[PATCH net-next 0/3] flow_offload: add tc vlan push_eth and pop_eth actions

[PATCH net-next 0/3] flow_offload: add tc vlan push_eth and pop_eth actions

[Roi Dayan]

Offloading vlan push_eth and pop_eth actions is needed in order to
correctly offload MPLSoUDP encap and decap flows, this series extends
the flow offload API to support these actions and updates mlx5 to
parse them.

Maor Dickman (3):
  net/sched: add vlan push_eth and pop_eth action to the hardware IR
  net/mlx5e: MPLSoUDP decap, use vlan push_eth instead of pedit
  net/mlx5e: MPLSoUDP encap, support action vlan pop_eth explicitly

 .../mellanox/mlx5/core/en/tc/act/act.c        |  7 +++
 .../mellanox/mlx5/core/en/tc/act/act.h        |  2 +
 .../mellanox/mlx5/core/en/tc/act/mirred.c     | 10 ++++
 .../mellanox/mlx5/core/en/tc/act/mpls.c       |  7 ++-
 .../mellanox/mlx5/core/en/tc/act/pedit.c      | 59 +++----------------
 .../mellanox/mlx5/core/en/tc/act/pedit.h      |  3 +-
 .../mellanox/mlx5/core/en/tc/act/vlan.c       | 19 ++++--
 .../mlx5/core/en/tc/act/vlan_mangle.c         |  4 +-
 .../ethernet/mellanox/mlx5/core/en/tc_priv.h  |  1 -
 .../mellanox/mlx5/core/en/tc_tun_encap.c      | 10 ++--
 .../net/ethernet/mellanox/mlx5/core/eswitch.h |  1 +
 include/net/flow_offload.h                    |  4 ++
 include/net/tc_act/tc_vlan.h                  | 14 +++++
 net/sched/act_vlan.c                          | 14 +++++
 14 files changed, 85 insertions(+), 70 deletions(-)

-- 
2.34.1

[PATCH net-next 1/3] net/sched: add vlan push_eth and pop_eth action to the hardware IR

[Roi Dayan]

From: Maor Dickman <maord@nvidia.com>

Add vlan push_eth and pop_eth action to the hardware intermediate
representation model which would subsequently allow it to be used
by drivers for offload.

Signed-off-by: Maor Dickman <maord@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
---
 include/net/flow_offload.h   |  4 ++++
 include/net/tc_act/tc_vlan.h | 14 ++++++++++++++
 net/sched/act_vlan.c         | 14 ++++++++++++++
 3 files changed, 32 insertions(+)

diff --git a/include/net/flow_offload.h b/include/net/flow_offload.h
index 92267d23083e..2bfa666842c5 100644
--- a/include/net/flow_offload.h
+++ b/include/net/flow_offload.h
@@ -150,6 +150,8 @@ enum flow_action_id {
 	FLOW_ACTION_PPPOE_PUSH,
 	FLOW_ACTION_JUMP,
 	FLOW_ACTION_PIPE,
+	FLOW_ACTION_VLAN_PUSH_ETH,
+	FLOW_ACTION_VLAN_POP_ETH,
 	NUM_FLOW_ACTIONS,
 };
 
@@ -211,6 +213,8 @@ struct flow_action_entry {
 			__be16		proto;
 			u8		prio;
 		} vlan;
+		unsigned char vlan_push_eth_dst[ETH_ALEN];
+		unsigned char vlan_push_eth_src[ETH_ALEN];
 		struct {				/* FLOW_ACTION_MANGLE */
 							/* FLOW_ACTION_ADD */
 			enum flow_action_mangle_base htype;
diff --git a/include/net/tc_act/tc_vlan.h b/include/net/tc_act/tc_vlan.h
index f94b8bc26f9e..8a3422c70f9f 100644
--- a/include/net/tc_act/tc_vlan.h
+++ b/include/net/tc_act/tc_vlan.h
@@ -78,4 +78,18 @@ static inline u8 tcf_vlan_push_prio(const struct tc_action *a)
 
