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

MPLS over UDP is supported by adding a rule on a representor net device
which does tunnel_key set, push mpls and forward to a baredup device. At
the hardware level we use a packet_reformat_context object to do the
encapsulation of the packet.

The resulting packet looks as follows (left side transmitted first):
outer L2 | outer IP | UDP | MPLS | inner L3 and data |

Example usage:
  tc filter add dev $rep0 protocol ip prio 1 root flower skip_sw  \
     action tunnel_key set src_ip 8.8.8.21 dst_ip 8.8.8.24 id 555 \
     dst_port 6635 tos 4 ttl 6 csum action mpls push protocol 0x8847 \
     label 555 tc 3 action mirred egress redirect dev bareudp0

This is how the filter is shown with tc filter show:
tc filter show dev enp59s0f0_0 ingress
filter protocol ip pref 1 flower chain 0
filter protocol ip pref 1 flower chain 0 handle 0x1
  eth_type ipv4
  skip_sw
  in_hw in_hw_count 1
        action order 1: tunnel_key  set
        src_ip 8.8.8.21
        dst_ip 8.8.8.24
        key_id 555
        dst_port 6635
        csum
        tos 0x4
        ttl 6 pipe
         index 1 ref 1 bind 1

        action order 2: mpls  push protocol mpls_uc label 555 tc 3 ttl 255 pipe
         index 1 ref 1 bind 1

        action order 3: mirred (Egress Redirect to device bareudp0) stolen
        index 1 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>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>

drivers/net/ethernet/mellanox/mlx5/core/Makefile

@@ -38,7 +38,7 @@ mlx5_core-$(CONFIG_MLX5_ESWITCH)     += en_rep.o lib/geneve.o lib/port_tun.o lag
 mlx5_core-$(CONFIG_MLX5_CLS_ACT)     += en_tc.o en/rep/tc.o en/rep/neigh.o \
 					en/mapping.o esw/chains.o en/tc_tun.o \
 					en/tc_tun_vxlan.o en/tc_tun_gre.o en/tc_tun_geneve.o \
-					diag/en_tc_tracepoint.o
+					en/tc_tun_mplsoudp.o diag/en_tc_tracepoint.o
 mlx5_core-$(CONFIG_MLX5_TC_CT)	     += en/tc_ct.o
 
 #

drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.c

@@ -4,6 +4,7 @@
 #include <net/vxlan.h>
 #include <net/gre.h>
 #include <net/geneve.h>
+#include <net/bareudp.h>
 #include "en/tc_tun.h"
 #include "en_tc.h"
 #include "rep/tc.h"
@@ -18,6 +19,8 @@ struct mlx5e_tc_tunnel *mlx5e_get_tc_tun(struct net_device *tunnel_dev)
 	else if (netif_is_gretap(tunnel_dev) ||
 		 netif_is_ip6gretap(tunnel_dev))
 		return &gre_tunnel;
+	else if (netif_is_bareudp(tunnel_dev))
+		return &mplsoudp_tunnel;
 	else
 		return NULL;
 }

drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.h

@@ -16,6 +16,7 @@ enum {
 	MLX5E_TC_TUNNEL_TYPE_VXLAN,
 	MLX5E_TC_TUNNEL_TYPE_GENEVE,
 	MLX5E_TC_TUNNEL_TYPE_GRETAP,
+	MLX5E_TC_TUNNEL_TYPE_MPLSOUDP,
 };
 
 struct mlx5e_tc_tunnel {
@@ -46,6 +47,7 @@ struct mlx5e_tc_tunnel {
 extern struct mlx5e_tc_tunnel vxlan_tunnel;
 extern struct mlx5e_tc_tunnel geneve_tunnel;
 extern struct mlx5e_tc_tunnel gre_tunnel;
+extern struct mlx5e_tc_tunnel mplsoudp_tunnel;
 
 struct mlx5e_tc_tunnel *mlx5e_get_tc_tun(struct net_device *tunnel_dev);
 

drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun_mplsoudp.c

+// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
+/* Copyright (c) 2018 Mellanox Technologies. */
+
+#include <net/bareudp.h>
+#include <net/mpls.h>
+#include "en/tc_tun.h"
+
+static bool can_offload(struct mlx5e_priv *priv)
+{
+	return MLX5_CAP_ESW_FLOWTABLE_FDB(priv->mdev, reformat_l3_tunnel_to_l2);
+}
+
+static int calc_hlen(struct mlx5e_encap_entry *e)
+{
+	return sizeof(struct udphdr) + MPLS_HLEN;
+}
+
+static int init_encap_attr(struct net_device *tunnel_dev,
+			   struct mlx5e_priv *priv,
+			   struct mlx5e_encap_entry *e,
+			   struct netlink_ext_ack *extack)
+{
+	e->tunnel = &mplsoudp_tunnel;
+	e->reformat_type = MLX5_REFORMAT_TYPE_L2_TO_L3_TUNNEL;
+	return 0;
+}
+
+static inline __be32 mpls_label_id_field(__be32 label, u8 tos, u8 ttl)
+{
+	u32 res;
+
+	/* mpls label is 32 bits long and construction as follows:
+	 * 20 bits label
+	 * 3 bits tos
+	 * 1 bit bottom of stack. Since we support only one label, this bit is
+	 *       always set.
+	 * 8 bits TTL
+	 */
+	res = be32_to_cpu(label) << 12 | 1 << 8 | (tos & 7) <<  9 | ttl;
+	return cpu_to_be32(res);
+}
+
+static int generate_ip_tun_hdr(char buf[],
+			       __u8 *ip_proto,
+			       struct mlx5e_encap_entry *r)
+{
+	const struct ip_tunnel_key *tun_key = &r->tun_info->key;
+	__be32 tun_id = tunnel_id_to_key32(tun_key->tun_id);
+	struct udphdr *udp = (struct udphdr *)(buf);
+	struct mpls_shim_hdr *mpls;
+
+	mpls = (struct mpls_shim_hdr *)(udp + 1);
+	*ip_proto = IPPROTO_UDP;
+
+	udp->dest = tun_key->tp_dst;
+	mpls->label_stack_entry = mpls_label_id_field(tun_id, tun_key->tos, tun_key->ttl);
+
+	return 0;
+}
+
+static int parse_udp_ports(struct mlx5e_priv *priv,
+			   struct mlx5_flow_spec *spec,
+			   struct flow_cls_offload *f,
+			   void *headers_c,
+			   void *headers_v)
+{
+	return mlx5e_tc_tun_parse_udp_ports(priv, spec, f, headers_c, headers_v);
+}
+
+static int parse_tunnel(struct mlx5e_priv *priv,
+			struct mlx5_flow_spec *spec,
+			struct flow_cls_offload *f,
+			void *headers_c,
+			void *headers_v)
+{
+	return 0;
+}
+
+struct mlx5e_tc_tunnel mplsoudp_tunnel = {
+	.tunnel_type          = MLX5E_TC_TUNNEL_TYPE_MPLSOUDP,
+	.match_level          = MLX5_MATCH_L4,
+	.can_offload          = can_offload,
+	.calc_hlen            = calc_hlen,
+	.init_encap_attr      = init_encap_attr,
+	.generate_ip_tun_hdr  = generate_ip_tun_hdr,
+	.parse_udp_ports      = parse_udp_ports,
+	.parse_tunnel         = parse_tunnel,
+};

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

@@ -48,6 +48,7 @@
 #include <net/tc_act/tc_csum.h>
 #include <net/arp.h>
 #include <net/ipv6_stubs.h>
+#include <net/bareudp.h>
 #include "en.h"
 #include "en_rep.h"
 #include "en/rep/tc.h"
@@ -3685,6 +3686,7 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv,
 	bool encap = false, decap = false;
 	u32 action = attr->action;
 	int err, i, if_count = 0;
+	bool mpls_push = false;
 
 	if (!flow_action_has_entries(flow_action))
 		return -EINVAL;
@@ -3699,6 +3701,16 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv,
 			action |= MLX5_FLOW_CONTEXT_ACTION_DROP |
 				  MLX5_FLOW_CONTEXT_ACTION_COUNT;
 			break;
+		case FLOW_ACTION_MPLS_PUSH:
+			if (!MLX5_CAP_ESW_FLOWTABLE_FDB(priv->mdev,
+							reformat_l2_to_l3_tunnel) ||
+			    act->mpls_push.proto != htons(ETH_P_MPLS_UC)) {
+				NL_SET_ERR_MSG_MOD(extack,
+						   "mpls push is supported only for mpls_uc protocol");
+				return -EOPNOTSUPP;
+			}
+			mpls_push = true;
+			break;
 		case FLOW_ACTION_MANGLE:
 		case FLOW_ACTION_ADD:
 			err = parse_tc_pedit_action(priv, act, MLX5_FLOW_NAMESPACE_FDB,
@@ -3729,6 +3741,12 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv,
 				return -EINVAL;
 			}
 
+			if (mpls_push && !netif_is_bareudp(out_dev)) {
+				NL_SET_ERR_MSG_MOD(extack,
+						   "mpls is supported only through a bareudp device");
+				return -EOPNOTSUPP;
+			}
+
 			if (ft_flow && out_dev == priv->netdev) {
 				/* Ignore forward to self rules generated
 				 * by adding both mlx5 devs to the flow table

drivers/net/ethernet/mellanox/mlx5/core/lib/port_tun.c

@@ -144,11 +144,11 @@ static int mlx5_set_entropy(struct mlx5_tun_entropy *tun_entropy,
 int mlx5_tun_entropy_refcount_inc(struct mlx5_tun_entropy *tun_entropy,
 				  int reformat_type)
 {
-	/* the default is error for unknown (non VXLAN/GRE tunnel types) */
 	int err = -EOPNOTSUPP;
 
 	mutex_lock(&tun_entropy->lock);
-	if (reformat_type == MLX5_REFORMAT_TYPE_L2_TO_VXLAN &&
+	if ((reformat_type == MLX5_REFORMAT_TYPE_L2_TO_VXLAN ||
+	     reformat_type == MLX5_REFORMAT_TYPE_L2_TO_L3_TUNNEL) &&
 	    tun_entropy->enabled) {
 		/* in case entropy calculation is enabled for all tunneling
 		 * types, it is ok for VXLAN, so approve.