bridge: enable multicast querier functionality by default
[project/netifd.git] / system-linux.c
index 604c206..ea3a138 100644 (file)
@@ -3,6 +3,7 @@
  * 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>
+ * Copyright (C) 2014 Gioacchino Mazzurco <gio@eigenlab.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
 #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>
+#include <linux/version.h>
+
+#ifndef RTN_FAILED_POLICY
+#define RTN_FAILED_POLICY 12
+#endif
 
-#include <unistd.h>
 #include <string.h>
 #include <fcntl.h>
 #include <glob.h>
@@ -53,7 +60,7 @@
 struct event_socket {
        struct uloop_fd uloop;
        struct nl_sock *sock;
-       struct nl_cb *cb;
+       int bufsize;
 };
 
 static int sock_ioctl = -1;
@@ -68,7 +75,38 @@ static void
 handler_nl_event(struct uloop_fd *u, unsigned int events)
 {
        struct event_socket *ev = container_of(u, struct event_socket, uloop);
-       nl_recvmsgs(ev->sock, ev->cb);
+       int err;
+       socklen_t errlen = sizeof(err);
+
+       if (!u->error) {
+               nl_recvmsgs_default(ev->sock);
+               return;
+       }
+
+       if (getsockopt(u->fd, SOL_SOCKET, SO_ERROR, (void *)&err, &errlen))
+               goto abort;
+
+       switch(err) {
+       case ENOBUFS:
+               // Increase rx buffer size on netlink socket
+               ev->bufsize *= 2;
+               if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
+                       goto abort;
+
+               // Request full dump since some info got dropped
+               struct rtgenmsg msg = { .rtgen_family = AF_UNSPEC };
+               nl_send_simple(ev->sock, RTM_GETLINK, NLM_F_DUMP, &msg, sizeof(msg));
+               break;
+
+       default:
+               goto abort;
+       }
+       u->error = false;
+       return;
+
+abort:
+       uloop_fd_delete(&ev->uloop);
+       return;
 }
 
 static struct nl_sock *
@@ -91,7 +129,7 @@ create_socket(int protocol, int groups)
 
 static bool
 create_raw_event_socket(struct event_socket *ev, int protocol, int groups,
-                       uloop_fd_handler cb)
+                        uloop_fd_handler cb, int flags)
 {
        ev->sock = create_socket(protocol, groups);
        if (!ev->sock)
@@ -99,7 +137,9 @@ create_raw_event_socket(struct event_socket *ev, int protocol, int groups,
 
        ev->uloop.fd = nl_socket_get_fd(ev->sock);
        ev->uloop.cb = cb;
-       uloop_fd_add(&ev->uloop, ULOOP_READ | ULOOP_EDGE_TRIGGER);
+       if (uloop_fd_add(&ev->uloop, ULOOP_READ|flags))
+               return false;
+
        return true;
 }
 
@@ -107,14 +147,61 @@ static bool
 create_event_socket(struct event_socket *ev, int protocol,
                    int (*cb)(struct nl_msg *msg, void *arg))
 {
-       // Prepare socket for link events
-       ev->cb = nl_cb_alloc(NL_CB_DEFAULT);
-       if (!ev->cb)
+       if (!create_raw_event_socket(ev, protocol, 0, handler_nl_event, ULOOP_ERROR_CB))
                return false;
 
-       nl_cb_set(ev->cb, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL);
+       // Install the valid custom callback handler
+       nl_socket_modify_cb(ev->sock, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL);
+
+       // Disable sequence number checking on event sockets
+       nl_socket_disable_seq_check(ev->sock);
 
