MLPPP not working in cluster mode.
[l2tpns.git] / l2tpns.c
index f47ae66..7478560 100644 (file)
--- a/l2tpns.c
+++ b/l2tpns.c
@@ -4,8 +4,6 @@
 // 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 1.173 2009/12/08 14:49:28 bodea Exp $";
-
 #include <arpa/inet.h>
 #include <assert.h>
 #include <errno.h>
@@ -14,7 +12,6 @@ char const *cvs_id_l2tpns = "$Id: l2tpns.c,v 1.173 2009/12/08 14:49:28 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 +28,7 @@ char const *cvs_id_l2tpns = "$Id: l2tpns.c,v 1.173 2009/12/08 14:49:28 bodea Exp
 #include <sys/time.h>
 #include <sys/resource.h>
 #include <sys/wait.h>
-#include <linux/if.h>
+#include <net/if.h>
 #include <stddef.h>
 #include <time.h>
 #include <dlfcn.h>
@@ -39,6 +36,8 @@ char const *cvs_id_l2tpns = "$Id: l2tpns.c,v 1.173 2009/12/08 14:49:28 bodea Exp
 #include <sched.h>
 #include <sys/sysinfo.h>
 #include <libcli.h>
+#include <linux/netlink.h>
+#include <linux/rtnetlink.h>
 
 #include "md5.h"
 #include "l2tpns.h"
@@ -56,6 +55,7 @@ char const *cvs_id_l2tpns = "$Id: l2tpns.c,v 1.173 2009/12/08 14:49:28 bodea Exp
 
 // Globals
 configt *config = NULL;                // all configuration
+int nlfd = -1;                 // netlink socket
 int tunfd = -1;                        // tun interface file handle. (network device)
 int udpfd = -1;                        // UDP file handle
 int controlfd = -1;            // Control signal handle
@@ -63,14 +63,14 @@ 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.
+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
@@ -102,6 +102,7 @@ uint32_t eth_tx = 0;
 
 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.
+uint64_t time_now_ms = 0;              // Current time in milliseconds since epoch.
 static char time_now_string[64] = {0}; // Current time as a string.
 static int time_changed = 0;           // time_now changed
 char main_quit = 0;                    // True if we're in the process of exiting.
@@ -157,6 +158,12 @@ 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),
+       CONFIG("echo_timeout", echo_timeout, INT),
+       CONFIG("idle_echo_timeout", idle_echo_timeout, INT),
        { NULL, 0, 0, 0 },
 };
 
@@ -164,8 +171,6 @@ 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",
@@ -197,6 +202,8 @@ 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);
@@ -236,6 +243,10 @@ static clockt now(double *f)
            time_now = t.tv_sec;
            time_changed++;
        }
+
+       // Time in milliseconds
+       time_now_ms = (t.tv_sec * 1000) + (t.tv_usec/1000);
+
        return (t.tv_sec - basetime) * 10 + t.tv_usec / 100000 + 1;
 }
 
@@ -416,43 +427,65 @@ 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 (!prefixlen) prefixlen = 32;
+
+       ip &= 0xffffffff << (32 - prefixlen);;  // Force the ip to be the first one in the route.
+
+       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;
+       }
 
-       if (!mask) mask = 0xffffffff;
+       req.nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.rt));
 
-       ip &= mask;             // Force the ip to be the first one in the route.
+       req.rt.rtm_family = AF_INET;
+       req.rt.rtm_dst_len = prefixlen;
+       req.rt.rtm_table = RT_TABLE_MAIN;
+       req.rt.rtm_protocol = 42;
+       req.rt.rtm_scope = RT_SCOPE_LINK;
+       req.rt.rtm_type = RTN_UNICAST;
 
-       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);
+       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, session[s].tunnel, "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.
@@ -468,40 +501,68 @@ 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 = 42;
+       req.rt.rtm_scope = RT_SCOPE_LINK;
+       req.rt.rtm_type = RTN_UNICAST;
 