 	return tcfv_push_prio;
 }
+
+static inline void tcf_vlan_push_dst(unsigned char *dest, const struct tc_action *a)
+{
+	rcu_read_lock();
+	memcpy(dest, rcu_dereference(to_vlan(a)->vlan_p)->tcfv_push_dst, ETH_ALEN);
+	rcu_read_unlock();
+}
+
+static inline void tcf_vlan_push_src(unsigned char *dest, const struct tc_action *a)
+{
+	rcu_read_lock();
+	memcpy(dest, rcu_dereference(to_vlan(a)->vlan_p)->tcfv_push_src, ETH_ALEN);
+	rcu_read_unlock();
+}
 #endif /* __NET_TC_VLAN_H */
diff --git a/net/sched/act_vlan.c b/net/sched/act_vlan.c
index 756e2dcde1cd..d27604204f17 100644
--- a/net/sched/act_vlan.c
+++ b/net/sched/act_vlan.c
@@ -390,6 +390,14 @@ static int tcf_vlan_offload_act_setup(struct tc_action *act, void *entry_data,
 			entry->vlan.proto = tcf_vlan_push_proto(act);
 			entry->vlan.prio = tcf_vlan_push_prio(act);
 			break;
+		case TCA_VLAN_ACT_POP_ETH:
+			entry->id = FLOW_ACTION_VLAN_POP_ETH;
+			break;
+		case TCA_VLAN_ACT_PUSH_ETH:
+			entry->id = FLOW_ACTION_VLAN_PUSH_ETH;
+			tcf_vlan_push_dst(entry->vlan_push_eth_dst, act);
+			tcf_vlan_push_src(entry->vlan_push_eth_src, act);
+			break;
 		default:
 			return -EOPNOTSUPP;
 		}
@@ -407,6 +415,12 @@ static int tcf_vlan_offload_act_setup(struct tc_action *act, void *entry_data,
 		case TCA_VLAN_ACT_MODIFY:
 			fl_action->id = FLOW_ACTION_VLAN_MANGLE;
 			break;
+		case TCA_VLAN_ACT_POP_ETH:
+			fl_action->id = FLOW_ACTION_VLAN_POP_ETH;
+			break;
+		case TCA_VLAN_ACT_PUSH_ETH:
+			fl_action->id = FLOW_ACTION_VLAN_PUSH_ETH;
+			break;
 		default:
 			return -EOPNOTSUPP;
 		}
-- 
2.34.1

[PATCH net-next 2/3] net/mlx5e: MPLSoUDP decap, use vlan push_eth instead of pedit

[Roi Dayan]

From: Maor Dickman <maord@nvidia.com>

Currently action pedit of source and destination MACs is used
to fill the MACs in L2 push step in MPLSoUDP decap offload,
this isn't aligned to tc SW which use vlan eth_push action
to do this.

To fix that, offload support for vlan veth_push action is
added together with mpls pop action, and deprecate the use
of pedit of MACs.

Flow example:
filter protocol mpls_uc pref 1 flower chain 0
filter protocol mpls_uc pref 1 flower chain 0 handle 0x1
  eth_type 8847
  mpls_label 555
  enc_dst_port 6635
  in_hw in_hw_count 1
        action order 1: tunnel_key  unset pipe
         index 2 ref 1 bind 1
        used_hw_stats delayed

        action order 2: mpls  pop protocol ip pipe
         index 2 ref 1 bind 1
        used_hw_stats delayed

        action order 3: vlan  push_eth dst_mac de:a2:ec:d6:69:c8 src_mac de:a2:ec:d6:69:c8 pipe
         index 2 ref 1 bind 1
        used_hw_stats delayed

        action order 4: mirred (Egress Redirect to device enp8s0f0_0) stolen
        index 2 ref 1 bind 1
        used_hw_stats delayed