-       return create_raw_event_socket(ev, protocol, 0, handler_nl_event);
+       // Increase rx buffer size to 65K on event sockets
+       ev->bufsize = 65535;
+       if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
+               return false;
+
+       return true;
+}
+
+static bool
+system_rtn_aton(const char *src, unsigned int *dst)
+{
+       char *e;
+       unsigned int n;
+
+       if (!strcmp(src, "local"))
+               n = RTN_LOCAL;
+       else if (!strcmp(src, "nat"))
+               n = RTN_NAT;
+       else if (!strcmp(src, "broadcast"))
+               n = RTN_BROADCAST;
+       else if (!strcmp(src, "anycast"))
+               n = RTN_ANYCAST;
+       else if (!strcmp(src, "multicast"))
+               n = RTN_MULTICAST;
+       else if (!strcmp(src, "prohibit"))
+               n = RTN_PROHIBIT;
+       else if (!strcmp(src, "unreachable"))
+               n = RTN_UNREACHABLE;
+       else if (!strcmp(src, "blackhole"))
+               n = RTN_BLACKHOLE;
+       else if (!strcmp(src, "xresolve"))
+               n = RTN_XRESOLVE;
+       else if (!strcmp(src, "unicast"))
+               n = RTN_UNICAST;
+       else if (!strcmp(src, "throw"))
+               n = RTN_THROW;
+       else if (!strcmp(src, "failed_policy"))
+               n = RTN_FAILED_POLICY;
+       else {
+               n = strtoul(src, &e, 0);
+               if (!e || *e || e == src || n > 255)
+                       return false;
+       }
+
+       *dst = n;
+       return true;
 }
 
 int system_init(void)
@@ -123,7 +210,7 @@ int system_init(void)
        static struct event_socket hotplug_event;
 
        sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
-       fcntl(sock_ioctl, F_SETFD, fcntl(sock_ioctl, F_GETFD) | FD_CLOEXEC);
+       system_fd_set_cloexec(sock_ioctl);
 
        // Prepare socket for routing / address control
        sock_rtnl = create_socket(NETLINK_ROUTE, 0);
@@ -134,7 +221,7 @@ int system_init(void)
                return -1;
 
        if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1,
-                                    handle_hotplug_event))
+                                    handle_hotplug_event, 0))
                return -1;
 
        // Receive network link events form kernel
@@ -151,7 +238,7 @@ static void system_set_sysctl(const char *path, const char *val)
        if (fd < 0)
                return;
 
-       write(fd, val, strlen(val));
+       if (write(fd, val, strlen(val))) {}
        close(fd);
 }
 
@@ -166,6 +253,44 @@ static void system_set_disable_ipv6(struct device *dev, const char *val)
        system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val);
 }
 
+static int system_get_sysctl(const char *path, char *buf, const size_t buf_sz)
+{
+       int fd = -1, ret = -1;
+
+       fd = open(path, O_RDONLY);
+       if (fd < 0)
+               goto out;
+
+       ssize_t len = read(fd, buf, buf_sz - 1);
+       if (len < 0)
+               goto out;
+
+       ret = buf[len] = 0;
+
+out:
+       if (fd >= 0)
+               close(fd);
+
+       return ret;
+}
+
+static int
+system_get_dev_sysctl(const char *path, const char *device, char *buf, const size_t buf_sz)
+{
+       snprintf(dev_buf, sizeof(dev_buf), path, device);
+       return system_get_sysctl(dev_buf, buf, buf_sz);
+}
+
+static int system_get_disable_ipv6(struct device *dev, char *buf, const size_t buf_sz)
+{
+       return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6",
+                       dev->ifname, buf, buf_sz);
+}
+
+#ifndef IFF_LOWER_UP
+#define IFF_LOWER_UP   0x10000
+#endif
+
 // Evaluate netlink messages
 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
 {
@@ -173,19 +298,19 @@ static int cb_rtnl_event(struct nl_msg *msg, void *arg)
        struct ifinfomsg *ifi = NLMSG_DATA(nh);
        struct nlattr *nla[__IFLA_MAX];
 
-       if (nh->nlmsg_type != RTM_DELLINK && nh->nlmsg_type != RTM_NEWLINK)
+       if (nh->nlmsg_type != RTM_NEWLINK)
                goto out;
 
        nlmsg_parse(nh, sizeof(*ifi), nla, __IFLA_MAX - 1, NULL);
        if (!nla[IFLA_IFNAME])
                goto out;
 
-       struct device *dev = device_get(RTA_DATA(nla[IFLA_IFNAME]), false);
+       struct device *dev = device_get(nla_data(nla[IFLA_IFNAME]), false);
        if (!dev)
                goto out;
 