-       LOG(1, 0, 0, "Route %s %s/%d\n",
+       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, s, session[s].tunnel, "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)
        {
@@ -514,12 +575,86 @@ void route6set(sessionidt s, struct in6_addr ip, int prefixlen, int add)
        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 netlink socket
+static void initnetlink(void)
+{
+       struct sockaddr_nl nladdr;
+
+       nlfd = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
+       if (nlfd < 0)
+       {
+               LOG(0, 0, 0, "Can't create netlink socket: %s\n", strerror(errno));
+               exit(1);
+       }
+
+       memset(&nladdr, 0, sizeof(nladdr));
+       nladdr.nl_family = AF_NETLINK;
+       nladdr.nl_pid = getpid();
+
+       if (bind(nlfd, (struct sockaddr *)&nladdr, sizeof(nladdr)) < 0)
+       {
+               LOG(0, 0, 0, "Can't bind netlink socket: %s\n", strerror(errno));
+               exit(1);
+       }
+}
+
+static ssize_t netlink_send(struct nlmsghdr *nh)
+{
+       struct sockaddr_nl nladdr;
+       struct iovec iov;
+       struct msghdr msg;
+
+       nh->nlmsg_pid = getpid();
+       nh->nlmsg_seq = ++nlseqnum;
+
+       // set kernel address
+       memset(&nladdr, 0, sizeof(nladdr));
+       nladdr.nl_family = AF_NETLINK;
+
+       iov = (struct iovec){ (void *)nh, nh->nlmsg_len };
+       msg = (struct msghdr){ (void *)&nladdr, sizeof(nladdr), &iov, 1, NULL, 0, 0 };
+
+       return sendmsg(nlfd, &msg, 0);
+}
+
+static ssize_t netlink_recv(void *buf, ssize_t len)
+{
+       struct sockaddr_nl nladdr;
+       struct iovec iov;
+       struct msghdr msg;
+
+       // set kernel address
+       memset(&nladdr, 0, sizeof(nladdr));
+       nladdr.nl_family = AF_NETLINK;
+
+       iov = (struct iovec){ buf, len };
+       msg = (struct msghdr){ (void *)&nladdr, sizeof(nladdr), &iov, 1, NULL, 0, 0 };
+
+       return recvmsg(nlfd, &msg, 0);
+}
+
+/* adapted from iproute2 */
+static void netlink_addattr(struct nlmsghdr *nh, int type, const void *data, int alen)
+{
+       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",
 };
 
 //
@@ -527,8 +662,7 @@ struct in6_ifreq {
 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;
 
@@ -547,75 +681,136 @@ 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);
-       ifrfd = socket(PF_INET, SOCK_DGRAM, IPPROTO_IP);
-
-       sin.sin_family = AF_INET;
-       sin.sin_addr.s_addr = config->bind_address ? config->bind_address : 0x01010101; // 1.1.1.1
-       memcpy(&ifr.ifr_addr, &sin, sizeof(struct sockaddr));
+       assert(strlen(ifr.ifr_name) < sizeof(config->tundevice) - 1);
+       strncpy(config->tundevice, ifr.ifr_name, sizeof(config->tundevice));
 
-       if (ioctl(ifrfd, SIOCSIFADDR, (void *) &ifr) < 0)
-       {
-               LOG(0, 0, 0, "Error setting tun address: %s\n", strerror(errno));
-               exit(1);
-       }
-       /* Bump up the qlen to deal with bursts from the network */
-       ifr.ifr_qlen = 1000;
-       if (ioctl(ifrfd, SIOCSIFTXQLEN, (void *) &ifr) < 0)
-       {
-               LOG(0, 0, 0, "Error setting tun queue length: %s\n", strerror(errno));
-               exit(1);
-       }
-       /* set MTU to modem MRU */
-       ifr.ifr_mtu = MRU;
-       if (ioctl(ifrfd, SIOCSIFMTU, (void *) &ifr) < 0)
+       tunidx = if_nametoindex(config->tundevice);
+       if (tunidx == 0)
        {
-               LOG(0, 0, 0, "Error setting tun MTU: %s\n", strerror(errno));
+               LOG(0, 0, 0, "Can't get tun interface index\n");
                exit(1);
        }
-       ifr.ifr_flags = IFF_UP;
-       if (ioctl(ifrfd, SIOCSIFFLAGS, (void *) &ifr) < 0)
-       {
-               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;
+               struct {
+                       // interface setting
+                       struct nlmsghdr nh;
+                       union {
+                               struct ifinfomsg ifinfo;
+                               struct ifaddrmsg ifaddr;
+                       } ifmsg;
+                       char rtdata[32]; // 32 should be enough
+               } req;
+               uint32_t txqlen, mtu;
+               in_addr_t ip;
 
-       // 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);
+               memset(&req, 0, sizeof(req));
 
-               // 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));
-               }
+               req.nh.nlmsg_type = RTM_NEWLINK;
+               req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_MULTI;
+               req.nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.ifmsg.ifinfo));
 