Signed-off-by: Maor Dickman <maord@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
---
 .../mellanox/mlx5/core/en/tc/act/act.c        |  6 ++
 .../mellanox/mlx5/core/en/tc/act/act.h        |  1 +
 .../mellanox/mlx5/core/en/tc/act/mirred.c     |  5 ++
 .../mellanox/mlx5/core/en/tc/act/mpls.c       |  7 ++-
 .../mellanox/mlx5/core/en/tc/act/pedit.c      | 59 +++----------------
 .../mellanox/mlx5/core/en/tc/act/pedit.h      |  3 +-
 .../mellanox/mlx5/core/en/tc/act/vlan.c       | 16 +++--
 .../mlx5/core/en/tc/act/vlan_mangle.c         |  4 +-
 .../ethernet/mellanox/mlx5/core/en/tc_priv.h  |  1 -
 .../mellanox/mlx5/core/en/tc_tun_encap.c      | 10 ++--
 .../net/ethernet/mellanox/mlx5/core/eswitch.h |  1 +
 11 files changed, 43 insertions(+), 70 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.c
index cb8f7593a00c..24403593b952 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.c
@@ -35,6 +35,12 @@ static struct mlx5e_tc_act *tc_acts_fdb[NUM_FLOW_ACTIONS] = {
 	NULL, /* FLOW_ACTION_CT_METADATA, */
 	&mlx5e_tc_act_mpls_push,
 	&mlx5e_tc_act_mpls_pop,
+	NULL, /* FLOW_ACTION_MPLS_MANGLE, */
+	NULL, /* FLOW_ACTION_GATE, */
+	NULL, /* FLOW_ACTION_PPPOE_PUSH, */
+	NULL, /* FLOW_ACTION_JUMP, */
+	NULL, /* FLOW_ACTION_PIPE, */
+	&mlx5e_tc_act_vlan,
 };
 
 /* Must be aligned with enum flow_action_id. */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.h b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.h
index 94a7cf38d6b1..2616aee6ebf0 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.h
@@ -22,6 +22,7 @@ struct mlx5e_tc_act_parse_state {
 	bool encap;
 	bool decap;
 	bool mpls_push;
+	bool eth_push;
 	bool ptype_host;
 	const struct ip_tunnel_info *tun_info;
 	struct mlx5e_mpls_info mpls_info;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/mirred.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/mirred.c
index 05a42fb4ba97..14cfa39d30f7 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/mirred.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/mirred.c
@@ -125,6 +125,11 @@ tc_act_can_offload_mirred(struct mlx5e_tc_act_parse_state *parse_state,
 		return false;
 	}
 
+	if (flow_flag_test(parse_state->flow, L3_TO_L2_DECAP) && !parse_state->eth_push) {
+		NL_SET_ERR_MSG_MOD(extack, "mpls pop is only supported with vlan eth push");
+		return false;
+	}
+
 	if (mlx5e_is_ft_flow(flow) && out_dev == priv->netdev) {
 		/* Ignore forward to self rules generated
 		 * by adding both mlx5 devs to the flow table
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/mpls.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/mpls.c
index 96a80e03d129..f106190bf37c 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/mpls.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/mpls.c
@@ -57,12 +57,13 @@ tc_act_can_offload_mpls_pop(struct mlx5e_tc_act_parse_state *parse_state,
 	filter_dev = attr->parse_attr->filter_dev;
 
 	/* we only support mpls pop if it is the first action
+	 * or it is second action after tunnel key unset
 	 * and the filter net device is bareudp. Subsequent
 	 * actions can be pedit and the last can be mirred
 	 * egress redirect.
 	 */
-	if (act_index) {
-		NL_SET_ERR_MSG_MOD(extack, "mpls pop supported only as first action");
+	if ((act_index == 1 && !parse_state->decap) || act_index > 1) {
+		NL_SET_ERR_MSG_MOD(extack, "mpls pop supported only as first action or with decap");
 		return false;
 	}
 
@@ -80,7 +81,7 @@ tc_act_parse_mpls_pop(struct mlx5e_tc_act_parse_state *parse_state,
 		      struct mlx5e_priv *priv,
 		      struct mlx5_flow_attr *attr)
 {
-	attr->parse_attr->eth.h_proto = act->mpls_pop.proto;
+	attr->esw_attr->eth.h_proto = act->mpls_pop.proto;
 	attr->action |= MLX5_FLOW_CONTEXT_ACTION_PACKET_REFORMAT;
 	flow_flag_set(parse_state->flow, L3_TO_L2_DECAP);
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/pedit.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/pedit.c
index 39f8f71bed9e..47597c524e59 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/pedit.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/pedit.c
@@ -42,13 +42,12 @@ set_pedit_val(u8 hdr_type, u32 mask, u32 val, u32 offset,
 	return -EOPNOTSUPP;
 }
 
