disable ipv6 for bridge member interfaces
[project/netifd.git] / system-linux.c
index f1acc19..2425445 100644 (file)
@@ -1,14 +1,23 @@
+#define _GNU_SOURCE
+
 #include <sys/socket.h>
 #include <sys/ioctl.h>
+#include <sys/stat.h>
+#include <sys/syscall.h>
 
 #include <linux/rtnetlink.h>
 #include <linux/sockios.h>
 #include <linux/if_vlan.h>
+#include <linux/if_bridge.h>
 
+#include <unistd.h>
 #include <string.h>
 #include <fcntl.h>
+#include <glob.h>
 
 #include <netlink/msg.h>
+#include <netlink/attr.h>
+#include <netlink/socket.h>
 #include <libubox/uloop.h>
 
 #include "netifd.h"
@@ -30,42 +39,46 @@ int system_init(void)
        fcntl(sock_ioctl, F_SETFD, fcntl(sock_ioctl, F_GETFD) | FD_CLOEXEC);
 
        // Prepare socket for routing / address control
-       if ((sock_rtnl = nl_socket_alloc())) {
-               if (nl_connect(sock_rtnl, NETLINK_ROUTE)) {
-                       nl_socket_free(sock_rtnl);
-                       sock_rtnl = NULL;
-               }
-       }
+       sock_rtnl = nl_socket_alloc();
+       if (!sock_rtnl)
+               return -1;
+
+       if (nl_connect(sock_rtnl, NETLINK_ROUTE))
+               goto error_free_sock;
 
        // Prepare socket for link events
-       if ((nl_cb_rtnl_event = nl_cb_alloc(NL_CB_DEFAULT)))
-               nl_cb_set(nl_cb_rtnl_event, NL_CB_VALID, NL_CB_CUSTOM,
-                                                       cb_rtnl_event, NULL);
-
-       if (nl_cb_rtnl_event && (sock_rtnl_event = nl_socket_alloc())) {
-               if (nl_connect(sock_rtnl_event, NETLINK_ROUTE)) {
-                       nl_socket_free(sock_rtnl_event);
-                       sock_rtnl_event = NULL;
-               }
-               // Receive network link events form kernel
-               nl_socket_add_membership(sock_rtnl_event, RTNLGRP_LINK);
-
-               // Synthesize initial link messages
-               struct nl_msg *m = nlmsg_alloc_simple(RTM_GETLINK, NLM_F_DUMP);
-               if (m && nlmsg_reserve(m, sizeof(struct ifinfomsg), 0)) {
-                       nl_send_auto_complete(sock_rtnl_event, m);
-                       nlmsg_free(m);
-               }
+       nl_cb_rtnl_event = nl_cb_alloc(NL_CB_DEFAULT);
+       if (!nl_cb_rtnl_event)
+               goto error_free_sock;
 
-#ifdef NLA_PUT_DATA
-               rtnl_event.fd = nl_socket_get_fd(sock_rtnl_event);
-#else
-               rtnl_event.fd = sock_rtnl_event->s_fd; // libnl-tiny hack...
-#endif
-               uloop_fd_add(&rtnl_event, ULOOP_READ | ULOOP_EDGE_TRIGGER);
-       }
+       nl_cb_set(nl_cb_rtnl_event, NL_CB_VALID, NL_CB_CUSTOM,
+                 cb_rtnl_event, NULL);
+
+       sock_rtnl_event = nl_socket_alloc();
+       if (!sock_rtnl_event)
+               goto error_free_cb;
+
+       if (nl_connect(sock_rtnl_event, NETLINK_ROUTE))
+               goto error_free_event;
+
+       // Receive network link events form kernel
+       nl_socket_add_membership(sock_rtnl_event, RTNLGRP_LINK);
 
-       return -(sock_ioctl < 0 || !sock_rtnl);
+       rtnl_event.fd = nl_socket_get_fd(sock_rtnl_event);
+       uloop_fd_add(&rtnl_event, ULOOP_READ | ULOOP_EDGE_TRIGGER);
+
+       return 0;
+
+error_free_event:
+       nl_socket_free(sock_rtnl_event);
+       sock_rtnl_event = NULL;
+error_free_cb:
+       nl_cb_put(nl_cb_rtnl_event);
+       nl_cb_rtnl_event = NULL;
+error_free_sock:
+       nl_socket_free(sock_rtnl);
+       sock_rtnl = NULL;
+       return -1;
 }
 