-               // 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));
+               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
+
+               /* Bump up the qlen to deal with bursts from the network */
+               txqlen = 1000;
+               netlink_addattr(&req.nh, IFLA_TXQLEN, &txqlen, sizeof(txqlen));
+               /* set MTU to modem MRU */
+               mtu = MRU;
+               netlink_addattr(&req.nh, IFLA_MTU, &mtu, sizeof(mtu));
+
+               if (netlink_send(&req.nh) < 0)
+                       goto senderror;
+
+               memset(&req, 0, sizeof(req));
+
+               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_INET;
+               req.ifmsg.ifaddr.ifa_prefixlen = 32;
+               req.ifmsg.ifaddr.ifa_scope = RT_SCOPE_UNIVERSE;
+               req.ifmsg.ifaddr.ifa_index = tunidx;
+
+               if (config->bind_address)
+                       ip = config->bind_address;
+               else
+                       ip = 0x01010101; // 1.1.1.1
+               netlink_addattr(&req.nh, IFA_LOCAL, &ip, sizeof(ip));
+
+               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]) {
+                       struct in6_addr ip6;
+
+                       memset(&req, 0, sizeof(req));
+
+                       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;
+
+                       // Link local address is FE80::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));
+
+                       if (netlink_send(&req.nh) < 0)
+                               goto senderror;
+
+                       memset(&req, 0, sizeof(req));
+
+                       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_UNIVERSE;
+                       req.ifmsg.ifaddr.ifa_index = tunidx;
+
+                       // Global address is prefix::1
+                       ip6 = config->ipv6_prefix;
+                       ip6.s6_addr[15] = 1;
+                       netlink_addattr(&req.nh, IFA_LOCAL, &ip6, sizeof(ip6));
+
+                       if (netlink_send(&req.nh) < 0)
+                               goto senderror;
                }
+
+               memset(&req, 0, sizeof(req));
+
+               req.nh.nlmsg_type = NLMSG_DONE;
+               req.nh.nlmsg_len = NLMSG_LENGTH(0);
+
+               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
@@ -1135,7 +1330,8 @@ static void update_session_out_stat(sessionidt s, sessiont *sp, int len)
 
 // process outgoing (to tunnel) IP
 //
-static void processipout(uint8_t *buf, int len)
+// (i.e. this routine writes to data[-8]).
+void processipout(uint8_t *buf, int len)
 {
        sessionidt s;
        sessiont *sp;
@@ -1272,75 +1468,123 @@ static void processipout(uint8_t *buf, int len)
                return;
        }
 
-       // Add on L2TP header
-        {
-                bundleidt bid = 0;
-                if(session[s].bundle != 0 && bundle[session[s].bundle].num_of_links > 1)
-                {
-                        bid = session[s].bundle;
-                        s = bundle[bid].members[bundle[bid].current_ses = ++bundle[bid].current_ses % bundle[bid].num_of_links];
-                       t = session[s].tunnel;
-                       sp = &session[s];
-                        LOG(4, s, t, "MPPP: (1)Session number becomes: %d\n", s);
-                        if(len > MINFRAGLEN)
-                        {
-                                // Partition the packet to "bundle[b].num_of_links" fragments
-                               bundlet *b = &bundle[bid];
-                               uint32_t num_of_links = b->num_of_links;
-                                uint32_t fraglen = len / num_of_links;
-                               fraglen = (fraglen > session[s].mru ? session[s].mru : fraglen);
-                               uint32_t last_fraglen = fraglen + len % num_of_links;
-                               last_fraglen = (last_fraglen > session[s].mru ? len % num_of_links : last_fraglen);
+       if(session[s].bundle != 0 && bundle[session[s].bundle].num_of_links > 1)
+       {
+
+               if (!config->cluster_iam_master)
+               {
+                       // The MPPP packets must be managed by the Master.
+                       master_forward_mppp_packet(s, data, size);
+                       return;
+               }
+
+               // Add on L2TP header
+               sessionidt members[MAXBUNDLESES];
+               bundleidt bid = session[s].bundle;
+               bundlet *b = &bundle[bid];
+               uint32_t num_of_links, nb_opened;
+               int i;
+
+               num_of_links = b->num_of_links;
+               nb_opened = 0;
+               for (i = 0;i < num_of_links;i++)
+               {
+                       s = b->members[i];
+                       if (session[s].ppp.lcp == Opened)
+                       {
+                               members[nb_opened] = s;
+                               nb_opened++;
+                       }
+               }
+
+               if (nb_opened < 1)
+               {
+                       LOG(3, s, t, "MPPP: PROCESSIPOUT ERROR, no session opened in bundle:%d\n", bid);
+                       return;
+               }
+
+               num_of_links = nb_opened;
+               b->current_ses = (b->current_ses + 1) % num_of_links;
+               s = members[b->current_ses];
+               t = session[s].tunnel;
+               sp = &session[s];
+               LOG(4, s, t, "MPPP: (1)Session number becomes: %d\n", s);
+
+               if (num_of_links > 1)
+               {
+                       if(len > MINFRAGLEN)
+                       {
+                               //for rotate traffic among the member links
+                               uint32_t divisor = num_of_links;
+                               if (divisor > 2)
+                                       divisor = divisor/2 + (divisor & 1);
+
+                               // Partition the packet to "num_of_links" fragments
+                               uint32_t fraglen = len / divisor;
+                               uint32_t last_fraglen = fraglen + len % divisor;
                                uint32_t remain = len;
 
