+ VLOG_INFO("RTM_GETLINK failed (%s), obtaining netdev stats "
+ "via proc (you are probably running a pre-2.6.19 "
+ "kernel)", strerror(error));
+ return false;
+ }
+ }
+}
+
+/* Retrieves current device stats for 'netdev'.
+ *
+ * XXX All of the members of struct netdev_stats are 64 bits wide, but on
+ * 32-bit architectures the Linux network stats are only 32 bits. */
+static int
+netdev_linux_get_stats(const struct netdev *netdev, struct netdev_stats *stats)
+{
+ static int use_netlink_stats = -1;
+ int error;
+
+ COVERAGE_INC(netdev_get_stats);
+ if (use_netlink_stats < 0) {
+ use_netlink_stats = check_for_working_netlink_stats();
+ }
+ if (use_netlink_stats) {
+ int ifindex;
+
+ error = get_ifindex(netdev, &ifindex);
+ if (!error) {
+ error = get_stats_via_netlink(ifindex, stats);
+ }
+ } else {
+ error = get_stats_via_proc(netdev->name, stats);
+ }
+ return error;
+}
+
+/* Stores the features supported by 'netdev' into each of '*current',
+ * '*advertised', '*supported', and '*peer' that are non-null. Each value is a
+ * bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
+ * successful, otherwise a positive errno value. On failure, all of the
+ * passed-in values are set to 0. */
+static int
+netdev_linux_get_features(struct netdev *netdev,
+ uint32_t *current, uint32_t *advertised,
+ uint32_t *supported, uint32_t *peer)
+{
+ struct ethtool_cmd ecmd;
+ int error;
+
+ memset(&ecmd, 0, sizeof ecmd);
+ error = netdev_linux_do_ethtool(netdev, &ecmd,
+ ETHTOOL_GSET, "ETHTOOL_GSET");
+ if (error) {
+ return error;
+ }
+
+ /* Supported features. */
+ *supported = 0;
+ if (ecmd.supported & SUPPORTED_10baseT_Half) {
+ *supported |= OFPPF_10MB_HD;
+ }
+ if (ecmd.supported & SUPPORTED_10baseT_Full) {
+ *supported |= OFPPF_10MB_FD;
+ }
+ if (ecmd.supported & SUPPORTED_100baseT_Half) {
+ *supported |= OFPPF_100MB_HD;
+ }
+ if (ecmd.supported & SUPPORTED_100baseT_Full) {
+ *supported |= OFPPF_100MB_FD;
+ }
+ if (ecmd.supported & SUPPORTED_1000baseT_Half) {
+ *supported |= OFPPF_1GB_HD;
+ }
+ if (ecmd.supported & SUPPORTED_1000baseT_Full) {
+ *supported |= OFPPF_1GB_FD;
+ }
+ if (ecmd.supported & SUPPORTED_10000baseT_Full) {
+ *supported |= OFPPF_10GB_FD;
+ }
+ if (ecmd.supported & SUPPORTED_TP) {
+ *supported |= OFPPF_COPPER;
+ }
+ if (ecmd.supported & SUPPORTED_FIBRE) {
+ *supported |= OFPPF_FIBER;
+ }
+ if (ecmd.supported & SUPPORTED_Autoneg) {
+ *supported |= OFPPF_AUTONEG;
+ }
+ if (ecmd.supported & SUPPORTED_Pause) {
+ *supported |= OFPPF_PAUSE;
+ }
+ if (ecmd.supported & SUPPORTED_Asym_Pause) {