-static int
-parse_pedit_to_modify_hdr(struct mlx5e_priv *priv,
-			  const struct flow_action_entry *act, int namespace,
-			  struct mlx5e_tc_flow_parse_attr *parse_attr,
-			  struct netlink_ext_ack *extack)
+int
+mlx5e_tc_act_pedit_parse_action(struct mlx5e_priv *priv,
+				const struct flow_action_entry *act, int namespace,
+				struct pedit_headers_action *hdrs,
+				struct netlink_ext_ack *extack)
 {
-	struct pedit_headers_action *hdrs = parse_attr->hdrs;
 	u8 cmd = (act->id == FLOW_ACTION_MANGLE) ? 0 : 1;
 	u8 htype = act->mangle.htype;
 	int err = -EOPNOTSUPP;
@@ -79,46 +78,6 @@ parse_pedit_to_modify_hdr(struct mlx5e_priv *priv,
 	return err;
 }
 
-static int
-parse_pedit_to_reformat(const struct flow_action_entry *act,
-			struct mlx5e_tc_flow_parse_attr *parse_attr,
-			struct netlink_ext_ack *extack)
-{
-	u32 mask, val, offset;
-	u32 *p;
-
-	if (act->id != FLOW_ACTION_MANGLE) {
-		NL_SET_ERR_MSG_MOD(extack, "Unsupported action id");
-		return -EOPNOTSUPP;
-	}
-
-	if (act->mangle.htype != FLOW_ACT_MANGLE_HDR_TYPE_ETH) {
-		NL_SET_ERR_MSG_MOD(extack, "Only Ethernet modification is supported");
-		return -EOPNOTSUPP;
-	}
-
-	mask = ~act->mangle.mask;
-	val = act->mangle.val;
-	offset = act->mangle.offset;
-	p = (u32 *)&parse_attr->eth;
-	*(p + (offset >> 2)) |= (val & mask);
-
-	return 0;
-}
-
-int
-mlx5e_tc_act_pedit_parse_action(struct mlx5e_priv *priv,
-				const struct flow_action_entry *act, int namespace,
-				struct mlx5e_tc_flow_parse_attr *parse_attr,
-				struct mlx5e_tc_flow *flow,
-				struct netlink_ext_ack *extack)
-{
-	if (flow && flow_flag_test(flow, L3_TO_L2_DECAP))
-		return parse_pedit_to_reformat(act, parse_attr, extack);
-
-	return parse_pedit_to_modify_hdr(priv, act, namespace, parse_attr, extack);
-}
-
 static bool
 tc_act_can_offload_pedit(struct mlx5e_tc_act_parse_state *parse_state,
 			 const struct flow_action_entry *act,
@@ -141,20 +100,16 @@ tc_act_parse_pedit(struct mlx5e_tc_act_parse_state *parse_state,
 
 	ns_type = mlx5e_get_flow_namespace(flow);
 
-	err = mlx5e_tc_act_pedit_parse_action(flow->priv, act, ns_type, attr->parse_attr,
-					      flow, parse_state->extack);
+	err = mlx5e_tc_act_pedit_parse_action(flow->priv, act, ns_type, attr->parse_attr->hdrs,
+					      parse_state->extack);
 	if (err)
 		return err;
 
-	if (flow_flag_test(flow, L3_TO_L2_DECAP))
-		goto out;
-
 	attr->action |= MLX5_FLOW_CONTEXT_ACTION_MOD_HDR;
 
 	if (ns_type == MLX5_FLOW_NAMESPACE_FDB)
 		esw_attr->split_count = esw_attr->out_count;
 
-out:
 	return 0;
 }
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/pedit.h b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/pedit.h
index 258f030a2dc6..434c8bd710a2 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/pedit.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/pedit.h
@@ -24,8 +24,7 @@ struct pedit_headers_action {
 int
 mlx5e_tc_act_pedit_parse_action(struct mlx5e_priv *priv,
 				const struct flow_action_entry *act, int namespace,
-				struct mlx5e_tc_flow_parse_attr *parse_attr,
-				struct mlx5e_tc_flow *flow,
+				struct pedit_headers_action *hdrs,
 				struct netlink_ext_ack *extack);
 