-       dev->ifindex = ifi->ifi_index;
-       /* TODO: parse link status */
+       device_set_ifindex(dev, ifi->ifi_index);
+       device_set_link(dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
 
 out:
        return 0;
@@ -327,21 +452,31 @@ static char *system_get_bridge(const char *name, char *buf, int buflen)
        return path + 1;
 }
 
+static void system_bridge_set_wireless(const char *bridge, const char *dev)
+{
+       snprintf(dev_buf, sizeof(dev_buf),
+                "/sys/devices/virtual/net/%s/brif/%s/multicast_to_unicast",
+                bridge, dev);
+       system_set_sysctl(dev_buf, "1");
+}
+
 int system_bridge_addif(struct device *bridge, struct device *dev)
 {
        char *oldbr;
+       int ret = 0;
 
-       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;
+       if (!oldbr || strcmp(oldbr, bridge->ifname) != 0)
+               ret = system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
+
+       if (dev->wireless)
+               system_bridge_set_wireless(bridge->ifname, dev->ifname);
 
-       return system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
+       return ret;
 }
 
 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);
 }
 
@@ -451,8 +586,9 @@ static int cb_clear_event(struct nl_msg *msg, void *arg)
        hdr->nlmsg_type = type;
        hdr->nlmsg_flags = NLM_F_REQUEST;
 
-       if (!nl_send_auto_complete(sock_rtnl, clr->msg))
-               nl_wait_for_ack(sock_rtnl);
+       nl_socket_disable_auto_ack(sock_rtnl);
+       nl_send_auto_complete(sock_rtnl, clr->msg);
+       nl_socket_enable_auto_ack(sock_rtnl);
 
        return NL_SKIP;
 }
@@ -529,11 +665,8 @@ void system_if_clear_state(struct device *dev)
        static char buf[256];
        char *bridge;
 
-       if (dev->external)
-               return;
-
-       dev->ifindex = system_if_resolve(dev);
-       if (!dev->ifindex)
+       device_set_ifindex(dev, system_if_resolve(dev));
+       if (dev->external || !dev->ifindex)
                return;
 
        system_if_flags(dev->ifname, 0, IFF_UP);
@@ -581,11 +714,12 @@ 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");
 
-       if (cfg->flags & BRIDGE_OPT_PRIORITY) {
-               args[0] = BRCTL_SET_BRIDGE_PRIORITY;
-               args[1] = cfg->priority;
-               system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
-       }
+       system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_querier",
+               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;
@@ -608,6 +742,93 @@ 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_UNSPEC, };
+       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_string(msg, IFLA_IFNAME, macvlan->ifname);
+       nla_put_u32(msg, IFLA_LINK, ifindex);
+
+       if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
+               goto nla_put_failure;
+
+       nla_put_string(msg, IFLA_INFO_KIND, "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;
+}
+
+static int system_link_del(struct device *dev)
+{
+       struct nl_msg *msg;
+       struct ifinfomsg iim = {
+               .ifi_family = AF_UNSPEC,
+               .ifi_index = 0,
+       };
+
+       msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
+
+       if (!msg)
+               return -1;
+
+       nlmsg_append(msg, &iim, sizeof(iim), 0);
+       nla_put_string(msg, IFLA_IFNAME, dev->ifname);
+       return system_rtnl_call(msg);
+}
+
+int system_macvlan_del(struct device *macvlan)
+{
+       return system_link_del(macvlan);
+}
+
 static int system_vlan(struct device *dev, int id)
 {
        struct vlan_ioctl_args ifr = {
@@ -638,10 +859,67 @@ int system_vlan_del(struct device *dev)
        return system_vlan(dev, -1);
 }
 