+ *supported |= OFPPF_PAUSE_ASYM;
+ }
+
+ /* Advertised features. */
+ *advertised = 0;
+ if (ecmd.advertising & ADVERTISED_10baseT_Half) {
+ *advertised |= OFPPF_10MB_HD;
+ }
+ if (ecmd.advertising & ADVERTISED_10baseT_Full) {
+ *advertised |= OFPPF_10MB_FD;
+ }
+ if (ecmd.advertising & ADVERTISED_100baseT_Half) {
+ *advertised |= OFPPF_100MB_HD;
+ }
+ if (ecmd.advertising & ADVERTISED_100baseT_Full) {
+ *advertised |= OFPPF_100MB_FD;
+ }
+ if (ecmd.advertising & ADVERTISED_1000baseT_Half) {
+ *advertised |= OFPPF_1GB_HD;
+ }
+ if (ecmd.advertising & ADVERTISED_1000baseT_Full) {
+ *advertised |= OFPPF_1GB_FD;
+ }
+ if (ecmd.advertising & ADVERTISED_10000baseT_Full) {
+ *advertised |= OFPPF_10GB_FD;
+ }
+ if (ecmd.advertising & ADVERTISED_TP) {
+ *advertised |= OFPPF_COPPER;
+ }
+ if (ecmd.advertising & ADVERTISED_FIBRE) {
+ *advertised |= OFPPF_FIBER;
+ }
+ if (ecmd.advertising & ADVERTISED_Autoneg) {
+ *advertised |= OFPPF_AUTONEG;
+ }
+ if (ecmd.advertising & ADVERTISED_Pause) {
+ *advertised |= OFPPF_PAUSE;
+ }
+ if (ecmd.advertising & ADVERTISED_Asym_Pause) {
+ *advertised |= OFPPF_PAUSE_ASYM;
+ }
+
+ /* Current settings. */
+ if (ecmd.speed == SPEED_10) {
+ *current = ecmd.duplex ? OFPPF_10MB_FD : OFPPF_10MB_HD;
+ } else if (ecmd.speed == SPEED_100) {
+ *current = ecmd.duplex ? OFPPF_100MB_FD : OFPPF_100MB_HD;
+ } else if (ecmd.speed == SPEED_1000) {
+ *current = ecmd.duplex ? OFPPF_1GB_FD : OFPPF_1GB_HD;
+ } else if (ecmd.speed == SPEED_10000) {
+ *current = OFPPF_10GB_FD;
+ } else {
+ *current = 0;
+ }
+
+ if (ecmd.port == PORT_TP) {
+ *current |= OFPPF_COPPER;
+ } else if (ecmd.port == PORT_FIBRE) {
+ *current |= OFPPF_FIBER;
+ }
+
+ if (ecmd.autoneg) {
+ *current |= OFPPF_AUTONEG;
+ }
+
+ /* Peer advertisements. */
+ *peer = 0; /* XXX */
+
+ return 0;
+}
+
+/* Set the features advertised by 'netdev' to 'advertise'. */
+static int
+netdev_linux_set_advertisements(struct netdev *netdev, uint32_t advertise)
+{
+ struct ethtool_cmd ecmd;
+ int error;
+
+ memset(&ecmd, 0, sizeof ecmd);
+ error = netdev_linux_do_ethtool(netdev, &ecmd,
+ ETHTOOL_GSET, "ETHTOOL_GSET");
+ if (error) {
+ return error;
+ }
+
+ ecmd.advertising = 0;
+ if (advertise & OFPPF_10MB_HD) {
+ ecmd.advertising |= ADVERTISED_10baseT_Half;
+ }
+ if (advertise & OFPPF_10MB_FD) {
+ ecmd.advertising |= ADVERTISED_10baseT_Full;
+ }
+ if (advertise & OFPPF_100MB_HD) {
+ ecmd.advertising |= ADVERTISED_100baseT_Half;
+ }
+ if (advertise & OFPPF_100MB_FD) {
+ ecmd.advertising |= ADVERTISED_100baseT_Full;
+ }
+ if (advertise & OFPPF_1GB_HD) {