 // If socket is ready for reading parse netlink events
@@ -74,6 +87,26 @@ static void handler_rtnl_event(struct uloop_fd *u, unsigned int events)
        nl_recvmsgs(sock_rtnl_event, nl_cb_rtnl_event);
 }
 
+static void system_set_sysctl(const char *path, const char *val)
+{
+       int fd;
+
+       fd = open(path, O_WRONLY);
+       if (fd < 0)
+               return;
+
+       write(fd, val, strlen(val));
+       close(fd);
+}
+
+static void system_set_disable_ipv6(struct device *dev, const char *val)
+{
+       char buf[256];
+
+       snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname);
+       system_set_sysctl(buf, "0");
+}
+
 // Evaluate netlink messages
 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
 {
@@ -107,88 +140,386 @@ static int system_rtnl_call(struct nl_msg *msg)
        return s;
 }
 
-int system_bridge_addbr(struct device *bridge)
-{
-       return ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname);
-}
-
 int system_bridge_delbr(struct device *bridge)
 {
        return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
 }
 
-static int system_bridge_if(struct device *bridge, struct device *dev, int cmd)
+static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
 {
        struct ifreq ifr;
-       ifr.ifr_ifindex = dev->ifindex;
-       strncpy(ifr.ifr_name, bridge->ifname, sizeof(ifr.ifr_name));
+       if (dev)
+               ifr.ifr_ifindex = dev->ifindex;
+       else
+               ifr.ifr_data = data;
+       strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
        return ioctl(sock_ioctl, cmd, &ifr);
 }
 
 int system_bridge_addif(struct device *bridge, struct device *dev)
 {
-       return system_bridge_if(bridge, dev, SIOCBRADDIF);
+       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)
 {
-       return system_bridge_if(bridge, dev, SIOCBRDELIF);
+       system_set_disable_ipv6(dev, "0");
+       return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
 }
 
-static int system_vlan(struct device *dev, int id)
+static bool system_is_bridge(const char *name, char *buf, int buflen)
 {
-       struct vlan_ioctl_args ifr = {
-               .cmd = (id == 0) ? DEL_VLAN_CMD : ADD_VLAN_CMD,
-               .u = {.VID = id},
-       };
-       strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
-       return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
+       struct stat st;
+
+       snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
+       if (stat(buf, &st) < 0)
+               return false;
+
+       return true;
 }
 
-int system_vlan_add(struct device *dev, int id)
+static char *system_get_bridge(const char *name, char *buf, int buflen)
 {
-       return system_vlan(dev, id);
+       char *path;
+       ssize_t len;
+       glob_t gl;
+
+       snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
+       if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
+               return NULL;
+
+       if (gl.gl_pathc == 0)
+               return NULL;
+
+       len = readlink(gl.gl_pathv[0], buf, buflen);
+       if (len < 0)
+               return NULL;
+
+       buf[len] = 0;
+       path = strrchr(buf, '/');
+       if (!path)
+               return NULL;
+
+       return path + 1;
 }
 
-int system_vlan_del(struct device *dev)
+static int system_if_resolve(struct device *dev)
 {
-       return system_vlan(dev, 0);
+       struct ifreq ifr;
+       strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
+       if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
+               return ifr.ifr_ifindex;
+       else
+               return 0;
 }
 
-static int system_if_flags(struct device *dev, unsigned add, unsigned rem)
+static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
 {
        struct ifreq ifr;
-       strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
+       strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
        ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
        ifr.ifr_flags |= add;
        ifr.ifr_flags &= ~rem;
        return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
 }
 