+int system_vlandev_add(struct device *vlandev, struct device *dev, struct vlandev_config *cfg)
+{
+       struct nl_msg *msg;
+       struct nlattr *linkinfo, *data;
+       struct ifinfomsg iim = { .ifi_family = AF_UNSPEC };
+       int ifindex = system_if_resolve(dev);
+       int rv;
+
+       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);
+       nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
+       nla_put_u32(msg, IFLA_LINK, ifindex);
+       
+       if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
+               goto nla_put_failure;
+       
+       nla_put_string(msg, IFLA_INFO_KIND, "vlan");
+
+       if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
+               goto nla_put_failure;
+
+       nla_put_u16(msg, IFLA_VLAN_ID, cfg->vid);
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3,10,0)
+       nla_put_u16(msg, IFLA_VLAN_PROTOCOL, htons(cfg->proto));
+#else
+       if(cfg->proto == VLAN_PROTO_8021AD)
+               netifd_log_message(L_WARNING, "%s Your kernel is older than linux 3.10.0, 802.1ad is not supported defaulting to 802.1q", vlandev->type->name);
+#endif
+
+       nla_nest_end(msg, data);
+       nla_nest_end(msg, linkinfo);
+
+       rv = system_rtnl_call(msg);
+       if (rv)
+               D(SYSTEM, "Error adding vlandev '%s' over '%s': %d\n", vlandev->ifname, dev->ifname, rv);
+
+       return rv;
+
+nla_put_failure:
+       nlmsg_free(msg);
+       return -ENOMEM;
+}
+
+int system_vlandev_del(struct device *vlandev)
+{
+       return system_link_del(vlandev);
+}
+
 static void
 system_if_get_settings(struct device *dev, struct device_settings *s)
 {
        struct ifreq ifr;
+       char buf[10];
 
        memset(&ifr, 0, sizeof(ifr));
        strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
@@ -660,38 +938,48 @@ system_if_get_settings(struct device *dev, struct device_settings *s)
                memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
                s->flags |= DEV_OPT_MACADDR;
        }
+
+       if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
+               s->ipv6 = !strtoul(buf, NULL, 0);
+               s->flags |= DEV_OPT_IPV6;
+       }
 }
 
 void
-system_if_apply_settings(struct device *dev, struct device_settings *s)
+system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
 {
        struct ifreq ifr;
 
+       if (!apply_mask)
+               return;
+
        memset(&ifr, 0, sizeof(ifr));
        strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
-       if (s->flags & DEV_OPT_MTU) {
+       if (s->flags & DEV_OPT_MTU & apply_mask) {
                ifr.ifr_mtu = s->mtu;
                if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
                        s->flags &= ~DEV_OPT_MTU;
        }
-       if (s->flags & DEV_OPT_TXQUEUELEN) {
+       if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
                ifr.ifr_qlen = s->txqueuelen;
                if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
                        s->flags &= ~DEV_OPT_TXQUEUELEN;
        }
-       if ((s->flags & DEV_OPT_MACADDR) && !dev->external) {
+       if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !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;
        }
+       if (s->flags & DEV_OPT_IPV6 & apply_mask)
+               system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
 }
 
 int system_if_up(struct device *dev)
 {
        system_if_get_settings(dev, &dev->orig_settings);
-       system_if_apply_settings(dev, &dev->settings);
-       dev->ifindex = system_if_resolve(dev);
+       system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
+       device_set_ifindex(dev, system_if_resolve(dev));
        return system_if_flags(dev->ifname, IFF_UP, 0);
 }
 
@@ -699,14 +987,82 @@ 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);
+       system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
        return ret;
 }
 
