key to data elements point at wrong memory area
[project/netifd.git] / system-linux.c
index 58ac5d4..e5364e0 100644 (file)
@@ -1,3 +1,18 @@
+/*
+ * netifd - network interface daemon
+ * Copyright (C) 2012 Felix Fietkau <nbd@openwrt.org>
+ * Copyright (C) 2013 Jo-Philipp Wich <jow@openwrt.org>
+ * Copyright (C) 2013 Steven Barth <steven@midlink.org>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2
+ * as published by the Free Software Foundation
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
 #define _GNU_SOURCE
 
 #include <sys/socket.h>
 #include <sys/stat.h>
 #include <sys/syscall.h>
 
+#include <net/if.h>
+#include <net/if_arp.h>
+
+#include <arpa/inet.h>
+#include <netinet/in.h>
+
 #include <linux/rtnetlink.h>
 #include <linux/sockios.h>
+#include <linux/ip.h>
+#include <linux/if_link.h>
 #include <linux/if_vlan.h>
 #include <linux/if_bridge.h>
+#include <linux/if_tunnel.h>
+#include <linux/ip6_tunnel.h>
 #include <linux/ethtool.h>
+#include <linux/fib_rules.h>
+
+#ifndef RTN_FAILED_POLICY
+#define RTN_FAILED_POLICY 12
+#endif
 
 #include <unistd.h>
 #include <string.h>
@@ -38,6 +68,8 @@ static struct nl_sock *sock_rtnl = NULL;
 static int cb_rtnl_event(struct nl_msg *msg, void *arg);
 static void handle_hotplug_event(struct uloop_fd *u, unsigned int events);
 
+static char dev_buf[256];
+
 static void
 handler_nl_event(struct uloop_fd *u, unsigned int events)
 {
@@ -131,10 +163,8 @@ static void system_set_sysctl(const char *path, const char *val)
 
 static void system_set_dev_sysctl(const char *path, const char *device, const char *val)
 {
-       static char buf[256];
-
-       snprintf(buf, sizeof(buf), path, val);
-       system_set_sysctl(buf, val);
+       snprintf(dev_buf, sizeof(dev_buf), path, device);
+       system_set_sysctl(dev_buf, val);
 }
 
 static void system_set_disable_ipv6(struct device *dev, const char *val)
@@ -214,6 +244,9 @@ found:
        if (dev->type != &simple_device_type)
                return;
 
+       if (add && system_if_force_external(dev->ifname))
+               return;
+
        device_set_present(dev, add);
 }
 
@@ -235,10 +268,15 @@ handle_hotplug_event(struct uloop_fd *u, unsigned int events)
 
 static int system_rtnl_call(struct nl_msg *msg)
 {
-       int s = -(nl_send_auto_complete(sock_rtnl, msg)
-                       || nl_wait_for_ack(sock_rtnl));
+       int ret;
+
+       ret = nl_send_auto_complete(sock_rtnl, msg);
        nlmsg_free(msg);
-       return s;
+
+       if (ret < 0)
+               return ret;
+
+       return nl_wait_for_ack(sock_rtnl);
 }
 
 int system_bridge_delbr(struct device *bridge)
@@ -249,6 +287,8 @@ int system_bridge_delbr(struct device *bridge)
 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
 {
        struct ifreq ifr;
+
+       memset(&ifr, 0, sizeof(ifr));
        if (dev)
                ifr.ifr_ifindex = dev->ifindex;
        else
@@ -257,18 +297,6 @@ static int system_bridge_if(const char *bridge, struct device *dev, int cmd, voi
        return ioctl(sock_ioctl, cmd, &ifr);
 }
 
-int system_bridge_addif(struct device *bridge, struct device *dev)
-{
-       system_set_disable_ipv6(dev, "1");
-       return system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
-}
-
-int system_bridge_delif(struct device *bridge, struct device *dev)
-{
-       system_set_disable_ipv6(dev, "0");
-       return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
-}
-
 static bool system_is_bridge(const char *name, char *buf, int buflen)
 {
        struct stat st;
@@ -305,6 +333,24 @@ static char *system_get_bridge(const char *name, char *buf, int buflen)
        return path + 1;
 }
 