-static int system_if_resolve(struct device *dev)
+struct clear_data {
+       struct nl_msg *msg;
+       struct device *dev;
+       int type;
+       int size;
+       int af;
+};
+
+
+static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
 {
-       struct ifreq ifr;
-       strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
-       if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
-               return ifr.ifr_ifindex;
-       else
-               return 0;
+       struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
+
+       return ifa->ifa_index == ifindex;
+}
+
+static bool check_route(struct nlmsghdr *hdr, int ifindex)
+{
+       struct nlattr *tb[__RTA_MAX];
+
+       nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
+       if (!tb[RTA_OIF])
+               return false;
+
+       return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
+}
+
+static int cb_clear_event(struct nl_msg *msg, void *arg)
+{
+       struct clear_data *clr = arg;
+       struct nlmsghdr *hdr = nlmsg_hdr(msg);
+       bool (*cb)(struct nlmsghdr *, int ifindex);
+       int type;
+
+       switch(clr->type) {
+       case RTM_GETADDR:
+               type = RTM_DELADDR;
+               if (hdr->nlmsg_type != RTM_NEWADDR)
+                       return NL_SKIP;
+
+               cb = check_ifaddr;
+               break;
+       case RTM_GETROUTE:
+               type = RTM_DELROUTE;
+               if (hdr->nlmsg_type != RTM_NEWROUTE)
+                       return NL_SKIP;
+
+               cb = check_route;
+               break;
+       default:
+               return NL_SKIP;
+       }
+
+       if (!cb(hdr, clr->dev->ifindex))
+               return NL_SKIP;
+
+       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;
+       hdr->nlmsg_flags = NLM_F_REQUEST;
+
+       if (!nl_send_auto_complete(sock_rtnl, clr->msg))
+               nl_wait_for_ack(sock_rtnl);
+
+       return NL_SKIP;
+}
+
+static int
+cb_finish_event(struct nl_msg *msg, void *arg)
+{
+       int *pending = arg;
+       *pending = 0;
+       return NL_STOP;
+}
+
+static int
+error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
+{
+       int *pending = arg;
+       *pending = err->error;
+       return NL_STOP;
+}
+
+static void
+system_if_clear_entries(struct device *dev, int type, int af)
+{
+       struct clear_data clr;
+       struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
+       struct rtmsg rtm = {
+               .rtm_family = af,
+               .rtm_flags = RTM_F_CLONED,
+       };
+       int flags = NLM_F_DUMP;
+       int pending = 1;
+
+       clr.af = af;
+       clr.dev = dev;
+       clr.type = type;
+       switch (type) {
+       case RTM_GETADDR:
+               clr.size = sizeof(struct rtgenmsg);
+               break;
+       case RTM_GETROUTE:
+               clr.size = sizeof(struct rtmsg);
+               break;
+       default:
+               return;
+       }
+
+       if (!cb)
+               return;
+
+       clr.msg = nlmsg_alloc_simple(type, flags);
+       if (!clr.msg)
+               goto out;
+
+       nlmsg_append(clr.msg, &rtm, clr.size, 0);
+       nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
+       nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
+       nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
+
+       nl_send_auto_complete(sock_rtnl, clr.msg);
+       while (pending > 0)
+               nl_recvmsgs(sock_rtnl, cb);
+
+       nlmsg_free(clr.msg);
+out:
+       nl_cb_put(cb);
+}
+
+/*
+ * Clear bridge (membership) state and bring down device
+ */
+void system_if_clear_state(struct device *dev)
+{
+       char buf[256];
+       char *bridge;
+
+       dev->ifindex = system_if_resolve(dev);
+       if (!dev->ifindex)
+               return;
+
+       system_if_flags(dev->ifname, 0, IFF_UP);
+
+       if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
+               D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
+               system_bridge_delbr(dev);
+               return;
+       }
+
+       bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
+       if (bridge) {
+               D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
+               system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
+       }
+
+       system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
+       system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
+       system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
+       system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
+       system_set_disable_ipv6(dev, "0");
+}
+
+static inline unsigned long
+sec_to_jiffies(int val)
+{
+       return (unsigned long) val * 100;
+}
+
+int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
+{
+       unsigned long args[4] = {};
+
+       if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
+               return -1;
+
+       args[0] = BRCTL_SET_BRIDGE_STP_STATE;
+       args[1] = !!cfg->stp;
+       system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
+
+       args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
+       args[1] = sec_to_jiffies(cfg->forward_delay);
+       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);
+               system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
+       }
+
+       if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
+               args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
+               args[1] = sec_to_jiffies(cfg->hello_time);
+               system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
+       }
+
+       if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
+               args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
+               args[1] = sec_to_jiffies(cfg->max_age);
+               system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
+       }
+
+       return 0;
+}
+
+static int system_vlan(struct device *dev, int id)
+{
+       struct vlan_ioctl_args ifr = {
+               .cmd = SET_VLAN_NAME_TYPE_CMD,
+               .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
+       };
+
+       ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
+
+       if (id < 0) {
+               ifr.cmd = DEL_VLAN_CMD;
+               ifr.u.VID = 0;
+       } else {
+               ifr.cmd = ADD_VLAN_CMD;
+               ifr.u.VID = id;
+       }
+       strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
+       return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
+}
+
+int system_vlan_add(struct device *dev, int id)
+{
+       return system_vlan(dev, id);
+}
+
+int system_vlan_del(struct device *dev)
+{
+       return system_vlan(dev, -1);
 }
 
