nfp: flower: ensure deletion of old offloaded MACs
authorJohn Hurley <john.hurley@netronome.com>
Wed, 16 Jan 2019 03:06:56 +0000 (19:06 -0800)
committerDavid S. Miller <davem@davemloft.net>
Wed, 16 Jan 2019 23:23:15 +0000 (15:23 -0800)
When a potential tunnel end point goes down then its MAC address should
not be matchable on the NFP.

Implement a delete message for offloaded MACs and call this on net device
down. While at it, remove the actions on register and unregister netdev
events. A MAC should only be offloaded if the device is up. Note that the
netdev notifier will replay any notifications for UP devices on
registration so NFP can still offload ports that exist before the driver
is loaded. Similarly, devices need to go down before they can be
unregistered so removal of offloaded MACs is only required on down events.

Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c

index 9b8c73bfcb5bfe13a924b4903b00b9daefc2b02c..46d8a222bd55b660701cdcda2c2dc8ad89760096 100644 (file)
@@ -98,20 +98,27 @@ struct nfp_ipv4_addr_entry {
        struct list_head list;
 };
 
+#define NFP_TUN_MAC_OFFLOAD_DEL_FLAG   0x2
+
 /**
  * struct nfp_tun_mac_addr_offload - configure MAC address of tunnel EP on NFP
- * @reserved:  reserved for future use
+ * @flags:     MAC address offload options
  * @count:     number of MAC addresses in the message (should be 1)
  * @index:     index of MAC address in the lookup table
  * @addr:      interface MAC address
  */
 struct nfp_tun_mac_addr_offload {
-       __be16 reserved;
+       __be16 flags;
        __be16 count;
        __be16 index;
        u8 addr[ETH_ALEN];
 };
 
+enum nfp_flower_mac_offload_cmd {
+       NFP_TUNNEL_MAC_OFFLOAD_ADD =            0,
+       NFP_TUNNEL_MAC_OFFLOAD_DEL =            1,
+};
+
 #define NFP_MAX_MAC_INDEX       0xff
 
 /**
@@ -515,12 +522,16 @@ static void nfp_tun_del_mac_idx(struct nfp_app *app, int ifindex)
        mutex_unlock(&priv->tun.mac_index_lock);
 }
 
-static int __nfp_tunnel_offload_mac(struct nfp_app *app, u8 *mac, u16 idx)
+static int
+__nfp_tunnel_offload_mac(struct nfp_app *app, u8 *mac, u16 idx, bool del)
 {
        struct nfp_tun_mac_addr_offload payload;
 
        memset(&payload, 0, sizeof(payload));
 
+       if (del)
+               payload.flags = cpu_to_be16(NFP_TUN_MAC_OFFLOAD_DEL_FLAG);
+
        /* FW supports multiple MACs per cmsg but restrict to single. */
        payload.count = cpu_to_be16(1);
        payload.index = cpu_to_be16(idx);
@@ -532,22 +543,12 @@ static int __nfp_tunnel_offload_mac(struct nfp_app *app, u8 *mac, u16 idx)
 }
 
 static int
-nfp_tunnel_offload_mac(struct nfp_app *app, struct net_device *netdev)
+nfp_tunnel_get_mac_idx_from_port(struct nfp_app *app, struct net_device *netdev,
+                                int port, u16 *nfp_mac_idx)
 {
-       u16 nfp_mac_idx;
-       int port = 0;
-
-       if (nfp_netdev_is_nfp_repr(netdev))
-               port = nfp_repr_get_port_id(netdev);
-       else if (!nfp_fl_is_netdev_to_offload(netdev))
-               return 0;
-
-       if (!is_valid_ether_addr(netdev->dev_addr))
-               return -EINVAL;
-
        if (FIELD_GET(NFP_FLOWER_CMSG_PORT_TYPE, port) ==
            NFP_FLOWER_CMSG_PORT_TYPE_PHYS_PORT) {
-               nfp_mac_idx = port << 8 | NFP_FLOWER_CMSG_PORT_TYPE_PHYS_PORT;
+               *nfp_mac_idx = port << 8 | NFP_FLOWER_CMSG_PORT_TYPE_PHYS_PORT;
        } else if (!port) {
                /* Must assign our own unique 8-bit index. */
                int idx = nfp_tun_get_mac_idx(app, netdev->ifindex);
@@ -555,12 +556,58 @@ nfp_tunnel_offload_mac(struct nfp_app *app, struct net_device *netdev)
                if (idx < 0)
                        return idx;
 
