, session[s].route[r].ip, session[s].route[r].mask, 0, 1);
+                       routeset(s, session[s].route[r].ip, session[s].route[r].prefixlen, 0, 1);
                }
 
                // Static IPs need to be routed if not already
@@ -5259,7 +5289,7 @@ int load_session(sessionidt s, sessiont *new)
 
        for (i = 0; !newip && i < MAXROUTE && (session[s].route[i].ip || new->route[i].ip); i++)
                if (new->route[i].ip != session[s].route[i].ip ||
-                   new->route[i].mask != session[s].route[i].mask)
+                   new->route[i].prefixlen != session[s].route[i].prefixlen)
                        newip++;
 
        // needs update
@@ -5270,11 +5300,11 @@ int load_session(sessionidt s, sessiont *new)
                // remove old routes...
                for (i = 0; i < MAXROUTE && session[s].route[i].ip; i++)
                {
-                       if ((session[s].ip & session[s].route[i].mask) ==
-                           (session[s].route[i].ip & session[s].route[i].mask))
+                       if ((session[s].ip >> (32-session[s].route[i].prefixlen)) ==
+                           (session[s].route[i].ip >> (32-session[s].route[i].prefixlen)))
                                routed++;
 
-                       routeset(s, session[s].route[i].ip, session[s].route[i].mask, 0, 0);
+                       routeset(s, session[s].route[i].ip, session[s].route[i].prefixlen, 0, 0);
                }
 
                // ...ip
@@ -5293,11 +5323,11 @@ int load_session(sessionidt s, sessiont *new)
                // add new routes...
                for (i = 0; i < MAXROUTE && new->route[i].ip; i++)
                {
-                       if ((new->ip & new->route[i].mask) ==
-                           (new->route[i].ip & new->route[i].mask))
+                       if ((new->ip >> (32-new->route[i].prefixlen)) ==
+                           (new->route[i].ip >> (32-new->route[i].prefixlen)))
                                routed++;
 
-                       routeset(s, new->route[i].ip, new->route[i].mask, 0, 1);
+                       routeset(s, new->route[i].ip, new->route[i].prefixlen, 0, 1);
                }
 
                // ...ip