 int system_if_up(struct device *dev)
 {
        dev->ifindex = system_if_resolve(dev);
-       return system_if_flags(dev, IFF_UP, 0);
+       return system_if_flags(dev->ifname, IFF_UP, 0);
 }
 
 int system_if_down(struct device *dev)
 {
-       return system_if_flags(dev, 0, IFF_UP);
+       return system_if_flags(dev->ifname, 0, IFF_UP);
 }
 
 int system_if_check(struct device *dev)
 {
-       return -!(system_if_resolve(dev));
+       device_set_present(dev, (system_if_resolve(dev) >= 0));
+       return 0;
+}
+
+int system_if_dump_stats(struct device *dev, struct blob_buf *b)
+{
+       const char *const counters[] = {
+               "collisions",     "rx_frame_errors",   "tx_compressed",
+               "multicast",      "rx_length_errors",  "tx_dropped",
+               "rx_bytes",       "rx_missed_errors",  "tx_errors",
+               "rx_compressed",  "rx_over_errors",    "tx_fifo_errors",
+               "rx_crc_errors",  "rx_packets",        "tx_heartbeat_errors",
+               "rx_dropped",     "tx_aborted_errors", "tx_packets",
+               "rx_errors",      "tx_bytes",          "tx_window_errors",
+               "rx_fifo_errors", "tx_carrier_errors",
+       };
+       char buf[64];
+       int stats_dir;
+       int i, fd, len;
+
+       snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
+       stats_dir = open(buf, O_DIRECTORY);
+       if (stats_dir < 0)
+               return -1;
+
+       for (i = 0; i < ARRAY_SIZE(counters); i++) {
+               fd = openat(stats_dir, counters[i], O_RDONLY);
+               if (fd < 0)
+                       continue;
+
+retry:
+               len = read(fd, buf, sizeof(buf));
+               if (len < 0) {
+                       if (errno == EINTR)
+                               goto retry;
+                       continue;
+               }
+
+               buf[len] = 0;
+               blobmsg_add_u32(b, counters[i], strtoul(buf, NULL, 0));
+               close(fd);
+       }
+
+       close(stats_dir);
+       return 0;
 }
 
 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
@@ -223,6 +554,7 @@ 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;
 
        if (alen == 4)
                have_gw = !!route->nexthop.in.s_addr;
@@ -244,7 +576,10 @@ static int system_rt(struct device *dev, struct device_route *route, int cmd)
                .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
        };
 
-       struct nl_msg *msg = nlmsg_alloc_simple(cmd, 0);
+       if (cmd == RTM_NEWROUTE)
+               flags |= NLM_F_CREATE | NLM_F_REPLACE;
+
+       struct nl_msg *msg = nlmsg_alloc_simple(cmd, flags);
        if (!msg)
                return -1;
 
@@ -271,3 +606,36 @@ int system_del_route(struct device *dev, struct device_route *route)
 {
        return system_rt(dev, route, RTM_DELROUTE);
 }
+
+int system_flush_routes(void)
+{
+       const char *names[] = {
+               "/proc/sys/net/ipv4/route/flush",
+               "/proc/sys/net/ipv6/route/flush"
+       };
+       int fd, i;
+
+       for (i = 0; i < ARRAY_SIZE(names); i++) {
+               fd = open(names[i], O_WRONLY);
+               if (fd < 0)
+                       continue;
+
+               write(fd, "-1", 2);
+               close(fd);
+       }
+       return 0;
+}
+
+time_t system_get_rtime(void)
+{
+       struct timespec ts;
+       struct timeval tv;
+
+       if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
+               return ts.tv_sec;
+
+       if (gettimeofday(&tv, NULL) == 0)
+               return tv.tv_sec;
+
+       return 0;
+}