+int system_bridge_addif(struct device *bridge, struct device *dev)
+{
+       char *oldbr;
+
+       system_set_disable_ipv6(dev, "1");
+       oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
+       if (oldbr && !strcmp(oldbr, bridge->ifname))
+               return 0;
+
+       return system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
+}
+
+int system_bridge_delif(struct device *bridge, struct device *dev)
+{
+       system_set_disable_ipv6(dev, "0");
+       return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
+}
+
 static int system_if_resolve(struct device *dev)
 {
        struct ifreq ifr;
@@ -318,6 +364,8 @@ static int system_if_resolve(struct device *dev)
 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
 {
        struct ifreq ifr;
+
+       memset(&ifr, 0, sizeof(ifr));
        strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
        ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
        ifr.ifr_flags |= add;
@@ -343,8 +391,13 @@ static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
 
 static bool check_route(struct nlmsghdr *hdr, int ifindex)
 {
+       struct rtmsg *r = NLMSG_DATA(hdr);
        struct nlattr *tb[__RTA_MAX];
 
+       if (r->rtm_protocol == RTPROT_KERNEL &&
+           r->rtm_family == AF_INET6)
+               return false;
+
        nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
        if (!tb[RTA_OIF])
                return false;
@@ -352,6 +405,11 @@ static bool check_route(struct nlmsghdr *hdr, int ifindex)
        return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
 }
 
+static bool check_rule(struct nlmsghdr *hdr, int ifindex)
+{
+       return true;
+}
+
 static int cb_clear_event(struct nl_msg *msg, void *arg)
 {
        struct clear_data *clr = arg;
@@ -374,16 +432,26 @@ static int cb_clear_event(struct nl_msg *msg, void *arg)
 
                cb = check_route;
                break;
+       case RTM_GETRULE:
+               type = RTM_DELRULE;
+               if (hdr->nlmsg_type != RTM_NEWRULE)
+                       return NL_SKIP;
+
+               cb = check_rule;
+               break;
        default:
                return NL_SKIP;
        }
 
-       if (!cb(hdr, clr->dev->ifindex))
+       if (!cb(hdr, clr->dev ? clr->dev->ifindex : 0))
                return NL_SKIP;
 
-       D(SYSTEM, "Remove %s from device %s\n",
-         type == RTM_DELADDR ? "an address" : "a route",
-         clr->dev->ifname);
+       if (type == RTM_DELRULE)
+               D(SYSTEM, "Remove a rule\n");
+       else
+               D(SYSTEM, "Remove %s from device %s\n",
+                 type == RTM_DELADDR ? "an address" : "a route",
+                 clr->dev->ifname);
        memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
        hdr = nlmsg_hdr(clr->msg);
        hdr->nlmsg_type = type;
@@ -428,6 +496,7 @@ system_if_clear_entries(struct device *dev, int type, int af)
        clr.type = type;
        switch (type) {
        case RTM_GETADDR:
+       case RTM_GETRULE:
                clr.size = sizeof(struct rtgenmsg);
                break;
        case RTM_GETROUTE:
@@ -518,6 +587,10 @@ int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
        system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
                bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 
+       args[0] = BRCTL_SET_BRIDGE_PRIORITY;
+       args[1] = cfg->priority;
+       system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
+
        if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
                args[0] = BRCTL_SET_AGEING_TIME;
                args[1] = sec_to_jiffies(cfg->ageing_time);
@@ -539,6 +612,93 @@ int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
        return 0;
 }
 
