Fix the inverted "delete/add" of the routes, in cluster mode.
[l2tpns.git] / l2tpns.c
index 110029a..ed85a91 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 1.176 2011/01/20 12:48:40 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 1.176 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 +28,7 @@ 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>
@@ -70,7 +67,7 @@ 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
@@ -105,6 +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.
@@ -160,6 +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 },
 };
 
@@ -167,8 +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",
@@ -241,6 +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;
 }
 
@@ -445,13 +451,17 @@ static void routeset(sessionidt s, in_addr_t ip, int prefixlen, in_addr_t gw, in
                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_protocol = 42;
        req.rt.rtm_scope = RT_SCOPE_LINK;
        req.rt.rtm_type = RTN_UNICAST;
 
@@ -464,7 +474,7 @@ static void routeset(sessionidt s, in_addr_t ip, int prefixlen, in_addr_t gw, in
                netlink_addattr(&req.nh, RTA_GATEWAY, &n_ip, sizeof(n_ip));
        }
 
-       LOG(1, s, 0, "Route %s %s/%d%s%s\n", add ? "add" : "del",
+       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) : "");
 
@@ -520,13 +530,17 @@ void route6set(sessionidt s, struct in6_addr ip, int prefixlen, int add)
                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_protocol = 42;
        req.rt.rtm_scope = RT_SCOPE_LINK;
        req.rt.rtm_type = RTN_UNICAST;
 
@@ -535,7 +549,7 @@ void route6set(sessionidt s, struct in6_addr ip, int prefixlen, int add)
        metric = 1;
        netlink_addattr(&req.nh, RTA_METRICS, &metric, sizeof(metric));
 
-       LOG(1, 0, 0, "Route %s %s/%d\n",
+       LOG(1, s, session[s].tunnel, "Route %s %s/%d\n",
            add ? "add" : "del",
            inet_ntop(AF_INET6, &ip, ipv6addr, INET6_ADDRSTRLEN),
            prefixlen);
@@ -543,7 +557,12 @@ void route6set(sessionidt s, struct in6_addr ip, int prefixlen, int add)
        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)
        {
@@ -642,7 +661,6 @@ static char *tun_nl_phase_msg[] = {
 // Set up TUN interface
 static void inittun(void)
 {
-       struct ifinfomsg ifinfo;
        struct ifreq ifr;
 
        memset(&ifr, 0, sizeof(ifr));
@@ -666,41 +684,13 @@ static void inittun(void)
        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
-               struct {
-                       struct nlmsghdr nh;
-                       struct ifinfomsg ifinfo;
-               } req;
-               char buf[4096];
-               ssize_t len;
-               struct nlmsghdr *resp_nh;
-
-               req.nh.nlmsg_type = RTM_GETLINK;
-               req.nh.nlmsg_flags = NLM_F_REQUEST;
-               req.nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.ifinfo));
-
-               req.ifinfo.ifi_family = AF_UNSPEC; // as the man says
-
-               netlink_addattr(&req.nh, IFLA_IFNAME, config->tundevice, strlen(config->tundevice)+1);
-
-               if(netlink_send(&req.nh) < 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))
-               {
-                       LOG(0, 0, 0, "Malformed answer getting tun ifindex %ld\n", len);
-                       exit(1);
-               }
-
-               memcpy(&ifinfo, NLMSG_DATA(resp_nh), sizeof(ifinfo));
-               // got index
-               tunidx = ifinfo.ifi_index;
+               LOG(0, 0, 0, "Can't get tun interface index\n");
+               exit(1);
        }
+
        {
                struct {
                        // interface setting
@@ -712,15 +702,16 @@ static void inittun(void)
                        char rtdata[32]; // 32 should be enough
                } req;
                uint32_t txqlen, mtu;
-               struct in_addr ip;
+               in_addr_t ip;
 
                memset(&req, 0, sizeof(req));
 
-               req.nh.nlmsg_type = RTM_SETLINK;
+               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));
 
-               req.ifmsg.ifinfo = ifinfo;
+               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
 
@@ -737,17 +728,18 @@ static void inittun(void)
                memset(&req, 0, sizeof(req));
 
                req.nh.nlmsg_type = RTM_NEWADDR;
-               req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_MULTI;
+               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_index = ifinfo.ifi_index;
+               req.ifmsg.ifaddr.ifa_scope = RT_SCOPE_UNIVERSE;
+               req.ifmsg.ifaddr.ifa_index = tunidx;
 
                if (config->bind_address)
-                       ip.s_addr = config->bind_address;
+                       ip = config->bind_address;
                else
-                       ip.s_addr = 0x01010101, // 1.1.1.1
+                       ip = 0x01010101; // 1.1.1.1
                netlink_addattr(&req.nh, IFA_LOCAL, &ip, sizeof(ip));
 
                if (netlink_send(&req.nh) < 0)
@@ -760,13 +752,13 @@ static void inittun(void)
                        memset(&req, 0, sizeof(req));
 
