Merge branch 'devlink-eswitch-inactive-mode'

Saeed Mahameed says:

====================
devlink eswitch inactive mode

Before having traffic flow through an eswitch, a user may want to have the
ability to block traffic towards the FDB until FDB is fully programmed and the
user is ready to send traffic to it. For example: when two eswitches are present
for vports in a multi-PF setup, one eswitch may take over the traffic from the
other when the user chooses. Before this take over, a user may want to first
program the inactive eswitch and then once ready redirect traffic to this new
eswitch.

This series introduces a user-configurable mode for an eswitch that allows
dynamically switching between active and inactive modes. When inactive, traffic
does not flow through the eswitch. While inactive, steering pipeline
configuration can be done (e.g. adding TC rules, discovering representors,
enabling the desired SDN modes such as bridge/OVS/DPDK/etc). Once configuration
is completed, a user can set the eswitch mode to active and have traffic flow
through. This allows admins to upgrade forwarding pipeline rules with very
minimal downtime and packet drops.

A user can start the eswitch in switchdev or switchdev_inactive mode.

Active: Traffic is enabled on this eswitch FDB.
Inactive: Traffic is ignored/dropped on this eswitch FDB.

An example use case:
$ devlink dev eswitch set pci/0000:08:00.1 mode switchdev_inactive
Setup FDB pipeline and netdev representors
...
Once ready to start receiving traffic
$ devlink dev eswitch set pci/0000:08:00.1 mode switchdev

v2: https://lore.kernel.org/all/[email protected]/
v1: https://lore.kernel.org/all/[email protected]/
====================

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Paolo Abeni <[email protected]>

Author: Paolo Abeni

Documentation/netlink/specs/devlink.yaml

@@ -99,6 +99,8 @@ definitions:
         name: legacy
       -
         name: switchdev
+      -
+        name: switchdev-inactive
   -
     type: enum
     name: eswitch-inline-mode

Documentation/networking/devlink/devlink-eswitch-attr.rst

@@ -39,6 +39,10 @@ The following is a list of E-Switch attributes.
          rules.
        * ``switchdev`` allows for more advanced offloading capabilities of
          the E-Switch to hardware.
+       * ``switchdev_inactive`` switchdev mode but starts inactive, doesn't allow traffic
+         until explicitly activated. This mode is useful for orchestrators that
+         want to prepare the device in switchdev mode but only activate it when
+         all configurations are done.
    * - ``inline-mode``
      - enum
      - Some HWs need the VF driver to put part of the packet
@@ -74,3 +78,12 @@ Example Usage
 
     # enable encap-mode with legacy mode
     $ devlink dev eswitch set pci/0000:08:00.0 mode legacy inline-mode none encap-mode basic
+
+    # start switchdev mode in inactive state
+    $ devlink dev eswitch set pci/0000:08:00.0 mode switchdev_inactive
+
+    # setup switchdev configurations, representors, FDB entries, etc..
+    ...
+
+    # activate switchdev mode to allow traffic
+    $ devlink dev eswitch set pci/0000:08:00.0 mode switchdev

drivers/net/ethernet/mellanox/mlx5/core/esw/adj_vport.c

@@ -4,13 +4,8 @@
 #include "fs_core.h"
 #include "eswitch.h"
 
-enum {
-	MLX5_ADJ_VPORT_DISCONNECT = 0x0,
-	MLX5_ADJ_VPORT_CONNECT = 0x1,
-};
-
-static int mlx5_esw_adj_vport_modify(struct mlx5_core_dev *dev,
-				     u16 vport, bool connect)
+int mlx5_esw_adj_vport_modify(struct mlx5_core_dev *dev, u16 vport,
+			      bool connect)
 {
 	u32 in[MLX5_ST_SZ_DW(modify_vport_state_in)] = {};
 
@@ -24,7 +19,7 @@ static int mlx5_esw_adj_vport_modify(struct mlx5_core_dev *dev,
 	MLX5_SET(modify_vport_state_in, in, egress_connect_valid, 1);
 	MLX5_SET(modify_vport_state_in, in, ingress_connect, connect);
 	MLX5_SET(modify_vport_state_in, in, egress_connect, connect);
-
+	MLX5_SET(modify_vport_state_in, in, admin_state, connect);
 	return mlx5_cmd_exec_in(dev, modify_vport_state, in);
 }
 
@@ -96,7 +91,6 @@ static int mlx5_esw_adj_vport_create(struct mlx5_eswitch *esw, u16 vhca_id,
 	if (err)
 		goto acl_ns_remove;
 
-	mlx5_esw_adj_vport_modify(esw->dev, vport_num, MLX5_ADJ_VPORT_CONNECT);
 	return 0;
 
 acl_ns_remove:
@@ -117,8 +111,7 @@ static void mlx5_esw_adj_vport_destroy(struct mlx5_eswitch *esw,
 
 	esw_debug(esw->dev, "Destroying adjacent vport %d for vhca_id 0x%x\n",
 		  vport_num, vport->vhca_id);
-	mlx5_esw_adj_vport_modify(esw->dev, vport_num,
-				  MLX5_ADJ_VPORT_DISCONNECT);
+
 	mlx5_esw_offloads_rep_remove(esw, vport);
 	mlx5_fs_vport_egress_acl_ns_remove(esw->dev->priv.steering,
 					   vport->index);

drivers/net/ethernet/mellanox/mlx5/core/eswitch.h

@@ -264,6 +264,9 @@ struct mlx5_eswitch_fdb {
 
 		struct offloads_fdb {
 			struct mlx5_flow_namespace *ns;
+			struct mlx5_flow_table *drop_root;
+			struct mlx5_flow_handle *drop_root_rule;
+			struct mlx5_fc *drop_root_fc;
 			struct mlx5_flow_table *tc_miss_table;
 			struct mlx5_flow_table *slow_fdb;
 			struct mlx5_flow_group *send_to_vport_grp;
@@ -392,6 +395,7 @@ struct mlx5_eswitch {
 	struct mlx5_esw_offload offloads;
 	u32 last_vport_idx;
 	int                     mode;
+	bool                    offloads_inactive;
 	u16                     manager_vport;
 	u16                     first_host_vport;
 	u8			num_peers;
@@ -634,6 +638,8 @@ const u32 *mlx5_esw_query_functions(struct mlx5_core_dev *dev);
 
 void mlx5_esw_adjacent_vhcas_setup(struct mlx5_eswitch *esw);
 void mlx5_esw_adjacent_vhcas_cleanup(struct mlx5_eswitch *esw);
+int mlx5_esw_adj_vport_modify(struct mlx5_core_dev *dev, u16 vport,
+			      bool connect);
 
 #define MLX5_DEBUG_ESWITCH_MASK BIT(3)