+struct if_check_data {
+       struct device *dev;
+       int pending;
+       int ret;
+};
+
+static int cb_if_check_valid(struct nl_msg *msg, void *arg)
+{
+       struct nlmsghdr *nh = nlmsg_hdr(msg);
+       struct ifinfomsg *ifi = NLMSG_DATA(nh);
+       struct if_check_data *chk = (struct if_check_data *)arg;
+
+       if (nh->nlmsg_type != RTM_NEWLINK)
+               return NL_SKIP;
+
+       device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
+       device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
+
+       return NL_OK;
+}
+
+static int cb_if_check_ack(struct nl_msg *msg, void *arg)
+{
+       struct if_check_data *chk = (struct if_check_data *)arg;
+       chk->pending = 0;
+       return NL_STOP;
+}
+
+static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
+{
+       struct if_check_data *chk = (struct if_check_data *)arg;
+
+       device_set_present(chk->dev, false);
+       device_set_link(chk->dev, false);
+       chk->pending = err->error;
+
+       return NL_STOP;
+}
+
 int system_if_check(struct device *dev)
 {
-       device_set_present(dev, (system_if_resolve(dev) > 0));
-       return 0;
+       struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
+       struct nl_msg *msg;
+       struct ifinfomsg ifi = {
+               .ifi_family = AF_UNSPEC,
+               .ifi_index = 0,
+       };
+       struct if_check_data chk = {
+               .dev = dev,
+               .pending = 1,
+       };
+       int ret = 1;
+
+       msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
+       if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
+           nla_put_string(msg, IFLA_IFNAME, dev->ifname))
+               goto out;
+
+       nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
+       nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
+       nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
+
+       nl_send_auto_complete(sock_rtnl, msg);
+       while (chk.pending > 0)
+               nl_recvmsgs(sock_rtnl, cb);
+
+       nlmsg_free(msg);
+       ret = chk.pending;
+
+out:
+       nl_cb_put(cb);
+       return ret;
 }
 
 struct device *
@@ -824,14 +1180,10 @@ system_if_dump_info(struct device *dev, struct blob_buf *b)
        char buf[64], *s;
        void *c;
        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_uint64_file(dir_fd, "carrier", &val))
-               blobmsg_add_u8(b, "link", !!val);
-
        memset(&ecmd, 0, sizeof(ecmd));
        memset(&ifr, 0, sizeof(ifr));
        strcpy(ifr.ifr_name, dev->ifname);
@@ -919,17 +1271,21 @@ static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
                struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
 
                if (addr->preferred_until) {
-                       int preferred = addr->preferred_until - now;
+                       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) {
-                       int valid = addr->valid_until - now;
+                       int64_t valid = addr->valid_until - now;
                        if (valid <= 0)
                                return -1;
+                       else if (valid > UINT32_MAX)
+                               valid = UINT32_MAX;
 
                        cinfo.ifa_valid = valid;
                }
@@ -964,18 +1320,18 @@ static int system_rt(struct device *dev, struct device_route *route, int cmd)
                        route->nexthop.in6.s6_addr32[2] ||
                        route->nexthop.in6.s6_addr32[3];
 
-       unsigned char scope = (cmd == RTM_DELROUTE) ? RT_SCOPE_NOWHERE :
-                       (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
-
-       unsigned int table = (route->flags & DEVROUTE_TABLE) ? route->table : RT_TABLE_MAIN;
+       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_src_len = route->sourcemask,
                .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
                .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
-               .rtm_scope = scope,
+               .rtm_scope = RT_SCOPE_NOWHERE,
                .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
+               .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
        };
        struct nl_msg *msg;
 
@@ -986,6 +1342,23 @@ static int system_rt(struct device *dev, struct device_route *route, int cmd)
                        rtm.rtm_scope = RT_SCOPE_UNIVERSE;
                        rtm.rtm_type = RTN_UNREACHABLE;
                }
+               else
+                       rtm.rtm_scope = (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
+       }
+
+       if (route->flags & DEVROUTE_TYPE) {
+               rtm.rtm_type = route->type;
+               if (!(route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))) {
+                       if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_BROADCAST ||
+                           rtm.rtm_type == RTN_NAT || rtm.rtm_type == RTN_ANYCAST)
+                               rtm.rtm_table = RT_TABLE_LOCAL;
+               }
+
+               if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_NAT)
+                       rtm.rtm_scope = RT_SCOPE_HOST;
+               else if (rtm.rtm_type == RTN_BROADCAST || rtm.rtm_type == RTN_MULTICAST ||
+                       rtm.rtm_type == RTN_ANYCAST)
+                       rtm.rtm_scope = RT_SCOPE_LINK;
        }
 
        msg = nlmsg_alloc_simple(cmd, flags);
@@ -997,6 +1370,9 @@ 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->sourcemask)
+               nla_put(msg, RTA_SRC, alen, &route->source);
+
        if (route->metric > 0)
                nla_put_u32(msg, RTA_PRIORITY, route->metric);
 