+int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
+{
+       struct nl_msg *msg;
+       struct nlattr *linkinfo, *data;
+       struct ifinfomsg iim = { .ifi_family = AF_INET };
+       int ifindex = system_if_resolve(dev);
+       int i, rv;
+       static const struct {
+               const char *name;
+               enum macvlan_mode val;
+       } modes[] = {
+               { "private", MACVLAN_MODE_PRIVATE },
+               { "vepa", MACVLAN_MODE_VEPA },
+               { "bridge", MACVLAN_MODE_BRIDGE },
+               { "passthru", MACVLAN_MODE_PASSTHRU },
+       };
+
+       if (ifindex == 0)
+               return -ENOENT;
+
+       msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
+
+       if (!msg)
+               return -1;
+
+       nlmsg_append(msg, &iim, sizeof(iim), 0);
+
+       if (cfg->flags & MACVLAN_OPT_MACADDR)
+               nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
+       nla_put(msg, IFLA_IFNAME, IFNAMSIZ, macvlan->ifname);
+       nla_put_u32(msg, IFLA_LINK, ifindex);
+
+       if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
+               goto nla_put_failure;
+
+       nla_put(msg, IFLA_INFO_KIND, strlen("macvlan"), "macvlan");
+
+       if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
+               goto nla_put_failure;
+
+       if (cfg->mode) {
+               for (i = 0; i < ARRAY_SIZE(modes); i++) {
+                       if (strcmp(cfg->mode, modes[i].name) != 0)
+                               continue;
+
+                       nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
+                       break;
+               }
+       }
+
+       nla_nest_end(msg, data);
+       nla_nest_end(msg, linkinfo);
+
+       rv = system_rtnl_call(msg);
+       if (rv)
+               D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
+
+       return rv;
+
+nla_put_failure:
+       nlmsg_free(msg);
+       return -ENOMEM;
+}
+
+int system_macvlan_del(struct device *macvlan)
+{
+       struct nl_msg *msg;
+       struct ifinfomsg iim;
+
+       iim.ifi_family = AF_INET;
+       iim.ifi_index  = 0;
+
+       msg = nlmsg_alloc_simple(RTM_DELLINK, 0);
+
+       if (!msg)
+               return -1;
+
+       nlmsg_append(msg, &iim, sizeof(iim), 0);
+
+       nla_put(msg, IFLA_INFO_KIND, strlen("macvlan"), "macvlan");
+       nla_put(msg, IFLA_IFNAME, sizeof(macvlan->ifname), macvlan->ifname);
+
+       system_rtnl_call(msg);
+
+       return 0;
+}
+
 static int system_vlan(struct device *dev, int id)
 {
        struct vlan_ioctl_args ifr = {
@@ -588,12 +748,12 @@ system_if_get_settings(struct device *dev, struct device_settings *s)
        }
 
        if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
-               memcpy(s->macaddr, &ifr.ifr_hwaddr, sizeof(s->macaddr));
+               memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
                s->flags |= DEV_OPT_MACADDR;
        }
 }
 
-static void
+void
 system_if_apply_settings(struct device *dev, struct device_settings *s)
 {
        struct ifreq ifr;
@@ -610,8 +770,9 @@ system_if_apply_settings(struct device *dev, struct device_settings *s)
                if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
                        s->flags &= ~DEV_OPT_TXQUEUELEN;
        }
-       if (s->flags & DEV_OPT_MACADDR) {
-               memcpy(&ifr.ifr_hwaddr, s->macaddr, sizeof(s->macaddr));
+       if ((s->flags & DEV_OPT_MACADDR) && !dev->external) {
+               ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
+               memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
                if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
                        s->flags &= ~DEV_OPT_MACADDR;
        }
@@ -628,6 +789,7 @@ int system_if_up(struct device *dev)
 int system_if_down(struct device *dev)
 {
        int ret = system_if_flags(dev->ifname, 0, IFF_UP);
+       dev->orig_settings.flags &= dev->settings.flags;
        system_if_apply_settings(dev, &dev->orig_settings);
        return ret;
 }
@@ -701,14 +863,14 @@ retry:
 }
 
 static bool
-read_int_file(int dir_fd, const char *file, int *val)
+read_uint64_file(int dir_fd, const char *file, uint64_t *val)
 {
        char buf[64];
        bool ret = false;
 
        ret = read_string_file(dir_fd, file, buf, sizeof(buf));
        if (ret)
-               *val = strtoul(buf, NULL, 0);
+               *val = strtoull(buf, NULL, 0);
 
        return ret;
 }
@@ -735,6 +897,16 @@ static void system_add_link_modes(struct blob_buf *b, __u32 mask)
        }
 }
 
+bool
+system_if_force_external(const char *ifname)
+{
+       char buf[64];
+       struct stat s;
+
+       snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
+       return stat(buf, &s) == 0;
+}
+
 int
 system_if_dump_info(struct device *dev, struct blob_buf *b)
 {
@@ -742,12 +914,13 @@ system_if_dump_info(struct device *dev, struct blob_buf *b)
        struct ifreq ifr;
        char buf[64], *s;
        void *c;
-       int dir_fd, val = 0;
+       int dir_fd;
+       uint64_t val = 0;
 
        snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
        dir_fd = open(buf, O_DIRECTORY);
 