+ ecmd.advertising |= ADVERTISED_1000baseT_Half;
+ }
+ if (advertise & OFPPF_1GB_FD) {
+ ecmd.advertising |= ADVERTISED_1000baseT_Full;
+ }
+ if (advertise & OFPPF_10GB_FD) {
+ ecmd.advertising |= ADVERTISED_10000baseT_Full;
+ }
+ if (advertise & OFPPF_COPPER) {
+ ecmd.advertising |= ADVERTISED_TP;
+ }
+ if (advertise & OFPPF_FIBER) {
+ ecmd.advertising |= ADVERTISED_FIBRE;
+ }
+ if (advertise & OFPPF_AUTONEG) {
+ ecmd.advertising |= ADVERTISED_Autoneg;
+ }
+ if (advertise & OFPPF_PAUSE) {
+ ecmd.advertising |= ADVERTISED_Pause;
+ }
+ if (advertise & OFPPF_PAUSE_ASYM) {
+ ecmd.advertising |= ADVERTISED_Asym_Pause;
+ }
+ return netdev_linux_do_ethtool(netdev, &ecmd,
+ ETHTOOL_SSET, "ETHTOOL_SSET");
+}
+
+/* If 'netdev_name' is the name of a VLAN network device (e.g. one created with
+ * vconfig(8)), sets '*vlan_vid' to the VLAN VID associated with that device
+ * and returns 0. Otherwise returns a errno value (specifically ENOENT if
+ * 'netdev_name' is the name of a network device that is not a VLAN device) and
+ * sets '*vlan_vid' to -1. */
+static int
+netdev_linux_get_vlan_vid(const struct netdev *netdev, int *vlan_vid)
+{
+ const char *netdev_name = netdev_get_name(netdev);
+ struct ds line = DS_EMPTY_INITIALIZER;
+ FILE *stream = NULL;
+ int error;
+ char *fn;
+
+ COVERAGE_INC(netdev_get_vlan_vid);
+ fn = xasprintf("/proc/net/vlan/%s", netdev_name);
+ stream = fopen(fn, "r");
+ if (!stream) {
+ error = errno;
+ goto done;
+ }
+
+ if (ds_get_line(&line, stream)) {
+ if (ferror(stream)) {
+ error = errno;
+ VLOG_ERR_RL(&rl, "error reading \"%s\": %s", fn, strerror(errno));
+ } else {
+ error = EPROTO;
+ VLOG_ERR_RL(&rl, "unexpected end of file reading \"%s\"", fn);
+ }
+ goto done;
+ }
+
+ if (!sscanf(ds_cstr(&line), "%*s VID: %d", vlan_vid)) {
+ error = EPROTO;
+ VLOG_ERR_RL(&rl, "parse error reading \"%s\" line 1: \"%s\"",
+ fn, ds_cstr(&line));
+ goto done;
+ }
+
+ error = 0;
+
+done:
+ free(fn);
+ if (stream) {
+ fclose(stream);
+ }
+ ds_destroy(&line);
+ if (error) {
+ *vlan_vid = -1;
+ }
+ return error;
+}
+
+#define POLICE_ADD_CMD "/sbin/tc qdisc add dev %s handle ffff: ingress"
+#define POLICE_CONFIG_CMD "/sbin/tc filter add dev %s parent ffff: protocol ip prio 50 u32 match ip src 0.0.0.0/0 police rate %dkbit burst %dk mtu 65535 drop flowid :1"
+/* We redirect stderr to /dev/null because we often want to remove all
+ * traffic control configuration on a port so its in a known state. If
+ * this done when there is no such configuration, tc complains, so we just
+ * always ignore it.