                                // send the first packet
-                                uint8_t *p = makeppp(fragbuf, sizeof(fragbuf), buf, fraglen, s, t, PPPIP, 0, bid, MP_BEGIN);
-                                if (!p) return;
-                                tunnelsend(fragbuf, fraglen + (p-fragbuf), t); // send it...
+                               uint8_t *p = makeppp(fragbuf, sizeof(fragbuf), buf, fraglen, s, t, PPPIP, 0, bid, MP_BEGIN);
+                               if (!p) return;
+                               tunnelsend(fragbuf, fraglen + (p-fragbuf), t); // send it...
+
                                // statistics
                                update_session_out_stat(s, sp, fraglen);
+
                                remain -= fraglen;
                                while (remain > last_fraglen)
-                               { 
-                                       s = b->members[b->current_ses = ++b->current_ses % num_of_links];
+                               {
+                                       b->current_ses = (b->current_ses + 1) % num_of_links;
+                                       s = members[b->current_ses];
                                        t = session[s].tunnel;
                                        sp = &session[s];
-                                       LOG(4, s, t, "MPPP: (2)Session number becomes: %d\n", s);
-                                       p = makeppp(fragbuf, sizeof(fragbuf), buf+(len - remain), fraglen, s, t, PPPIP, 0, bid, 0);
-                                       if (!p) return;
-                                       tunnelsend(fragbuf, fraglen + (p-fragbuf), t); // send it...
+                                       LOG(4, s, t, "MPPP: (2)Session number becomes: %d\n", s);
+                                       p = makeppp(fragbuf, sizeof(fragbuf), buf+(len - remain), fraglen, s, t, PPPIP, 0, bid, 0);
+                                       if (!p) return;
+                                       tunnelsend(fragbuf, fraglen + (p-fragbuf), t); // send it...
                                        update_session_out_stat(s, sp, fraglen);
                                        remain -= fraglen;
                                }
                                // send the last fragment
-                               s = b->members[b->current_ses = ++b->current_ses % num_of_links];
+                               b->current_ses = (b->current_ses + 1) % num_of_links;
+                               s = members[b->current_ses];
                                t = session[s].tunnel;
                                sp = &session[s];
-                                       LOG(4, s, t, "MPPP: (2)Session number becomes: %d\n", s);
-                                       p = makeppp(fragbuf, sizeof(fragbuf), buf+(len - remain), remain, s, t, PPPIP, 0, bid, MP_END);
-                                       if (!p) return;
-                                       tunnelsend(fragbuf, remain + (p-fragbuf), t); // send it...
+                               LOG(4, s, t, "MPPP: (2)Session number becomes: %d\n", s);
+                               p = makeppp(fragbuf, sizeof(fragbuf), buf+(len - remain), remain, s, t, PPPIP, 0, bid, MP_END);
+                               if (!p) return;
+                               tunnelsend(fragbuf, remain + (p-fragbuf), t); // send it...
                                update_session_out_stat(s, sp, remain);
                                if (remain != last_fraglen)
                                        LOG(3, s, t, "PROCESSIPOUT ERROR REMAIN != LAST_FRAGLEN, %d != %d\n", remain, last_fraglen);
-                        }
-                        else {
-                                // Send it as one frame
-                                uint8_t *p = makeppp(fragbuf, sizeof(fragbuf), buf, len, s, t, PPPIP, 0, bid, MP_BOTH_BITS);
-                                if (!p) return;
-                                tunnelsend(fragbuf, len + (p-fragbuf), t); // send it...
+                       }
+                       else
+                       {
+                               // Send it as one frame
+                               uint8_t *p = makeppp(fragbuf, sizeof(fragbuf), buf, len, s, t, PPPIP, 0, bid, MP_BOTH_BITS);
+                               if (!p) return;
+                               tunnelsend(fragbuf, len + (p-fragbuf), t); // send it...
                                LOG(4, s, t, "MPPP: packet sent as one frame\n");
                                update_session_out_stat(s, sp, len);
-                        }
-                }
-                else
-                {
-                        uint8_t *p = makeppp(fragbuf, sizeof(fragbuf), buf, len, s, t, PPPIP, 0, 0, 0);
-                        if (!p) return;
-                        tunnelsend(fragbuf, len + (p-fragbuf), t); // send it...
+                       }
+               }
+               else
+               {
+                       // Send it as one frame (NO MPPP Frame)
+                       uint8_t *p = makeppp(fragbuf, sizeof(fragbuf), buf, len, s, t, PPPIP, 0, 0, 0);
+                       if (!p) return;
+                       tunnelsend(fragbuf, len + (p-fragbuf), t); // send it...
                        update_session_out_stat(s, sp, len);
-                }
-        }
+               }
+       }
+       else
+       {
+               uint8_t *p = makeppp(fragbuf, sizeof(fragbuf), buf, len, s, t, PPPIP, 0, 0, 0);
+               if (!p) return;
+               tunnelsend(fragbuf, len + (p-fragbuf), t); // send it...
+               update_session_out_stat(s, sp, len);
+       }
 