-       if (read_int_file(dir_fd, "carrier", &val))
+       if (read_uint64_file(dir_fd, "carrier", &val))
                blobmsg_add_u8(b, "link", !!val);
 
        memset(&ecmd, 0, sizeof(ecmd));
@@ -790,7 +963,8 @@ system_if_dump_stats(struct device *dev, struct blob_buf *b)
        };
        char buf[64];
        int stats_dir;
-       int i, val = 0;
+       int i;
+       uint64_t val = 0;
 
        snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
        stats_dir = open(buf, O_DIRECTORY);
@@ -798,8 +972,8 @@ system_if_dump_stats(struct device *dev, struct blob_buf *b)
                return -1;
 
        for (i = 0; i < ARRAY_SIZE(counters); i++)
-               if (read_int_file(stats_dir, counters[i], &val))
-                       blobmsg_add_u32(b, counters[i], val);
+               if (read_uint64_file(stats_dir, counters[i], &val))
+                       blobmsg_add_u64(b, counters[i], val);
 
        close(stats_dir);
        return 0;
@@ -807,7 +981,9 @@ system_if_dump_stats(struct device *dev, struct blob_buf *b)
 
 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
 {
-       int alen = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
+       bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
+       int alen = v4 ? 4 : 16;
+       unsigned int flags = 0;
        struct ifaddrmsg ifa = {
                .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
                .ifa_prefixlen = addr->mask,
@@ -815,13 +991,47 @@ static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
        };
 
        struct nl_msg *msg;
+       if (cmd == RTM_NEWADDR)
+               flags |= NLM_F_CREATE | NLM_F_REPLACE;
 
-       msg = nlmsg_alloc_simple(cmd, 0);
+       msg = nlmsg_alloc_simple(cmd, flags);
        if (!msg)
                return -1;
 
        nlmsg_append(msg, &ifa, sizeof(ifa), 0);
        nla_put(msg, IFA_LOCAL, alen, &addr->addr);
+       if (v4) {
+               if (addr->broadcast)
+                       nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
+               if (addr->point_to_point)
+                       nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
+       } else {
+               time_t now = system_get_rtime();
+               struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
+
+               if (addr->preferred_until) {
+                       int64_t preferred = addr->preferred_until - now;
+                       if (preferred < 0)
+                               preferred = 0;
+                       else if (preferred > UINT32_MAX)
+                               preferred = UINT32_MAX;
+
+                       cinfo.ifa_prefered = preferred;
+               }
+
+               if (addr->valid_until) {
+                       int64_t valid = addr->valid_until - now;
+                       if (valid <= 0)
+                               return -1;
+                       else if (valid > UINT32_MAX)
+                               valid = UINT32_MAX;
+
+                       cinfo.ifa_valid = valid;
+               }
+
+               nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
+       }
+
        return system_rtnl_call(msg);
 }
 
@@ -840,7 +1050,6 @@ static int system_rt(struct device *dev, struct device_route *route, int cmd)
        int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
        bool have_gw;
        unsigned int flags = 0;
-       int ifindex = dev->ifindex;
 
        if (alen == 4)
                have_gw = !!route->nexthop.in.s_addr;
@@ -853,19 +1062,28 @@ static int system_rt(struct device *dev, struct device_route *route, int cmd)
        unsigned char scope = (cmd == RTM_DELROUTE) ? RT_SCOPE_NOWHERE :
                        (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
 
+       unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
+                       ? route->table : RT_TABLE_MAIN;
+
        struct rtmsg rtm = {
                .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
                .rtm_dst_len = route->mask,
-               .rtm_table = RT_TABLE_MAIN,
-               .rtm_protocol = RTPROT_BOOT,
+               .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
+               .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
                .rtm_scope = scope,
                .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
        };
        struct nl_msg *msg;
 
-       if (cmd == RTM_NEWROUTE)
+       if (cmd == RTM_NEWROUTE) {
                flags |= NLM_F_CREATE | NLM_F_REPLACE;
 
