Correctly form the delete route/route6 request.
[l2tpns.git] / l2tpns.c
index 57e3215..bab453f 100644 (file)
--- a/l2tpns.c
+++ b/l2tpns.c
@@ -14,7 +14,6 @@ char const *cvs_id_l2tpns = "$Id: l2tpns.c,v 1.176 2011/01/20 12:48:40 bodea Exp
 #define SYSLOG_NAMES
 #include <syslog.h>
 #include <malloc.h>
-#include <math.h>
 #include <net/route.h>
 #include <sys/mman.h>
 #include <netdb.h>
@@ -31,7 +30,7 @@ char const *cvs_id_l2tpns = "$Id: l2tpns.c,v 1.176 2011/01/20 12:48:40 bodea Exp
 #include <sys/time.h>
 #include <sys/resource.h>
 #include <sys/wait.h>
-#include <linux/if.h>
+#include <net/if.h>
 #include <stddef.h>
 #include <time.h>
 #include <dlfcn.h>
@@ -200,6 +199,8 @@ 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);
@@ -419,43 +420,65 @@ void random_data(uint8_t *buf, int len)
 // via BGP if enabled, and stuffs it into the
 // 'sessionbyip' cache.
 //
-// 'ip' and 'mask' must be in _host_ order.
+// 'ip' must be in _host_ order.
 //
-static void routeset(sessionidt s, in_addr_t ip, in_addr_t mask, in_addr_t gw, int add)
+static void routeset(sessionidt s, in_addr_t ip, int prefixlen, in_addr_t gw, int add)
 {
-       struct rtentry r;
+       struct {
+               struct nlmsghdr nh;
+               struct rtmsg rt;
+               char buf[32];
+       } req;
        int i;
+       in_addr_t n_ip;
 
-       if (!mask) mask = 0xffffffff;
+       if (!prefixlen) prefixlen = 32;
 
-       ip &= mask;             // Force the ip to be the first one in the route.
+       ip &= 0xffffffff << (32 - prefixlen);;  // Force the ip to be the first one in the route.
 
-       memset(&r, 0, sizeof(r));
-       r.rt_dev = config->tundevice;
-       r.rt_dst.sa_family = AF_INET;
-       *(uint32_t *) & (((struct sockaddr_in *) &r.rt_dst)->sin_addr.s_addr) = htonl(ip);
-       r.rt_gateway.sa_family = AF_INET;
-       *(uint32_t *) & (((struct sockaddr_in *) &r.rt_gateway)->sin_addr.s_addr) = htonl(gw);
-       r.rt_genmask.sa_family = AF_INET;
-       *(uint32_t *) & (((struct sockaddr_in *) &r.rt_genmask)->sin_addr.s_addr) = htonl(mask);
-       r.rt_flags = (RTF_UP | RTF_STATIC);
+       memset(&req, 0, sizeof(req));
+
+       if (add)
+       {
+               req.nh.nlmsg_type = RTM_NEWROUTE;
+               req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_REPLACE;
+       }
+       else
+       {
+               req.nh.nlmsg_type = RTM_DELROUTE;
+               req.nh.nlmsg_flags = NLM_F_REQUEST;
+       }
+
+       req.nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.rt));
+
+       req.rt.rtm_family = AF_INET;
+       req.rt.rtm_dst_len = prefixlen;
+       req.rt.rtm_table = RT_TABLE_MAIN;
+       req.rt.rtm_protocol = RTPROT_BOOT; // XXX
+       req.rt.rtm_scope = RT_SCOPE_LINK;
+       req.rt.rtm_type = RTN_UNICAST;
+
+       netlink_addattr(&req.nh, RTA_OIF, &tunidx, sizeof(int));
+       n_ip = htonl(ip);
+       netlink_addattr(&req.nh, RTA_DST, &n_ip, sizeof(n_ip));
        if (gw)
-               r.rt_flags |= RTF_GATEWAY;
-       else if (mask == 0xffffffff)
-               r.rt_flags |= RTF_HOST;
+       {
+               n_ip = htonl(gw);
+               netlink_addattr(&req.nh, RTA_GATEWAY, &n_ip, sizeof(n_ip));
+       }
 
-       LOG(1, s, 0, "Route %s %s/%s%s%s\n", add ? "add" : "del",
-           fmtaddr(htonl(ip), 0), fmtaddr(htonl(mask), 1),
+       LOG(1, s, 0, "Route %s %s/%d%s%s\n", add ? "add" : "del",
+           fmtaddr(htonl(ip), 0), prefixlen,
            gw ? " via" : "", gw ? fmtaddr(htonl(gw), 2) : "");
 
-       if (ioctl(ifrfd, add ? SIOCADDRT : SIOCDELRT, (void *) &r) < 0)
-               LOG(0, 0, 0, "routeset() error in ioctl: %s\n", strerror(errno));
+       if (netlink_send(&req.nh) < 0)
+               LOG(0, 0, 0, "routeset() error in sending netlink message: %s\n", strerror(errno));
 
 #ifdef BGP
        if (add)
-               bgp_add_route(htonl(ip), htonl(mask));
+               bgp_add_route(htonl(ip), prefixlen);
        else
-               bgp_del_route(htonl(ip), htonl(mask));
+               bgp_del_route(htonl(ip), prefixlen);
 #endif /* BGP */
 
                // Add/Remove the IPs to the 'sessionbyip' cache.
@@ -471,38 +494,61 @@ static void routeset(sessionidt s, in_addr_t ip, in_addr_t mask, in_addr_t gw, i
                if (!add)       // Are we deleting a route?
                        s = 0;  // Caching the session as '0' is the same as uncaching.
 