        // Snooping this session, send it to intercept box
        if (sp->snoop_ip && sp->snoop_port)
@@ -1420,7 +1664,10 @@ static void processipv6out(uint8_t * buf, int len)
        if (session[s].bundle && bundle[session[s].bundle].num_of_links > 1)
        {
                bundleidt bid = session[s].bundle;
-               s = bundle[bid].members[bundle[bid].current_ses = ++bundle[bid].current_ses % bundle[bid].num_of_links];
+               bundlet *b = &bundle[bid];
+
+               b->current_ses = (b->current_ses + 1) % b->num_of_links;
+               s = b->members[b->current_ses];
                LOG(3, s, session[s].tunnel, "MPPP: Session number becomes: %u\n", s);
        }
        t = session[s].tunnel;
@@ -1475,7 +1722,6 @@ static void send_ipout(sessionidt s, uint8_t *buf, int len)
 {
        sessiont *sp;
        tunnelidt t;
-       in_addr_t ip;
 
        uint8_t b[MAXETHER + 20];
 
@@ -1489,8 +1735,6 @@ static void send_ipout(sessionidt s, uint8_t *buf, int len)
        buf += 4;
        len -= 4;
 
-       ip = *(in_addr_t *)(buf + 16);
-
        if (!session[s].ip)
                return;
 
@@ -1523,10 +1767,10 @@ static void send_ipout(sessionidt s, uint8_t *buf, int len)
 static void control16(controlt * c, uint16_t avp, uint16_t val, uint8_t m)
 {
        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->buf16[c->length/2 + 0] = htons(l);
+       c->buf16[c->length/2 + 1] = htons(0);
+       c->buf16[c->length/2 + 2] = htons(avp);
+       c->buf16[c->length/2 + 3] = htons(val);
        c->length += 8;
 }
 
@@ -1534,10 +1778,10 @@ static void control16(controlt * c, uint16_t avp, uint16_t val, uint8_t m)
 static void control32(controlt * c, uint16_t avp, uint32_t val, uint8_t m)
 {
        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->buf16[c->length/2 + 0] = htons(l);
+       c->buf16[c->length/2 + 1] = htons(0);
+       c->buf16[c->length/2 + 2] = htons(avp);
+       *(uint32_t *) &c->buf[c->length + 6] = htonl(val);
        c->length += 10;
 }
 
@@ -1545,10 +1789,10 @@ static void control32(controlt * c, uint16_t avp, uint32_t val, uint8_t m)
 static void controls(controlt * c, uint16_t avp, char *val, uint8_t m)
 {
        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->buf16[c->length/2 + 0] = htons(l);
+       c->buf16[c->length/2 + 1] = htons(0);
+       c->buf16[c->length/2 + 2] = htons(avp);
+       memcpy(&c->buf[c->length + 6], val, strlen(val));
        c->length += 6 + strlen(val);
 }
 
@@ -1556,10 +1800,10 @@ static void controls(controlt * c, uint16_t avp, char *val, uint8_t m)
 static void controlb(controlt * c, uint16_t avp, uint8_t *val, unsigned int len, uint8_t m)
 {
        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->buf16[c->length/2 + 0] = htons(l);
+       c->buf16[c->length/2 + 1] = htons(0);
+       c->buf16[c->length/2 + 2] = htons(avp);
+       memcpy(&c->buf[c->length + 6], val, len);
        c->length += 6 + len;
 }
 
@@ -1576,7 +1820,7 @@ static controlt *controlnew(uint16_t mtype)
        }
        assert(c);
        c->next = 0;
-       *(uint16_t *) (c->buf + 0) = htons(0xC802); // flags/ver
+       c->buf16[0] = htons(0xC802); // flags/ver
        c->length = 12;
        control16(c, 0, mtype, 1);
        return c;
@@ -1586,26 +1830,26 @@ static controlt *controlnew(uint16_t mtype)
 // (ZLB send).
 static void controlnull(tunnelidt t)
 {
-       uint8_t buf[12];
+       uint16_t buf[6];
        if (tunnel[t].controlc) // Messages queued; They will carry the ack.
                return;
 
-       *(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);
+       buf[0] = htons(0xC802); // flags/ver
+       buf[1] = htons(12); // length
+       buf[2] = htons(tunnel[t].far); // tunnel
+       buf[3] = htons(0); // session
+       buf[4] = htons(tunnel[t].ns); // sequence
+       buf[5] = htons(tunnel[t].nr); // sequence
+       tunnelsend((uint8_t *)buf, 12, t);
 }
 