+               if (!dev) { // Add null-route
+                       rtm.rtm_scope = RT_SCOPE_UNIVERSE;
+                       rtm.rtm_type = RTN_UNREACHABLE;
+               }
+       }
+
        msg = nlmsg_alloc_simple(cmd, flags);
        if (!msg)
                return -1;
@@ -875,14 +1093,17 @@ static int system_rt(struct device *dev, struct device_route *route, int cmd)
        if (route->mask)
                nla_put(msg, RTA_DST, alen, &route->addr);
 
-       if (route->metric >= 0)
+       if (route->metric > 0)
                nla_put_u32(msg, RTA_PRIORITY, route->metric);
 
        if (have_gw)
                nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
 
-       if (route->flags & DEVADDR_DEVICE)
-               nla_put_u32(msg, RTA_OIF, ifindex);
+       if (dev)
+               nla_put_u32(msg, RTA_OIF, dev->ifindex);
+
+       if (table >= 256)
+               nla_put_u32(msg, RTA_TABLE, table);
 
        return system_rtnl_call(msg);
 }
@@ -916,6 +1137,228 @@ int system_flush_routes(void)
        return 0;
 }
 
+bool system_resolve_rt_table(const char *name, unsigned int *id)
+{
+       FILE *f;
+       char *e, buf[128];
+       unsigned int n, table = RT_TABLE_UNSPEC;
+
+       /* first try to parse table as number */
+       if ((n = strtoul(name, &e, 0)) > 0 && !*e)
+               table = n;
+
+       /* handle well known aliases */
+       else if (!strcmp(name, "default"))
+               table = RT_TABLE_DEFAULT;
+       else if (!strcmp(name, "main"))
+               table = RT_TABLE_MAIN;
+       else if (!strcmp(name, "local"))
+               table = RT_TABLE_LOCAL;
+
+       /* try to look up name in /etc/iproute2/rt_tables */
+       else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
+       {
+               while (fgets(buf, sizeof(buf) - 1, f) != NULL)
+               {
+                       if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
+                               continue;
+
+                       n = strtoul(e, NULL, 10);
+                       e = strtok(NULL, " \t\n");
+
+                       if (e && !strcmp(e, name))
+                       {
+                               table = n;
+                               break;
+                       }
+               }
+
+               fclose(f);
+       }
+
+       if (table == RT_TABLE_UNSPEC)
+               return false;
+
+       /* do not consider main table special */
+       if (table == RT_TABLE_MAIN)
+               table = RT_TABLE_UNSPEC;
+
+       *id = table;
+       return true;
+}
+
+static int system_iprule(struct iprule *rule, int cmd)
+{
+       int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
+
+       struct nl_msg *msg;
+       struct rtmsg rtm = {
+               .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
+               .rtm_protocol = RTPROT_STATIC,
+               .rtm_scope = RT_SCOPE_UNIVERSE,
+               .rtm_table = RT_TABLE_UNSPEC,
+               .rtm_type = RTN_UNSPEC,
+               .rtm_flags = 0,
+       };
+
+       if (cmd == RTM_NEWRULE) {
+               rtm.rtm_type = RTN_UNICAST;
+               rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
+       }
+
+       if (rule->invert)
+               rtm.rtm_flags |= FIB_RULE_INVERT;
+
+       if (rule->flags & IPRULE_SRC)
+               rtm.rtm_src_len = rule->src_mask;
+
+       if (rule->flags & IPRULE_DEST)
+               rtm.rtm_dst_len = rule->dest_mask;
+
+       if (rule->flags & IPRULE_TOS)
+               rtm.rtm_tos = rule->tos;
+
+       if (rule->flags & IPRULE_LOOKUP) {
+               if (rule->lookup < 256)
+                       rtm.rtm_table = rule->lookup;
+       }
+
+       if (rule->flags & IPRULE_ACTION)
+               rtm.rtm_type = rule->action;
+       else if (rule->flags & IPRULE_GOTO)
+               rtm.rtm_type = FR_ACT_GOTO;
+       else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
+               rtm.rtm_type = FR_ACT_NOP;
