net/mlx5e: Add support for hw decapsulation of MPLS over UDP

MPLS over UDP is supported in hardware by using a packet reformat object
with reformat type equal L3_TUNNEL_TO_L2 which both decapsulates the
outer L3, L4 and MPLS headers, and allows for setting the L2 headers of
the resulting decapsulated packet. For the hardware to operate
correctly, the configuration of the firmware must have
FLEX_PARSER_PROFILE_ENABLE = 1.

Example tc rule:
  tc filter add dev bareudp0 protocol all prio 1 root flower enc_dst_port \
      6635 enc_src_ip 8.8.8.23 action mpls pop protocol ip pipe \
      action pedit ex munge eth dst set 00:11:22:33:44:21 pipe action \
      mirred egress redirect dev enp59s0f0_0

We use pedit to set the correct destination MAC.

For MPLS over UDP decapsulation to take place, the driver logic requires
the following:

1. flower filter added on bareudp device.
2. action mpls pop
3. zero or more pedit munge actions
4. one redirect action

Current implementation supports only IPv4 and no VLAN.

tc filter show output looks like this:
   filter protocol all pref 1 flower chain 0
   filter protocol all pref 1 flower chain 0 handle 0x1
     enc_src_ip 8.8.8.24
     enc_dst_port 6635
     in_hw in_hw_count 1
            action order 1: mpls  pop protocol ip pipe
             index 2 ref 1 bind 1

            action order 2:  pedit action pipe keys 2
             index 1 ref 1 bind 1
             key #0  at eth+0: val 00112233 mask 00000000
             key #1  at eth+4: val 44210000 mask 0000ffff

            action order 3: mirred (Egress Redirect to device enp59s0f0_0) stolen
            index 2 ref 1 bind 1
Signed-off-by: Eli Cohen <eli@mellanox.com>
Reviewed-by: Roi Dayan <roid@mellanox.com>
Reviewed-by: Eli Britstein <elibr@mellanox.com>
Reviewed-by: Vlad Buslov <vladbu@mellanox.com>
Reviewed-by: Paul Blakey <paulb@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

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

@@ -158,6 +158,22 @@ struct mlx5e_neigh_hash_entry {
 enum {
 	/* set when the encap entry is successfully offloaded into HW */
 	MLX5_ENCAP_ENTRY_VALID     = BIT(0),
+	MLX5_REFORMAT_DECAP        = BIT(1),
+};
+
+struct mlx5e_decap_key {
+	struct ethhdr key;
+};
+
+struct mlx5e_decap_entry {
+	struct mlx5e_decap_key key;
+	struct list_head flows;
+	struct hlist_node hlist;
+	refcount_t refcnt;
+	struct completion res_ready;
+	int compl_result;
+	struct mlx5_pkt_reformat *pkt_reformat;
+	struct rcu_head rcu;
 };
 
 struct mlx5e_encap_entry {

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

@@ -2262,6 +2262,8 @@ int mlx5_eswitch_init(struct mlx5_core_dev *dev)
 	hash_init(esw->offloads.encap_tbl);
 	mutex_init(&esw->offloads.mod_hdr.lock);
 	hash_init(esw->offloads.mod_hdr.hlist);
+	mutex_init(&esw->offloads.decap_tbl_lock);
+	hash_init(esw->offloads.decap_tbl);
 	atomic64_set(&esw->offloads.num_flows, 0);
 	mutex_init(&esw->state_lock);
 	mutex_init(&esw->mode_lock);
@@ -2303,6 +2305,7 @@ void mlx5_eswitch_cleanup(struct mlx5_eswitch *esw)
 	mutex_destroy(&esw->state_lock);
 	mutex_destroy(&esw->offloads.mod_hdr.lock);
 	mutex_destroy(&esw->offloads.encap_tbl_lock);
+	mutex_destroy(&esw->offloads.decap_tbl_lock);
 	kfree(esw->vports);
 	kfree(esw);
 }

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

@@ -209,6 +209,8 @@ struct mlx5_esw_offload {
 	struct mutex peer_mutex;
 	struct mutex encap_tbl_lock; /* protects encap_tbl */
 	DECLARE_HASHTABLE(encap_tbl, 8);
+	struct mutex decap_tbl_lock; /* protects decap_tbl */
+	DECLARE_HASHTABLE(decap_tbl, 8);
 	struct mod_hdr_tbl mod_hdr;
 	DECLARE_HASHTABLE(termtbl_tbl, 8);
 	struct mutex termtbl_mutex; /* protects termtbl hash */
@@ -432,6 +434,7 @@ struct mlx5_esw_flow_attr {
 	struct mlx5_flow_table *fdb;
 	struct mlx5_flow_table *dest_ft;
 	struct mlx5_ct_attr ct_attr;
+	struct mlx5_pkt_reformat *decap_pkt_reformat;
 	struct mlx5e_tc_flow_parse_attr *parse_attr;
 };
 

drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c

@@ -366,6 +366,10 @@ mlx5_eswitch_add_offloaded_rule(struct mlx5_eswitch *esw,
 			}
 		}
 	}
+
+	if (attr->decap_pkt_reformat)
+		flow_act.pkt_reformat = attr->decap_pkt_reformat;
+
 	if (flow_act.action & MLX5_FLOW_CONTEXT_ACTION_COUNT) {
 		dest[i].type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
 		dest[i].counter_id = mlx5_fc_id(attr->counter);