 #endif /* __MLX5_EN_TC_ACT_PEDIT_H__ */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/vlan.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/vlan.c
index 6378b7558ba2..b78e99ab60c7 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/vlan.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/vlan.c
@@ -34,7 +34,8 @@ parse_tc_vlan_action(struct mlx5e_priv *priv,
 		     const struct flow_action_entry *act,
 		     struct mlx5_esw_flow_attr *attr,
 		     u32 *action,
-		     struct netlink_ext_ack *extack)
+		     struct netlink_ext_ack *extack,
+		     struct mlx5e_tc_act_parse_state *parse_state)
 {
 	u8 vlan_idx = attr->total_vlan;
 
@@ -84,6 +85,13 @@ parse_tc_vlan_action(struct mlx5e_priv *priv,
 			*action |= MLX5_FLOW_CONTEXT_ACTION_VLAN_PUSH;
 		}
 		break;
+	case FLOW_ACTION_VLAN_PUSH_ETH:
+		if (!flow_flag_test(parse_state->flow, L3_TO_L2_DECAP))
+			return -EOPNOTSUPP;
+		parse_state->eth_push = true;
+		memcpy(attr->eth.h_dest, act->vlan_push_eth_dst, ETH_ALEN);
+		memcpy(attr->eth.h_source, act->vlan_push_eth_src, ETH_ALEN);
+		break;
 	default:
 		NL_SET_ERR_MSG_MOD(extack, "Unexpected action id for VLAN");
 		return -EINVAL;
@@ -109,7 +117,7 @@ mlx5e_tc_act_vlan_add_push_action(struct mlx5e_priv *priv,
 	};
 	int err;
 
-	err = parse_tc_vlan_action(priv, &vlan_act, attr->esw_attr, &attr->action, extack);
+	err = parse_tc_vlan_action(priv, &vlan_act, attr->esw_attr, &attr->action, extack, NULL);
 	if (err)
 		return err;
 
@@ -139,7 +147,7 @@ mlx5e_tc_act_vlan_add_pop_action(struct mlx5e_priv *priv,
 						priv->netdev->lower_level;
 	while (nest_level--) {
 		err = parse_tc_vlan_action(priv, &vlan_act, attr->esw_attr, &attr->action,
-					   extack);
+					   extack, NULL);
 		if (err)
 			return err;
 	}
@@ -174,7 +182,7 @@ tc_act_parse_vlan(struct mlx5e_tc_act_parse_state *parse_state,
 							   parse_state->extack);
 	} else {
 		err = parse_tc_vlan_action(priv, act, esw_attr, &attr->action,
-					   parse_state->extack);
+					   parse_state->extack, parse_state);
 	}
 
 	if (err)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/vlan_mangle.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/vlan_mangle.c
index 28444d4ffd73..9a8a1a6bd99e 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/vlan_mangle.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/vlan_mangle.c
@@ -43,8 +43,8 @@ mlx5e_tc_act_vlan_add_rewrite_action(struct mlx5e_priv *priv, int namespace,
 		return -EOPNOTSUPP;
 	}
 
-	err = mlx5e_tc_act_pedit_parse_action(priv, &pedit_act, namespace, parse_attr,
-					      NULL, extack);
+	err = mlx5e_tc_act_pedit_parse_action(priv, &pedit_act, namespace, parse_attr->hdrs,
+					      extack);
 	*action |= MLX5_FLOW_CONTEXT_ACTION_MOD_HDR;
 
 	return err;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc_priv.h b/drivers/net/ethernet/mellanox/mlx5/core/en/tc_priv.h
index 03c953dacb09..3b74a6fd5c43 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc_priv.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc_priv.h
@@ -41,7 +41,6 @@ struct mlx5e_tc_flow_parse_attr {
 	struct pedit_headers_action hdrs[__PEDIT_CMD_MAX];
 	struct mlx5e_tc_mod_hdr_acts mod_hdr_acts;
 	int mirred_ifindex[MLX5_MAX_FLOW_FWD_VPORTS];
-	struct ethhdr eth;
 	struct mlx5e_tc_act_parse_state parse_state;
 };
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun_encap.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun_encap.c
index 5105c8018d37..5aff97914367 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun_encap.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun_encap.c
@@ -906,20 +906,18 @@ int mlx5e_attach_decap(struct mlx5e_priv *priv,
 	struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
 	struct mlx5_esw_flow_attr *attr = flow->attr->esw_attr;
 	struct mlx5_pkt_reformat_params reformat_params;
-	struct mlx5e_tc_flow_parse_attr *parse_attr;
 	struct mlx5e_decap_entry *d;
 	struct mlx5e_decap_key key;
 	uintptr_t hash_key;
 	int err = 0;
 