+
+       msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
+
+       if (!msg)
+               return -1;
+
+       nlmsg_append(msg, &rtm, sizeof(rtm), 0);
+
+       if (rule->flags & IPRULE_IN)
+               nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
+
+       if (rule->flags & IPRULE_OUT)
+               nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
+
+       if (rule->flags & IPRULE_SRC)
+               nla_put(msg, FRA_SRC, alen, &rule->src_addr);
+
+       if (rule->flags & IPRULE_DEST)
+               nla_put(msg, FRA_DST, alen, &rule->dest_addr);
+
+       if (rule->flags & IPRULE_PRIORITY)
+               nla_put_u32(msg, FRA_PRIORITY, rule->priority);
+       else if (cmd == RTM_NEWRULE)
+               nla_put_u32(msg, FRA_PRIORITY, rule->order);
+
+       if (rule->flags & IPRULE_FWMARK)
+               nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
+
+       if (rule->flags & IPRULE_FWMASK)
+               nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
+
+       if (rule->flags & IPRULE_LOOKUP) {
+               if (rule->lookup >= 256)
+                       nla_put_u32(msg, FRA_TABLE, rule->lookup);
+       }
+
+       if (rule->flags & IPRULE_GOTO)
+               nla_put_u32(msg, FRA_GOTO, rule->gotoid);
+
+       return system_rtnl_call(msg);
+}
+
+int system_add_iprule(struct iprule *rule)
+{
+       return system_iprule(rule, RTM_NEWRULE);
+}
+
+int system_del_iprule(struct iprule *rule)
+{
+       return system_iprule(rule, RTM_DELRULE);
+}
+
+int system_flush_iprules(void)
+{
+       int rv = 0;
+       struct iprule rule;
+
+       system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
+       system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
+
+       memset(&rule, 0, sizeof(rule));
+
+
+       rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
+
+       rule.priority = 0;
+       rule.lookup = RT_TABLE_LOCAL;
+       rv |= system_iprule(&rule, RTM_NEWRULE);
+
+       rule.priority = 32766;
+       rule.lookup = RT_TABLE_MAIN;
+       rv |= system_iprule(&rule, RTM_NEWRULE);
+
+       rule.priority = 32767;
+       rule.lookup = RT_TABLE_DEFAULT;
+       rv |= system_iprule(&rule, RTM_NEWRULE);
+
+
+       rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
+
+       rule.priority = 0;
+       rule.lookup = RT_TABLE_LOCAL;
+       rv |= system_iprule(&rule, RTM_NEWRULE);
+
+       rule.priority = 32766;
+       rule.lookup = RT_TABLE_MAIN;
+       rv |= system_iprule(&rule, RTM_NEWRULE);
+
+       return rv;
+}
+
+bool system_resolve_iprule_action(const char *action, unsigned int *id)
+{
+       char *e;
+       unsigned int n;
+
+       if (!strcmp(action, "local"))
+               n = RTN_LOCAL;
+       else if (!strcmp(action, "nat"))
+               n = RTN_NAT;
+       else if (!strcmp(action, "broadcast"))
+               n = RTN_BROADCAST;
+       else if (!strcmp(action, "anycast"))
+               n = RTN_ANYCAST;
+       else if (!strcmp(action, "multicast"))
+               n = RTN_MULTICAST;
+       else if (!strcmp(action, "prohibit"))
+               n = RTN_PROHIBIT;
+       else if (!strcmp(action, "unreachable"))
+               n = RTN_UNREACHABLE;
+       else if (!strcmp(action, "blackhole"))
+               n = RTN_BLACKHOLE;
+       else if (!strcmp(action, "xresolve"))
+               n = RTN_XRESOLVE;
+       else if (!strcmp(action, "unicast"))
+               n = RTN_UNICAST;
+       else if (!strcmp(action, "throw"))
+               n = RTN_THROW;
+       else if (!strcmp(action, "failed_policy"))
+               n = RTN_FAILED_POLICY;
+       else {
+               n = strtoul(action, &e, 0);
+               if (!e || *e || e == action || n > 255)
+                       return false;
+       }
+
+       *id = n;
+       return true;
+}
+
 time_t system_get_rtime(void)
 {
        struct timespec ts;
@@ -929,3 +1372,154 @@ time_t system_get_rtime(void)
 