 // add a control message to a tunnel, and send if within window
 static void controladd(controlt *c, sessionidt far, tunnelidt t)
 {
-       *(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
+       c->buf16[1] = htons(c->length); // length
+       c->buf16[2] = htons(tunnel[t].far); // tunnel
+       c->buf16[3] = htons(far); // session
+       c->buf16[4] = htons(tunnel[t].ns); // sequence
        tunnel[t].ns++;              // advance sequence
        // link in message in to queue
        if (tunnel[t].controlc)
@@ -1730,6 +1974,7 @@ void sessionshutdown(sessionidt s, char const *reason, int cdn_result, int cdn_e
                struct param_kill_session data = { &tunnel[session[s].tunnel], &session[s] };
                LOG(2, s, session[s].tunnel, "Shutting down session %u: %s\n", s, reason);
                run_plugins(PLUGIN_KILL_SESSION, &data);
+               session[s].die = TIME + 150; // Clean up in 15 seconds
        }
 
        if (session[s].ip && !walled_garden && !session[s].die)
@@ -1760,11 +2005,11 @@ 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;
                }
 
@@ -1780,36 +2025,67 @@ void sessionshutdown(sessionidt s, char const *reason, int cdn_result, int cdn_e
                if (session[s].ppp.ipv6cp == Opened && session[s].ipv6prefixlen && del_routes)
                        route6set(s, session[s].ipv6route, session[s].ipv6prefixlen, 0);
                
-               if (b) 
+               if (b)
                {
-                       // This session was part of a bundle
-                       bundle[b].num_of_links--;
-                       LOG(3, s, 0, "MPPP: Dropping member link: %d from bundle %d\n",s,b);
-                       if(bundle[b].num_of_links == 0) 
+                       // This session was part of a bundle
+                       bundle[b].num_of_links--;
+                       LOG(3, s, session[s].tunnel, "MPPP: Dropping member link: %d from bundle %d\n",s,b);
+                       if(bundle[b].num_of_links == 0)
                        {
-                               bundleclear(b);
-                               LOG(3, s, 0, "MPPP: Kill bundle: %d (No remaing member links)\n",b);
-                       }
-                       else 
+                               bundleclear(b);
+                               LOG(3, s, session[s].tunnel, "MPPP: Kill bundle: %d (No remaing member links)\n",b);
+                       }
+                       else 
                        {
-                               // Adjust the members array to accomodate the new change
-                               uint8_t mem_num = 0;
-                               // It should be here num_of_links instead of num_of_links-1 (previous instruction "num_of_links--")
-                               if(bundle[b].members[bundle[b].num_of_links] != s) 
+                               // Adjust the members array to accomodate the new change
+                               uint8_t mem_num = 0;
+                               // It should be here num_of_links instead of num_of_links-1 (previous instruction "num_of_links--")
+                               if(bundle[b].members[bundle[b].num_of_links] != s)
                                {
-                                       uint8_t ml;
-                                       for(ml = 0; ml<bundle[b].num_of_links; ml++)
-                                               if(bundle[b].members[ml] == s)
-                                               {
-                                                       mem_num = ml;
-                                                       break;
-                                               }
-                                       bundle[b].members[mem_num] = bundle[b].members[bundle[b].num_of_links];
-                                       LOG(3, s, 0, "MPPP: Adjusted member links array\n");
-                               }
-                       }
-                       cluster_send_bundle(b);
-               }
+                                       uint8_t ml;
+                                       for(ml = 0; ml<bundle[b].num_of_links; ml++)
+                                       if(bundle[b].members[ml] == s)
+                                       {
+                                                       mem_num = ml;
+                                                       break;
+                                       }
+                                       bundle[b].members[mem_num] = bundle[b].members[bundle[b].num_of_links];
+                                       LOG(3, s, session[s].tunnel, "MPPP: Adjusted member links array\n");
+
+                                       // If the killed session is the first of the bundle,
+                                       // the new first session must be stored in the cache_ipmap
+                                       // else the function sessionbyip return 0 and the sending not work any more (processipout).
+                                       if (mem_num == 0)
+                                       {
+                                               sessionidt new_s = bundle[b].members[0];
+
+                                               routed = 0;
+                                               // Add the route for this session.
+                                               for (r = 0; r < MAXROUTE && session[new_s].route[r].ip; r++)
+                                               {
+                                                       int i, prefixlen;
+                                                       in_addr_t ip;
+
+                                                       prefixlen = session[new_s].route[r].prefixlen;
+                                                       ip = session[new_s].route[r].ip;
+
+                                                       if (!prefixlen) prefixlen = 32;
+                                                       ip &= 0xffffffff << (32 - prefixlen);   // Force the ip to be the first one in the route.
+
+                                                       for (i = ip; i < ip+(1<<(32-prefixlen)) ; ++i)
+                                                               cache_ipmap(i, new_s);
+                                               }
+                                               cache_ipmap(session[new_s].ip, new_s);
+
+                                               // IPV6 route
+                                               if (session[new_s].ipv6prefixlen)
+                                                       cache_ipv6map(session[new_s].ipv6route, session[new_s].ipv6prefixlen, new_s);
+                                       }
+                               }
+                       }
+
+                       cluster_send_bundle(b);
+        }
        }
 
