bridge: support binding vlan dev link state to vlan member bridge ports
authorMike Manning <mmanning@vyatta.att-mail.com>
Thu, 18 Apr 2019 17:35:33 +0000 (18:35 +0100)
committerDavid S. Miller <davem@davemloft.net>
Fri, 19 Apr 2019 20:58:17 +0000 (13:58 -0700)
In the case of vlan filtering on bridges, the bridge may also have the
corresponding vlan devices as upper devices. A vlan bridge binding mode
is added to allow the link state of the vlan device to track only the
state of the subset of bridge ports that are also members of the vlan,
rather than that of all bridge ports. This mode is set with a vlan flag
rather than a bridge sysfs so that the 8021q module is aware that it
should not set the link state for the vlan device.

If bridge vlan is configured, the bridge device event handling results
in the link state for an upper device being set, if it is a vlan device
with the vlan bridge binding mode enabled. This also sets a
vlan_bridge_binding flag so that subsequent UP/DOWN/CHANGE events for
the ports in that bridge result in a link state update of the vlan
device if required.

The link state of the vlan device is up if there is at least one bridge
port that is a vlan member that is admin & oper up, otherwise its oper
state is IF_OPER_LOWERLAYERDOWN.

Signed-off-by: Mike Manning <mmanning@vyatta.att-mail.com>
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/bridge/br.c
net/bridge/br_private.h
net/bridge/br_vlan.c

index a5174e5001d86a5b16faadd519167820284961a9..e69fc87a13e053f2d8141a2aab430be5c62785e7 100644 (file)
@@ -40,10 +40,13 @@ static int br_device_event(struct notifier_block *unused, unsigned long event, v
        bool changed_addr;
        int err;
 
-       /* register of bridge completed, add sysfs entries */
-       if ((dev->priv_flags & IFF_EBRIDGE) && event == NETDEV_REGISTER) {
-               br_sysfs_addbr(dev);
-               return NOTIFY_DONE;
+       if (dev->priv_flags & IFF_EBRIDGE) {
+               if (event == NETDEV_REGISTER) {
+                       /* register of bridge completed, add sysfs entries */
+                       br_sysfs_addbr(dev);
+                       return NOTIFY_DONE;
+               }
+               br_vlan_bridge_event(dev, event, ptr);
        }
 
        /* not a port of a bridge */
@@ -126,6 +129,8 @@ static int br_device_event(struct notifier_block *unused, unsigned long event, v
                break;
        }
 
+       br_vlan_port_event(p, event);
+
        /* Events that may cause spanning tree to refresh */
        if (!notified && (event == NETDEV_CHANGEADDR || event == NETDEV_UP ||
                          event == NETDEV_CHANGE || event == NETDEV_DOWN))
index 4bea2f11da9bd024de9ce1ec8686ff36eb671821..334a8c496b50ca96211157491a95ba7eafdf9742 100644 (file)
@@ -321,6 +321,7 @@ enum net_bridge_opts {
        BROPT_MTU_SET_BY_USER,
        BROPT_VLAN_STATS_PER_PORT,
        BROPT_NO_LL_LEARN,
+       BROPT_VLAN_BRIDGE_BINDING,
 };
 
 struct net_bridge {
@@ -895,6 +896,9 @@ int nbp_vlan_init(struct net_bridge_port *port, struct netlink_ext_ack *extack);
 int nbp_get_num_vlan_infos(struct net_bridge_port *p, u32 filter_mask);
 void br_vlan_get_stats(const struct net_bridge_vlan *v,
                       struct br_vlan_stats *stats);
+void br_vlan_port_event(struct net_bridge_port *p, unsigned long event);
+void br_vlan_bridge_event(struct net_device *dev, unsigned long event,
+                         void *ptr);
 
 static inline struct net_bridge_vlan_group *br_vlan_group(
                                        const struct net_bridge *br)
@@ -1078,6 +1082,16 @@ static inline void br_vlan_get_stats(const struct net_bridge_vlan *v,
                                     struct br_vlan_stats *stats)
 {
 }
+
+static inline void br_vlan_port_event(struct net_bridge_port *p,
+                                     unsigned long event)
+{
+}
+
+static inline void br_vlan_bridge_event(struct net_device *dev,
+                                       unsigned long event, void *ptr)
+{
+}
 #endif
 