+ */
+#define POLICE_DEL_CMD "/sbin/tc qdisc del dev %s handle ffff: ingress 2>/dev/null"
+
+/* Attempts to set input rate limiting (policing) policy. */
+static int
+netdev_linux_set_policing(struct netdev *netdev,
+ uint32_t kbits_rate, uint32_t kbits_burst)
+{
+ const char *netdev_name = netdev_get_name(netdev);
+ char command[1024];
+
+ COVERAGE_INC(netdev_set_policing);
+ if (kbits_rate) {
+ if (!kbits_burst) {
+ /* Default to 10 kilobits if not specified. */
+ kbits_burst = 10;
+ }
+
+ /* xxx This should be more careful about only adding if it
+ * xxx actually exists, as opposed to always deleting it. */
+ snprintf(command, sizeof(command), POLICE_DEL_CMD, netdev_name);
+ if (system(command) == -1) {
+ VLOG_WARN_RL(&rl, "%s: problem removing policing", netdev_name);
+ }
+
+ snprintf(command, sizeof(command), POLICE_ADD_CMD, netdev_name);
+ if (system(command) != 0) {
+ VLOG_WARN_RL(&rl, "%s: problem adding policing", netdev_name);
+ return -1;
+ }
+
+ snprintf(command, sizeof(command), POLICE_CONFIG_CMD, netdev_name,
+ kbits_rate, kbits_burst);
+ if (system(command) != 0) {
+ VLOG_WARN_RL(&rl, "%s: problem configuring policing",
+ netdev_name);
+ return -1;
+ }
+ } else {
+ snprintf(command, sizeof(command), POLICE_DEL_CMD, netdev_name);
+ if (system(command) == -1) {
+ VLOG_WARN_RL(&rl, "%s: problem removing policing", netdev_name);
+ }
+ }
+
+ return 0;
+}
+
+static int
+netdev_linux_get_in4(const struct netdev *netdev_,
+ struct in_addr *address, struct in_addr *netmask)
+{
+ struct netdev_linux *netdev = netdev_linux_cast(netdev_);
+ if (!(netdev->cache->valid & VALID_IN4)) {
+ int error;
+
+ error = netdev_linux_get_ipv4(netdev_, &netdev->cache->address,
+ SIOCGIFADDR, "SIOCGIFADDR");
+ if (error) {
+ return error;
+ }
+
+ error = netdev_linux_get_ipv4(netdev_, &netdev->cache->netmask,
+ SIOCGIFNETMASK, "SIOCGIFNETMASK");
+ if (error) {
+ return error;
+ }
+
+ netdev->cache->valid |= VALID_IN4;
+ }
+ *address = netdev->cache->address;
+ *netmask = netdev->cache->netmask;
+ return address->s_addr == INADDR_ANY ? EADDRNOTAVAIL : 0;
+}
+
+static int
+netdev_linux_set_in4(struct netdev *netdev_, struct in_addr address,
+ struct in_addr netmask)
+{
+ struct netdev_linux *netdev = netdev_linux_cast(netdev_);
+ int error;
+
+ error = do_set_addr(netdev_, SIOCSIFADDR, "SIOCSIFADDR", address);
+ if (!error) {
+ netdev->cache->valid |= VALID_IN4;
+ netdev->cache->address = address;
+ netdev->cache->netmask = netmask;
+ if (address.s_addr != INADDR_ANY) {
+ error = do_set_addr(netdev_, SIOCSIFNETMASK,
+ "SIOCSIFNETMASK", netmask);
+ }
+ }
+ return error;
+}
+
+static bool
+parse_if_inet6_line(const char *line,
+ struct in6_addr *in6, char ifname[16 + 1])
+{
+ uint8_t *s6 = in6->s6_addr;
+#define X8 "%2"SCNx8
+ return sscanf(line,
+ " "X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8 X8
+ "%*x %*x %*x %*x %16s\n",
+ &s6[0], &s6[1], &s6[2], &s6[3],
+ &s6[4], &s6[5], &s6[6], &s6[7],
+ &s6[8], &s6[9], &s6[10], &s6[11],
+ &s6[12], &s6[13], &s6[14], &s6[15],
+ ifname) == 17;
+}
+
+/* If 'netdev' has an assigned IPv6 address, sets '*in6' to that address (if
+ * 'in6' is non-null) and returns true. Otherwise, returns false. */
+static int