                        req.nh.nlmsg_type = RTM_NEWADDR;
-                       req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_MULTI;
+                       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 = ifinfo.ifi_index;
+                       req.ifmsg.ifaddr.ifa_index = tunidx;
 
                        // Link local address is FE80::1
                        memset(&ip6, 0, sizeof(ip6));
@@ -781,13 +773,13 @@ static void inittun(void)
                        memset(&req, 0, sizeof(req));
 
                        req.nh.nlmsg_type = RTM_NEWADDR;
-                       req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_MULTI;
+                       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 = ifinfo.ifi_index;
+                       req.ifmsg.ifaddr.ifa_index = tunidx;
 
                        // Global address is prefix::1
                        ip6 = config->ipv6_prefix;
@@ -1338,7 +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;
@@ -1475,75 +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)
@@ -1623,7 +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;
@@ -1678,7 +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];
 
@@ -1692,8 +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;
 
@@ -1726,10 +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;
 }
 
@@ -1737,10 +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;
 }
 
@@ -1748,10 +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);
 }
 
@@ -1759,10 +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;
 }
 
@@ -1779,7 +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;
@@ -1789,26 +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)
@@ -1984,36 +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.
@@ -2024,10 +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);
@@ -2216,21 +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);
@@ -2552,7 +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
                                        {
@@ -2972,7 +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);
@@ -3106,8 +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;
                        }
@@ -3146,6 +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);
@@ -3419,8 +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);
@@ -3429,7 +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];
@@ -3600,35 +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)
        {
@@ -3709,6 +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);
@@ -3773,7 +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 */
 
@@ -3952,9 +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
@@ -3993,6 +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)
                {
@@ -4198,6 +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;
 
@@ -4717,8 +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);
 
@@ -5137,18 +5227,18 @@ int sessionsetup(sessionidt s, tunnelidt t)
        LOG(3, s, t, "Doing session setup for session\n");
 
        // Join a bundle if the MRRU option is accepted
-        if(session[s].mrru > 0 && session[s].bundle == 0)
-        {
-                LOG(3, s, t, "This session can be part of multilink bundle\n");
-                if (join_bundle(s) > 0)
-                       cluster_send_bundle(session[s].bundle);
+       if(session[s].mrru > 0 && session[s].bundle == 0)
+       {
+               LOG(3, s, t, "This session can be part of multilink bundle\n");
+               if (join_bundle(s) > 0)
+                       cluster_send_bundle(session[s].bundle);
                else
                {
                        LOG(0, s, t, "MPPP: Mismaching mssf option with other sessions in bundle\n");
                        sessionshutdown(s, "Mismaching mssf option.", CDN_NONE, TERM_SERVICE_UNAVAILABLE);
                        return 0;
                }
-        }
+       }
 
        if (!session[s].ip)
        {
@@ -5163,7 +5253,6 @@ int sessionsetup(sessionidt s, tunnelidt t)
                        fmtaddr(htonl(session[s].ip), 0));
        }
 
-
        // Make sure this is right
        session[s].tunnel = t;
 
@@ -5176,13 +5265,14 @@ int sessionsetup(sessionidt s, tunnelidt t)
                for (i = 1; i <= config->cluster_highest_sessionid; i++)
                {
                        if (i == s) continue;
-                       if (!session[s].opened) continue;
+                       if (!session[s].opened) break;
                        // Allow duplicate sessions for multilink ones of the same bundle.
-                        if (session[s].bundle && session[i].bundle && session[s].bundle == session[i].bundle)
-                                continue;
+                       if (session[s].bundle && session[i].bundle && session[s].bundle == session[i].bundle) continue;
+
                        if (ip == session[i].ip)
                        {
                                sessionkill(i, "Duplicate IP address");
+                               cluster_listinvert_session(s, i);
                                continue;
                        }
 
@@ -5190,16 +5280,16 @@ int sessionsetup(sessionidt s, tunnelidt t)
                        if (session[s].walled_garden || session[i].walled_garden) continue;
                        // Guest change
                        int found = 0;
-                        int gu;
-                        for (gu = 0; gu < guest_accounts_num; gu++)
-                        {
-                                if (!strcasecmp(user, guest_users[gu]))
-                                {
-                                        found = 1;
-                                        break;
-                                }
-                        }
-                        if (found) continue;
+                       int gu;
+                       for (gu = 0; gu < guest_accounts_num; gu++)
+                       {
+                               if (!strcasecmp(user, guest_users[gu]))
+                               {
+                                       found = 1;
+                                       break;
+                               }
+                       }
+                       if (found) continue;
 
                        // Drop the new session in case of duplicate sessionss, not the old one.
                        if (!strcasecmp(user, session[i].user))
@@ -5210,7 +5300,7 @@ int sessionsetup(sessionidt s, tunnelidt t)
        // no need to set a route for the same IP address of the bundle
        if (!session[s].bundle || (bundle[session[s].bundle].num_of_links == 1))
        {
-               int routed = 0;
+               int routed = 0;
 
                // Add the route for this session.
                for (r = 0; r < MAXROUTE && session[s].route[r].ip; r++)