 struct nf_br_ops {
index 0a02822b56675884998fb0ca328be0666cc73139..b903689a8fc5e7cc9032ab598231139628ef084e 100644 (file)
@@ -1264,3 +1264,154 @@ int br_vlan_get_info(const struct net_device *dev, u16 vid,
        return 0;
 }
 EXPORT_SYMBOL_GPL(br_vlan_get_info);
+
+static int br_vlan_is_bind_vlan_dev(const struct net_device *dev)
+{
+       return is_vlan_dev(dev) &&
+               !!(vlan_dev_priv(dev)->flags & VLAN_FLAG_BRIDGE_BINDING);
+}
+
+static int br_vlan_is_bind_vlan_dev_fn(struct net_device *dev,
+                                      __always_unused void *data)
+{
+       return br_vlan_is_bind_vlan_dev(dev);
+}
+
+static bool br_vlan_has_upper_bind_vlan_dev(struct net_device *dev)
+{
+       int found;
+
+       rcu_read_lock();
+       found = netdev_walk_all_upper_dev_rcu(dev, br_vlan_is_bind_vlan_dev_fn,
+                                             NULL);
+       rcu_read_unlock();
+
+       return !!found;
+}
+
+struct br_vlan_bind_walk_data {
+       u16 vid;
+       struct net_device *result;
+};
+
+static int br_vlan_match_bind_vlan_dev_fn(struct net_device *dev,
+                                         void *data_in)
+{
+       struct br_vlan_bind_walk_data *data = data_in;
+       int found = 0;
+
+       if (br_vlan_is_bind_vlan_dev(dev) &&
+           vlan_dev_priv(dev)->vlan_id == data->vid) {
+               data->result = dev;
+               found = 1;
+       }
+
+       return found;
+}
+
+static struct net_device *
+br_vlan_get_upper_bind_vlan_dev(struct net_device *dev, u16 vid)
+{
+       struct br_vlan_bind_walk_data data = {
+               .vid = vid,
+       };
+
+       rcu_read_lock();
+       netdev_walk_all_upper_dev_rcu(dev, br_vlan_match_bind_vlan_dev_fn,
+                                     &data);
+       rcu_read_unlock();
+
+       return data.result;
+}
+
+static bool br_vlan_is_dev_up(const struct net_device *dev)
+{
+       return  !!(dev->flags & IFF_UP) && netif_oper_up(dev);
+}
+
+static void br_vlan_set_vlan_dev_state(const struct net_bridge *br,
+                                      struct net_device *vlan_dev)
+{
+       u16 vid = vlan_dev_priv(vlan_dev)->vlan_id;
+       struct net_bridge_vlan_group *vg;
+       struct net_bridge_port *p;
+       bool has_carrier = false;
+
+       list_for_each_entry(p, &br->port_list, list) {
+               vg = nbp_vlan_group(p);
+               if (br_vlan_find(vg, vid) && br_vlan_is_dev_up(p->dev)) {
+                       has_carrier = true;
+                       break;
+               }
+       }
+
+       if (has_carrier)
+               netif_carrier_on(vlan_dev);
+       else
+               netif_carrier_off(vlan_dev);
+}
+
+static void br_vlan_set_all_vlan_dev_state(struct net_bridge_port *p)
+{
+       struct net_bridge_vlan_group *vg = nbp_vlan_group(p);
+       struct net_bridge_vlan *vlan;
+       struct net_device *vlan_dev;
+
+       list_for_each_entry(vlan, &vg->vlan_list, vlist) {
+               vlan_dev = br_vlan_get_upper_bind_vlan_dev(p->br->dev,
+                                                          vlan->vid);
+               if (vlan_dev) {
+                       if (br_vlan_is_dev_up(p->dev))
+                               netif_carrier_on(vlan_dev);
+                       else
+                               br_vlan_set_vlan_dev_state(p->br, vlan_dev);
+               }
+       }
+}
+
+static void br_vlan_upper_change(struct net_device *dev,
+                                struct net_device *upper_dev,
+                                bool linking)
+{
+       struct net_bridge *br = netdev_priv(dev);
+
+       if (!br_vlan_is_bind_vlan_dev(upper_dev))
+               return;
+
+       if (linking) {
+               br_vlan_set_vlan_dev_state(br, upper_dev);
+               br_opt_toggle(br, BROPT_VLAN_BRIDGE_BINDING, true);
+       } else {
+               br_opt_toggle(br, BROPT_VLAN_BRIDGE_BINDING,
+                             br_vlan_has_upper_bind_vlan_dev(dev));
+       }
+}
+
+/* Must be protected by RTNL. */
+void br_vlan_bridge_event(struct net_device *dev, unsigned long event,
+                         void *ptr)
+{
+       struct netdev_notifier_changeupper_info *info;
+
+       switch (event) {
+       case NETDEV_CHANGEUPPER:
+               info = ptr;
+               br_vlan_upper_change(dev, info->upper_dev, info->linking);
+               break;
+       }
+}
+
+/* Must be protected by RTNL. */
+void br_vlan_port_event(struct net_bridge_port *p, unsigned long event)
+{
+       if (!br_opt_get(p->br, BROPT_VLAN_BRIDGE_BINDING))
+               return;
+
+       switch (event) {
+       case NETDEV_CHANGE:
+       case NETDEV_DOWN:
+       case NETDEV_UP:
+               br_vlan_set_all_vlan_dev_state(p);
+               break;
+       }
+}