+netdev_linux_get_in6(const struct netdev *netdev_, struct in6_addr *in6)
+{
+ struct netdev_linux *netdev = netdev_linux_cast(netdev_);
+ if (!(netdev->cache->valid & VALID_IN6)) {
+ FILE *file;
+ char line[128];
+
+ netdev->cache->in6 = in6addr_any;
+
+ file = fopen("/proc/net/if_inet6", "r");
+ if (file != NULL) {
+ const char *name = netdev_get_name(netdev_);
+ while (fgets(line, sizeof line, file)) {
+ struct in6_addr in6;
+ char ifname[16 + 1];
+ if (parse_if_inet6_line(line, &in6, ifname)
+ && !strcmp(name, ifname))
+ {
+ netdev->cache->in6 = in6;
+ break;
+ }
+ }
+ fclose(file);
+ }
+ netdev->cache->valid |= VALID_IN6;
+ }
+ *in6 = netdev->cache->in6;
+ return 0;
+}
+
+static void
+make_in4_sockaddr(struct sockaddr *sa, struct in_addr addr)
+{
+ struct sockaddr_in sin;
+ memset(&sin, 0, sizeof sin);
+ sin.sin_family = AF_INET;
+ sin.sin_addr = addr;
+ sin.sin_port = 0;
+
+ memset(sa, 0, sizeof *sa);
+ memcpy(sa, &sin, sizeof sin);
+}
+
+static int
+do_set_addr(struct netdev *netdev,
+ int ioctl_nr, const char *ioctl_name, struct in_addr addr)
+{
+ struct ifreq ifr;
+ strncpy(ifr.ifr_name, netdev->name, sizeof ifr.ifr_name);
+ make_in4_sockaddr(&ifr.ifr_addr, addr);
+ return netdev_linux_do_ioctl(netdev, &ifr, ioctl_nr, ioctl_name);
+}
+
+/* Adds 'router' as a default IP gateway. */
+static int
+netdev_linux_add_router(struct netdev *netdev UNUSED, struct in_addr router)
+{
+ struct in_addr any = { INADDR_ANY };
+ struct rtentry rt;
+ int error;
+
+ memset(&rt, 0, sizeof rt);
+ make_in4_sockaddr(&rt.rt_dst, any);
+ make_in4_sockaddr(&rt.rt_gateway, router);
+ make_in4_sockaddr(&rt.rt_genmask, any);
+ rt.rt_flags = RTF_UP | RTF_GATEWAY;
+ COVERAGE_INC(netdev_add_router);
+ error = ioctl(af_inet_sock, SIOCADDRT, &rt) < 0 ? errno : 0;
+ if (error) {
+ VLOG_WARN("ioctl(SIOCADDRT): %s", strerror(error));
+ }
+ return error;
+}
+
+static int
+netdev_linux_get_next_hop(const struct in_addr *host, struct in_addr *next_hop,
+ char **netdev_name)
+{
+ static const char fn[] = "/proc/net/route";
+ FILE *stream;
+ char line[256];
+ int ln;
+
+ *netdev_name = NULL;
+ stream = fopen(fn, "r");
+ if (stream == NULL) {
+ VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(errno));
+ return errno;
+ }
+
+ ln = 0;
+ while (fgets(line, sizeof line, stream)) {
+ if (++ln >= 2) {
+ char iface[17];
+ uint32_t dest, gateway, mask;
+ int refcnt, metric, mtu;
+ unsigned int flags, use, window, irtt;
+
+ if (sscanf(line,
+ "%16s %"SCNx32" %"SCNx32" %04X %d %u %d %"SCNx32
+ " %d %u %u\n",
+ iface, &dest, &gateway, &flags, &refcnt,
+ &use, &metric, &mask, &mtu, &window, &irtt) != 11) {
+
+ VLOG_WARN_RL(&rl, "%s: could not parse line %d: %s",
+ fn, ln, line);
+ continue;
+ }
+ if (!(flags & RTF_UP)) {
+ /* Skip routes that aren't up. */
+ continue;
+ }
+
+ /* The output of 'dest', 'mask', and 'gateway' were given in
+ * network byte order, so we don't need need any endian
+ * conversions here. */
+ if ((dest & mask) == (host->s_addr & mask)) {
+ if (!gateway) {
+ /* The host is directly reachable. */
+ next_hop->s_addr = 0;
+ } else {
+ /* To reach the host, we must go through a gateway. */
+ next_hop->s_addr = gateway;
+ }
+ *netdev_name = xstrdup(iface);
+ fclose(stream);
+ return 0;