        if (session[s].throttle_in || session[s].throttle_out) // Unthrottle if throttled.
@@ -1820,10 +2096,10 @@ void sessionshutdown(sessionidt s, char const *reason, int cdn_result, int cdn_e
                controlt *c = controlnew(14); // sending CDN
                if (cdn_error)
                {
-                       uint8_t buf[4];
-                       *(uint16_t *) buf     = htons(cdn_result);
-                       *(uint16_t *) (buf+2) = htons(cdn_error);
-                       controlb(c, 1, buf, 4, 1);
+                       uint16_t buf[2];
+                       buf[0] = htons(cdn_result);
+                       buf[1] = htons(cdn_error);
+                       controlb(c, 1, (uint8_t *)buf, 4, 1);
                }
                else
                        control16(c, 1, cdn_result, 1);
@@ -1832,9 +2108,6 @@ void sessionshutdown(sessionidt s, char const *reason, int cdn_result, int cdn_e
                controladd(c, session[s].far, session[s].tunnel); // send the message
        }
 
-       if (!session[s].die)
-               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--;
@@ -1930,8 +2203,9 @@ void sessionkill(sessionidt s, char *reason)
                return;
        }
 
-       session[s].die = TIME;
-       sessionshutdown(s, reason, CDN_ADMIN_DISC, TERM_ADMIN_RESET);  // close radius/routes, etc.
+       if (!session[s].die)
+               sessionshutdown(s, reason, CDN_ADMIN_DISC, TERM_ADMIN_RESET);  // close radius/routes, etc.
+
        if (sess_local[s].radius)
                radiusclear(sess_local[s].radius, s); // cant send clean accounting data, session is killed
 
@@ -2014,21 +2288,21 @@ static void tunnelshutdown(tunnelidt t, char *reason, int result, int error, cha
                controlt *c = controlnew(4);    // sending StopCCN
                if (error)
                {
-                       uint8_t buf[64];
+                       uint16_t buf[32];
                        int l = 4;
-                       *(uint16_t *) buf     = htons(result);
-                       *(uint16_t *) (buf+2) = htons(error);
+                       buf[0] = htons(result);
+                       buf[1] = htons(error);
                        if (msg)
                        {
                                int m = strlen(msg);
                                if (m + 4 > sizeof(buf))
                                    m = sizeof(buf) - 4;
 
-                               memcpy(buf+4, msg, m);
+                               memcpy(buf+2, msg, m);
                                l += m;
                        }
 
-                       controlb(c, 1, buf, l, 1);
+                       controlb(c, 1, (uint8_t *)buf, l, 1);
                }
                else
                        control16(c, 1, result, 1);
@@ -2350,7 +2624,7 @@ void processudp(uint8_t *buf, int len, struct sockaddr_in *addr)
                                case 0:     // message type
                                        message = ntohs(*(uint16_t *) b);
                                        mandatory = flags & 0x80;
-                                       LOG(4, s, t, "   Message type = %u (%s)\n", *b, l2tp_code(message));
+                                       LOG(4, s, t, "   Message type = %u (%s)\n", message, l2tp_code(message));
                                        break;
                                case 1:     // result code
                                        {
@@ -2687,9 +2961,11 @@ void processudp(uint8_t *buf, int len, struct sockaddr_in *addr)
                                        break;
                                case 2:       // SCCRP
                                        tunnel[t].state = TUNNELOPEN;
+                                       tunnel[t].lastrec = time_now;
                                        break;
                                case 3:       // SCCN
                                        tunnel[t].state = TUNNELOPEN;
+                                       tunnel[t].lastrec = time_now;
                                        controlnull(t); // ack
                                        break;
                                case 4:       // StopCCN
@@ -2768,7 +3044,7 @@ 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);
@@ -2902,8 +3178,9 @@ void processudp(uint8_t *buf, int len, struct sockaddr_in *addr)
                        }
 
                        session[s].last_packet = session[s].last_data = time_now;
-                       if (session[s].walled_garden && !config->cluster_iam_master)
+                       if (!config->cluster_iam_master)
                        {
+                               // The fragments reconstruction is managed by the Master.
                                master_forward_packet(buf, len, addr->sin_addr.s_addr, addr->sin_port);
                                return;
                        }
@@ -2942,6 +3219,7 @@ void processudp(uint8_t *buf, int len, struct sockaddr_in *addr)
 }
 