@@ -1009,7 +1385,22 @@ static int system_rt(struct device *dev, struct device_route *route, int cmd)
        if (table >= 256)
                nla_put_u32(msg, RTA_TABLE, table);
 
+       if (route->flags & DEVROUTE_MTU) {
+               struct nlattr *metrics;
+
+               if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
+                       goto nla_put_failure;
+
+               nla_put_u32(msg, RTAX_MTU, route->mtu);
+
+               nla_nest_end(msg, metrics);
+       }
+
        return system_rtnl_call(msg);
+
+nla_put_failure:
+       nlmsg_free(msg);
+       return -ENOMEM;
 }
 
 int system_add_route(struct device *dev, struct device_route *route)
@@ -1035,12 +1426,17 @@ int system_flush_routes(void)
                if (fd < 0)
                        continue;
 
-               write(fd, "-1", 2);
+               if (write(fd, "-1", 2)) {}
                close(fd);
        }
        return 0;
 }
 
+bool system_resolve_rt_type(const char *type, unsigned int *id)
+{
+       return system_rtn_aton(type, id);
+}
+
 bool system_resolve_rt_table(const char *name, unsigned int *id)
 {
        FILE *f;
@@ -1083,14 +1479,15 @@ bool system_resolve_rt_table(const char *name, unsigned int *id)
        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;
 }
 
+bool system_is_default_rt_table(unsigned int id)
+{
+       return (id == RT_TABLE_MAIN);
+}
+
 static int system_iprule(struct iprule *rule, int cmd)
 {
        int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
@@ -1226,39 +1623,7 @@ int system_flush_iprules(void)
 
 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 {
-               n = strtoul(action, &e, 0);
-               if (!e || *e || e == action || n > 255)
-                       return false;
-       }
-
-       *id = n;
-       return true;
+       return system_rtn_aton(action, id);
 }
 
 time_t system_get_rtime(void)
@@ -1279,14 +1644,6 @@ time_t system_get_rtime(void)
 #define IP_DF       0x4000
 #endif
 
-static void tunnel_parm_init(struct ip_tunnel_parm *p)
-{
-       memset(p, 0, sizeof(*p));
-       p->iph.version = 4;
-       p->iph.ihl = 5;
-       p->iph.frag_off = htons(IP_DF);
-}
-
 static int tunnel_ioctl(const char *name, int cmd, void *p)
 {
        struct ifreq ifr;
@@ -1299,10 +1656,7 @@ static int tunnel_ioctl(const char *name, int cmd, void *p)
 
 int system_del_ip_tunnel(const char *name)
 {
-       struct ip_tunnel_parm p;
-
-       tunnel_parm_init(&p);
-       return tunnel_ioctl(name, SIOCDELTUNNEL, &p);
+       return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
 }
 
 int system_update_ipv6_mtu(struct device *dev, int mtu)
@@ -1332,86 +1686,196 @@ out:
        return ret;
 }
 
-static int parse_ipaddr(struct blob_attr *attr, __be32 *addr)
-{
-       if (!attr)
-               return 1;
-
-       return inet_pton(AF_INET, blobmsg_data(attr), (void *) addr);
-}
-
-
 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
 {
        struct blob_attr *tb[__TUNNEL_ATTR_MAX];
        struct blob_attr *cur;
-       struct ip_tunnel_parm p;
-       const char *base, *str;
-       bool is_sit;
+       bool set_df = true;
+       const char *str;
 
        system_del_ip_tunnel(name);
 
-       tunnel_parm_init(&p);
-
        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);
-       is_sit = !strcmp(str, "sit");
-
-       if (is_sit) {
-               p.iph.protocol = IPPROTO_IPV6;
-               base = "sit0";
-       } else
-               return -EINVAL;
 
-       if (!parse_ipaddr(tb[TUNNEL_ATTR_LOCAL], &p.iph.saddr))
-               return -EINVAL;
-
-       if (!parse_ipaddr(tb[TUNNEL_ATTR_REMOTE], &p.iph.daddr))
-               return -EINVAL;
+       if ((cur = tb[TUNNEL_ATTR_DF]))
+               set_df = blobmsg_get_bool(cur);
 