-               for (i = ip; (i&mask) == (ip&mask) ; ++i)
+               for (i = ip; i < ip+(1<<(32-prefixlen)) ; ++i)
                        cache_ipmap(i, s);
        }
 }
 
 void route6set(sessionidt s, struct in6_addr ip, int prefixlen, int add)
 {
-       struct in6_rtmsg rt;
+       struct {
+               struct nlmsghdr nh;
+               struct rtmsg rt;
+               char buf[64];
+       } req;
+       int metric;
        char ipv6addr[INET6_ADDRSTRLEN];
 
-       if (ifr6fd < 0)
+       if (!config->ipv6_prefix.s6_addr[0])
        {
                LOG(0, 0, 0, "Asked to set IPv6 route, but IPv6 not setup.\n");
                return;
        }
 
-       memset(&rt, 0, sizeof(rt));
+       memset(&req, 0, sizeof(req));
 
-       memcpy(&rt.rtmsg_dst, &ip, sizeof(struct in6_addr));
-       rt.rtmsg_dst_len = prefixlen;
-       rt.rtmsg_metric = 1;
-       rt.rtmsg_flags = RTF_UP;
-       rt.rtmsg_ifindex = tunidx;
+       if (add)
+       {
+               req.nh.nlmsg_type = RTM_NEWROUTE;
+               req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_CREATE | NLM_F_REPLACE;
+       }
+       else
+       {
+               req.nh.nlmsg_type = RTM_DELROUTE;
+               req.nh.nlmsg_flags = NLM_F_REQUEST;
+       }
+
+       req.nh.nlmsg_len = NLMSG_LENGTH(sizeof(req.rt));
+
+       req.rt.rtm_family = AF_INET6;
+       req.rt.rtm_dst_len = prefixlen;
+       req.rt.rtm_table = RT_TABLE_MAIN;
+       req.rt.rtm_protocol = RTPROT_BOOT; // XXX
+       req.rt.rtm_scope = RT_SCOPE_LINK;
+       req.rt.rtm_type = RTN_UNICAST;
+
+       netlink_addattr(&req.nh, RTA_OIF, &tunidx, sizeof(int));
+       netlink_addattr(&req.nh, RTA_DST, &ip, sizeof(ip));
+       metric = 1;
+       netlink_addattr(&req.nh, RTA_METRICS, &metric, sizeof(metric));
 
        LOG(1, 0, 0, "Route %s %s/%d\n",
            add ? "add" : "del",
            inet_ntop(AF_INET6, &ip, ipv6addr, INET6_ADDRSTRLEN),
            prefixlen);
 
-       if (ioctl(ifr6fd, add ? SIOCADDRT : SIOCDELRT, (void *) &rt) < 0)
-               LOG(0, 0, 0, "route6set() error in ioctl: %s\n",
-                               strerror(errno));
+       if (netlink_send(&req.nh) < 0)
+               LOG(0, 0, 0, "route6set() error in sending netlink message: %s\n", strerror(errno));
 
        // FIXME: need to add BGP routing (RFC2858)
 
@@ -577,7 +623,7 @@ static ssize_t netlink_recv(void *buf, ssize_t len)
 }
 
 /* adapted from iproute2 */
-void netlink_addattr(struct nlmsghdr *nh, int type, const void *data, int alen)
+static void netlink_addattr(struct nlmsghdr *nh, int type, const void *data, int alen)
 {
        int len = RTA_LENGTH(alen);
        struct rtattr *rta;
@@ -603,7 +649,6 @@ 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));
@@ -627,43 +672,13 @@ 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 __attribute__ ((aligned(NLMSG_ALIGNTO)));
-                       struct rtattr ifname_rta __attribute__ ((aligned(RTA_ALIGNTO)));
-                       char ifname[IFNAMSIZ];
-               } 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
@@ -671,19 +686,20 @@ static void inittun(void)
                        union {
                                struct ifinfomsg ifinfo;
                                struct ifaddrmsg ifaddr;
-                       } ifmsg __attribute__ ((aligned(NLMSG_ALIGNTO)));
+                       } ifmsg;
                        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
 
@@ -700,17 +716,18 @@ 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)
@@ -723,13 +740,13 @@ 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));
@@ -744,13 +761,13 @@ 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;
@@ -1927,11 +1944,11 @@ 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;
                }
 
@@ -4443,18 +4460,18 @@ 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.
@@ -4512,7 +4529,7 @@ 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;
@@ -4522,15 +4539,14 @@ 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
                {
@@ -5179,11 +5195,11 @@ 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
@@ -5254,7 +5270,7 @@ 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
@@ -5265,11 +5281,11 @@ 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
@@ -5288,11 +5304,11 @@ 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