 // read and process packet on tun
+// (i.e. this routine writes to buf[-8]).
 static void processtun(uint8_t * buf, int len)
 {
        LOG_HEX(5, "Receive TUN Data", buf, len);
@@ -3032,7 +3310,7 @@ static void regular_cleanups(double period)
                        if (tunnel[t].retry <= TIME)
                        {
                                controlt *c = tunnel[t].controls;
-                               uint8_t w = tunnel[t].window;
+                               uint16_t w = tunnel[t].window;
                                tunnel[t].try++; // another try
                                if (tunnel[t].try > 5)
                                        tunnelkill(t, "Timeout on control message"); // game over
@@ -3215,8 +3493,8 @@ static void regular_cleanups(double period)
                        }
                }
 
-               // Drop sessions who have not responded within IDLE_TIMEOUT seconds
-               if (session[s].last_packet && (time_now - session[s].last_packet >= IDLE_TIMEOUT))
+               // Drop sessions who have not responded within IDLE_ECHO_TIMEOUT seconds
+               if (session[s].last_packet && (time_now - session[s].last_packet >= config->idle_echo_timeout))
                {
                        sessionshutdown(s, "No response to LCP ECHO requests.", CDN_ADMIN_DISC, TERM_LOST_SERVICE);
                        STAT(session_timeout);
@@ -3225,7 +3503,7 @@ static void regular_cleanups(double period)
                }
 
                // No data in ECHO_TIMEOUT seconds, send LCP ECHO
-               if (session[s].ppp.phase >= Establish && (time_now - session[s].last_packet >= ECHO_TIMEOUT) &&
+               if (session[s].ppp.phase >= Establish && (time_now - session[s].last_packet >= config->echo_timeout) &&
                        (time_now - sess_local[s].last_echo >= ECHO_TIMEOUT))
                {
                        uint8_t b[MAXETHER];
@@ -3396,35 +3674,36 @@ 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)
        {
@@ -3490,8 +3769,8 @@ 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
@@ -3505,6 +3784,8 @@ static void mainloop(void)
 {
        int i;
        uint8_t buf[65536];
+       uint8_t *p = buf + 8; // for the hearder of the forwarded MPPP packet (see C_MPPP_FORWARD)
+       int size_bufp = sizeof(buf) - 8;
        clockt next_cluster_ping = 0;   // send initial ping immediately
        struct epoll_event events[BASE_FDS + RADIUS_FDS + EXTRA_FDS];
        int maxevent = sizeof(events)/sizeof(*events);
@@ -3515,8 +3796,8 @@ 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 */
        {
@@ -3552,6 +3833,10 @@ 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
@@ -3565,7 +3850,8 @@ 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 */
 
@@ -3688,6 +3974,32 @@ 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);
                                }
@@ -3718,9 +4030,9 @@ static void mainloop(void)
                                // incoming IP
                                if (tun_ready)
                                {
-                                       if ((s = read(tunfd, buf, sizeof(buf))) > 0)
+                                       if ((s = read(tunfd, p, size_bufp)) > 0)
                                        {
-                                               processtun(buf, s);
+                                               processtun(p, s);
                                                tun_pkts++;
                                        }
                                        else
@@ -3759,6 +4071,11 @@ 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)
                {
@@ -3964,6 +4281,9 @@ static void initdata(int optdebug, char *optconfig)
        config->ppp_max_failure = 5;
        config->kill_timedout_sessions = 1;
        strcpy(config->random_device, RANDOMDEVICE);
+       // Set default value echo_timeout and idle_echo_timeout
+       config->echo_timeout = ECHO_TIMEOUT;
+       config->idle_echo_timeout = IDLE_ECHO_TIMEOUT;
 
        log_stream = stderr;
 
@@ -4246,18 +4566,18 @@ 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.
@@ -4315,7 +4635,7 @@ 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;
@@ -4325,15 +4645,14 @@ 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
                {
@@ -4484,8 +4803,12 @@ int main(int argc, char *argv[])
        initplugins();
        initdata(optdebug, optconfig);
 
-       init_cli(hostname);
+       init_cli();
        read_config_file();
+       /* set hostname /after/ having read the config file */
+       if (*config->hostname)
+               strcpy(hostname, config->hostname);
+       cli_init_complete(hostname);
        update_config();
        init_tbf(config->num_tbfs);
 
@@ -4529,6 +4852,8 @@ int main(int argc, char *argv[])
                }
        }
 
+       initnetlink();
+
        /* Set up the cluster communications port. */
        if (cluster_init() < 0)
                exit(1);
@@ -4980,11 +5305,11 @@ 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
@@ -5055,7 +5380,7 @@ 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
@@ -5066,11 +5391,11 @@ 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
@@ -5089,11 +5414,11 @@ 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