+       unsigned int ttl = 0;
        if ((cur = tb[TUNNEL_ATTR_TTL])) {
-               unsigned int val = blobmsg_get_u32(cur);
+               ttl = blobmsg_get_u32(cur);
+               if (ttl > 255 || (!set_df && ttl))
+                       return -EINVAL;
+       }
 
-               if (val > 255)
+       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;
 
-               p.iph.ttl = val;
+               if (iface->l3_dev.dev)
+                       link = iface->l3_dev.dev->ifindex;
        }
 
-       strncpy(p.name, name, sizeof(p.name));
-       if (tunnel_ioctl(base, SIOCADDTUNNEL, &p) < 0)
-               return -1;
-
-#ifdef SIOCADD6RD
-       cur = tb[TUNNEL_ATTR_6RD_PREFIX];
-       if (cur && is_sit) {
-               unsigned int mask;
-               struct ip_tunnel_6rd p6;
+       if (!strcmp(str, "sit")) {
+               struct ip_tunnel_parm p = {
+                       .link = link,
+                       .iph = {
+                               .version = 4,
+                               .ihl = 5,
+                               .frag_off = set_df ? htons(IP_DF) : 0,
+                               .protocol = IPPROTO_IPV6,
+                               .ttl = ttl
+                       }
+               };
 
-               memset(&p6, 0, sizeof(p6));
+               if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
+                               inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
+                       return -EINVAL;
 
-               if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
-                                       &p6.prefix, &mask) || mask > 128)
+               if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
+                               inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
                        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)
+               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.relay_prefixlen = mask;
+                       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 nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
+                               NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
+               struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
+               int ret = 0;
 
-               if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
-                       system_del_ip_tunnel(name);
+               if (!nlm)
                        return -1;
+
+               nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
+               nla_put_string(nlm, IFLA_IFNAME, name);
+
+               if (link)
+                       nla_put_u32(nlm, IFLA_LINK, link);
+
+               struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
+               if (!linkinfo) {
+                       ret = -ENOMEM;
+                       goto failure;
+               }
+               nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
+               struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
+               if (!infodata) {
+                       ret = -ENOMEM;
+                       goto failure;
+               }
+
+               if (link)
+                       nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
+
+               nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
+               nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
+               nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
+
+               struct in6_addr in6buf;
+               if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
+                       if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
+                               ret = -EINVAL;
+                               goto failure;
+                       }
+                       nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
+               }
+
+               if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
+                       if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
+                               ret = -EINVAL;
+                               goto failure;
+                       }
+                       nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
+               }
+
+#ifdef IFLA_IPTUN_FMR_MAX
+               if ((cur = tb[TUNNEL_ATTR_FMRS])) {
+                       struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
+
+                       struct blob_attr *fmr;
+                       unsigned rem, fmrcnt = 0;
+                       blobmsg_for_each_attr(fmr, cur, rem) {
+                               if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
+                                       continue;
+
+                               unsigned ip4len, ip6len, ealen, offset = 6;
+                               char ip6buf[48];
+                               char ip4buf[16];
+
+                               if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
+                                               ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5) {
+                                       ret = -EINVAL;
+                                       goto failure;
+                               }
+
+                               struct in6_addr ip6prefix;
+                               struct in_addr ip4prefix;
+                               if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
+                                               inet_pton(AF_INET, ip4buf, &ip4prefix) != 1) {
+                                       ret = -EINVAL;
+                                       goto failure;
+                               }
+
+                               struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
+
+                               nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
+                               nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
+                               nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
+                               nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
+                               nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
+                               nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
+
+                               nla_nest_end(nlm, rule);
+                       }
+
+                       nla_nest_end(nlm, fmrs);
                }
-       }
 #endif
 
+               nla_nest_end(nlm, infodata);
+               nla_nest_end(nlm, linkinfo);
+
+               return system_rtnl_call(nlm);
+failure:
+               nlmsg_free(nlm);
+               return ret;
+       }
+       else
+               return -EINVAL;
+
        return 0;
 }