-	parse_attr = flow->attr->parse_attr;
-	if (sizeof(parse_attr->eth) > MLX5_CAP_ESW(priv->mdev, max_encap_header_size)) {
+	if (sizeof(attr->eth) > MLX5_CAP_ESW(priv->mdev, max_encap_header_size)) {
 		NL_SET_ERR_MSG_MOD(extack,
 				   "encap header larger than max supported");
 		return -EOPNOTSUPP;
 	}
 
-	key.key = parse_attr->eth;
+	key.key = attr->eth;
 	hash_key = hash_decap_info(&key);
 	mutex_lock(&esw->offloads.decap_tbl_lock);
 	d = mlx5e_decap_get(priv, &key, hash_key);
@@ -949,8 +947,8 @@ int mlx5e_attach_decap(struct mlx5e_priv *priv,
 
 	memset(&reformat_params, 0, sizeof(reformat_params));
 	reformat_params.type = MLX5_REFORMAT_TYPE_L3_TUNNEL_TO_L2;
-	reformat_params.size = sizeof(parse_attr->eth);
-	reformat_params.data = &parse_attr->eth;
+	reformat_params.size = sizeof(attr->eth);
+	reformat_params.data = &attr->eth;
 	d->pkt_reformat = mlx5_packet_reformat_alloc(priv->mdev,
 						     &reformat_params,
 						     MLX5_FLOW_NAMESPACE_FDB);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
index 973281bdb4a2..bac5160837c5 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
@@ -474,6 +474,7 @@ struct mlx5_esw_flow_attr {
 		int src_port_rewrite_act_id;
 	} dests[MLX5_MAX_FLOW_FWD_VPORTS];
 	struct mlx5_rx_tun_attr *rx_tun_attr;
+	struct ethhdr eth;
 	struct mlx5_pkt_reformat *decap_pkt_reformat;
 };
 
-- 
2.34.1

[PATCH net-next 3/3] net/mlx5e: MPLSoUDP encap, support action vlan pop_eth explicitly

[Roi Dayan]

From: Maor Dickman <maord@nvidia.com>

Currently the MPLSoUDP encap offload does the L2 pop implicitly
while adding such action explicitly (vlan eth_push) will cause
the rule to not be offloaded.

Solve it by adding offload support for vlan eth_push in case of
MPLSoUDP decap case.

Flow example:
filter root protocol ip pref 1 flower chain 0
filter root protocol ip pref 1 flower chain 0 handle 0x1
  eth_type ipv4
  dst_ip 2.2.2.22
  src_ip 2.2.2.21
  in_hw in_hw_count 1
        action order 1: vlan  pop_eth pipe
         index 1 ref 1 bind 1
        used_hw_stats delayed

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

        action order 3: tunnel_key  set
        src_ip 8.8.8.21
        dst_ip 8.8.8.22
        dst_port 6635
        csum
        tos 0x4
        ttl 6 pipe
         index 1 ref 1 bind 1
        used_hw_stats delayed

        action order 4: mirred (Egress Redirect to device bareudp0) stolen
        index 1 ref 1 bind 1
        used_hw_stats delayed

Signed-off-by: Maor Dickman <maord@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.c    | 1 +
 drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.h    | 1 +
 drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/mirred.c | 5 +++++
 drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/vlan.c   | 3 +++
 4 files changed, 10 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.c
index 24403593b952..af37a8d247a1 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.c
@@ -41,6 +41,7 @@ static struct mlx5e_tc_act *tc_acts_fdb[NUM_FLOW_ACTIONS] = {
 	NULL, /* FLOW_ACTION_JUMP, */
 	NULL, /* FLOW_ACTION_PIPE, */
 	&mlx5e_tc_act_vlan,
+	&mlx5e_tc_act_vlan,
 };
 