        return 0;
 }
+
+#ifndef IP_DF
+#define IP_DF       0x4000
+#endif
+
+static int tunnel_ioctl(const char *name, int cmd, void *p)
+{
+       struct ifreq ifr;
+
+       memset(&ifr, 0, sizeof(ifr));
+       strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
+       ifr.ifr_ifru.ifru_data = p;
+       return ioctl(sock_ioctl, cmd, &ifr);
+}
+
+int system_del_ip_tunnel(const char *name)
+{
+       return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
+}
+
+int system_update_ipv6_mtu(struct device *dev, int mtu)
+{
+       int ret = -1;
+       char buf[64];
+       snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
+                       dev->ifname);
+
+       int fd = open(buf, O_RDWR);
+       ssize_t len = read(fd, buf, sizeof(buf) - 1);
+       if (len < 0)
+               goto out;
+
+       buf[len] = 0;
+       ret = atoi(buf);
+
+       if (!mtu || ret <= mtu)
+               goto out;
+
+       lseek(fd, 0, SEEK_SET);
+       if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
+               ret = -1;
+
+out:
+       close(fd);
+       return ret;
+}
+
+
+int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
+{
+       struct blob_attr *tb[__TUNNEL_ATTR_MAX];
+       struct blob_attr *cur;
+       const char *str;
+
+       system_del_ip_tunnel(name);
+
+       blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
+               blob_data(attr), blob_len(attr));
+
+       if (!(cur = tb[TUNNEL_ATTR_TYPE]))
+               return -EINVAL;
+       str = blobmsg_data(cur);
+
+       unsigned int ttl = 0;
+       if ((cur = tb[TUNNEL_ATTR_TTL]) && (ttl = blobmsg_get_u32(cur)) > 255)
+               return -EINVAL;
+
+       unsigned int link = 0;
+       if ((cur = tb[TUNNEL_ATTR_LINK])) {
+               struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
+               if (!iface)
+                       return -EINVAL;
+
+               if (iface->l3_dev.dev)
+                       link = iface->l3_dev.dev->ifindex;
+       }
+
+
+       if (!strcmp(str, "sit")) {
+               struct ip_tunnel_parm p = {
+                       .link = link,
+                       .iph = {
+                               .version = 4,
+                               .ihl = 5,
+                               .frag_off = htons(IP_DF),
+                               .protocol = IPPROTO_IPV6,
+                               .ttl = ttl
+                       }
+               };
+
+               if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
+                               inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
+                       return -EINVAL;
+
+               if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
+                               inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
+                       return -EINVAL;
+
+               strncpy(p.name, name, sizeof(p.name));
+               if (tunnel_ioctl("sit0", SIOCADDTUNNEL, &p) < 0)
+                       return -1;
+
+#ifdef SIOCADD6RD
+               if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
+                       unsigned int mask;
+                       struct ip_tunnel_6rd p6;
+
+                       memset(&p6, 0, sizeof(p6));
+
+                       if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
+                                               &p6.prefix, &mask) || mask > 128)
+                               return -EINVAL;
+                       p6.prefixlen = mask;
+
+                       if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
+                               if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
+                                                       &p6.relay_prefix, &mask) || mask > 32)
+                                       return -EINVAL;
+                               p6.relay_prefixlen = mask;
+                       }
+
+                       if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
+                               system_del_ip_tunnel(name);
+                               return -1;
+                       }
+               }
+#endif
+       } else if (!strcmp(str, "ipip6")) {
+               struct ip6_tnl_parm p = {
+                       .link = link,
+                       .proto = IPPROTO_IPIP,
+                       .hop_limit = (ttl) ? ttl : 64,
+                       .encap_limit = 4,
+               };
+
+               if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
+                               inet_pton(AF_INET6, blobmsg_data(cur), &p.laddr) < 1)
+                       return -EINVAL;
+
+               if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
+                               inet_pton(AF_INET6, blobmsg_data(cur), &p.raddr) < 1)
+                       return -EINVAL;
+
+               strncpy(p.name, name, sizeof(p.name));
+               if (tunnel_ioctl("ip6tnl0", SIOCADDTUNNEL, &p) < 0)
+                       return -1;
+       } else
+               return -EINVAL;
+
+       return 0;
+}