-               nfp_mac_idx = idx << 8 | NFP_FLOWER_CMSG_PORT_TYPE_OTHER_PORT;
+               *nfp_mac_idx = idx << 8 | NFP_FLOWER_CMSG_PORT_TYPE_OTHER_PORT;
+       } else {
+               return -EOPNOTSUPP;
+       }
+
+       return 0;
+}
+
+static int
+nfp_tunnel_offload_mac(struct nfp_app *app, struct net_device *netdev,
+                      enum nfp_flower_mac_offload_cmd cmd)
+{
+       bool non_repr = false;
+       int err, port = 0;
+       u16 nfp_mac_idx;
+
+       if (nfp_netdev_is_nfp_repr(netdev)) {
+               struct nfp_repr *repr;
+
+               repr = netdev_priv(netdev);
+               if (repr->app != app)
+                       return 0;
+
+               port = nfp_repr_get_port_id(netdev);
+       } else if (nfp_fl_is_netdev_to_offload(netdev)) {
+               non_repr = true;
        } else {
                return 0;
        }
 
-       return __nfp_tunnel_offload_mac(app, netdev->dev_addr, nfp_mac_idx);
+       if (!is_valid_ether_addr(netdev->dev_addr))
+               return -EINVAL;
+
+       switch (cmd) {
+       case NFP_TUNNEL_MAC_OFFLOAD_ADD:
+               err = nfp_tunnel_get_mac_idx_from_port(app, netdev, port,
+                                                      &nfp_mac_idx);
+               if (err)
+                       return err;
+
+               return __nfp_tunnel_offload_mac(app, netdev->dev_addr,
+                                               nfp_mac_idx, false);
+       case NFP_TUNNEL_MAC_OFFLOAD_DEL:
+               if (non_repr)
+                       nfp_tun_del_mac_idx(app, netdev->ifindex);
+
+               return __nfp_tunnel_offload_mac(app, netdev->dev_addr, 0, true);
+       default:
+               return -EINVAL;
+       }
+
+       return 0;
 }
 
 int nfp_tunnel_mac_event_handler(struct nfp_app *app,
@@ -569,13 +616,15 @@ int nfp_tunnel_mac_event_handler(struct nfp_app *app,
 {
        int err;
 
-       if (event == NETDEV_DOWN || event == NETDEV_UNREGISTER) {
-               /* If non-nfp netdev then free its offload index. */
-               if (nfp_fl_is_netdev_to_offload(netdev))
-                       nfp_tun_del_mac_idx(app, netdev->ifindex);
-       } else if (event == NETDEV_UP || event == NETDEV_CHANGEADDR ||
-                  event == NETDEV_REGISTER) {
-               err = nfp_tunnel_offload_mac(app, netdev);
+       if (event == NETDEV_DOWN) {
+               err = nfp_tunnel_offload_mac(app, netdev,
+                                            NFP_TUNNEL_MAC_OFFLOAD_DEL);
+               if (err)
+                       nfp_flower_cmsg_warn(app, "Failed to delete offload MAC on %s.\n",
+                                            netdev_name(netdev));
+       } else if (event == NETDEV_UP || event == NETDEV_CHANGEADDR) {
+               err = nfp_tunnel_offload_mac(app, netdev,
+                                            NFP_TUNNEL_MAC_OFFLOAD_ADD);
                if (err)
                        nfp_flower_cmsg_warn(app, "Failed to offload MAC on %s.\n",
                                             netdev_name(netdev));