 /* Must be aligned with enum flow_action_id. */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.h b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.h
index 2616aee6ebf0..f34714c5ddd4 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/act.h
@@ -23,6 +23,7 @@ struct mlx5e_tc_act_parse_state {
 	bool decap;
 	bool mpls_push;
 	bool eth_push;
+	bool eth_pop;
 	bool ptype_host;
 	const struct ip_tunnel_info *tun_info;
 	struct mlx5e_mpls_info mpls_info;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/mirred.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/mirred.c
index 14cfa39d30f7..2b002c6a2e73 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/mirred.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/mirred.c
@@ -125,6 +125,11 @@ tc_act_can_offload_mirred(struct mlx5e_tc_act_parse_state *parse_state,
 		return false;
 	}
 
+	if (parse_state->eth_pop && !parse_state->mpls_push) {
+		NL_SET_ERR_MSG_MOD(extack, "vlan pop eth is supported only with mpls push");
+		return false;
+	}
+
 	if (flow_flag_test(parse_state->flow, L3_TO_L2_DECAP) && !parse_state->eth_push) {
 		NL_SET_ERR_MSG_MOD(extack, "mpls pop is only supported with vlan eth push");
 		return false;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/vlan.c b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/vlan.c
index b78e99ab60c7..37f14862f99a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/vlan.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/tc/act/vlan.c
@@ -85,6 +85,9 @@ parse_tc_vlan_action(struct mlx5e_priv *priv,
 			*action |= MLX5_FLOW_CONTEXT_ACTION_VLAN_PUSH;
 		}
 		break;
+	case FLOW_ACTION_VLAN_POP_ETH:
+		parse_state->eth_pop = true;
+		break;
 	case FLOW_ACTION_VLAN_PUSH_ETH:
 		if (!flow_flag_test(parse_state->flow, L3_TO_L2_DECAP))
 			return -EOPNOTSUPP;
-- 
2.34.1

Re: [PATCH net-next 1/3] net/sched: add vlan push_eth and pop_eth action to the hardware IR

[Jakub Kicinski]

Sorry for the late review.

On Wed, 9 Mar 2022 15:02:54 +0200 Roi Dayan wrote:
> @@ -211,6 +213,8 @@ struct flow_action_entry {
>  			__be16		proto;
>  			u8		prio;
>  		} vlan;
> +		unsigned char vlan_push_eth_dst[ETH_ALEN];
> +		unsigned char vlan_push_eth_src[ETH_ALEN];

Let's wrap these two in a struct, like all other members here, 
and add the customary comment indicating which action its for.

>  		struct {				/* FLOW_ACTION_MANGLE */
>  							/* FLOW_ACTION_ADD */
>  			enum flow_action_mangle_base htype;
> diff --git a/include/net/tc_act/tc_vlan.h b/include/net/tc_act/tc_vlan.h
> index f94b8bc26f9e..8a3422c70f9f 100644
> --- a/include/net/tc_act/tc_vlan.h
> +++ b/include/net/tc_act/tc_vlan.h
> @@ -78,4 +78,18 @@ static inline u8 tcf_vlan_push_prio(const struct tc_action *a)
>  
>  	return tcfv_push_prio;
>  }
> +
> +static inline void tcf_vlan_push_dst(unsigned char *dest, const struct tc_action *a)
> +{
> +	rcu_read_lock();
> +	memcpy(dest, rcu_dereference(to_vlan(a)->vlan_p)->tcfv_push_dst, ETH_ALEN);
> +	rcu_read_unlock();
> +}
> +
> +static inline void tcf_vlan_push_src(unsigned char *dest, const struct tc_action *a)
> +{
> +	rcu_read_lock();
> +	memcpy(dest, rcu_dereference(to_vlan(a)->vlan_p)->tcfv_push_src, ETH_ALEN);
> +	rcu_read_unlock();
> +}

The use of these two helpers separately makes no sense, we can't push
half a header. It should be one helper populating both src and dst, IMO.

Re: [PATCH net-next 2/3] net/mlx5e: MPLSoUDP decap, use vlan push_eth instead of pedit

[Jakub Kicinski]

On Wed, 9 Mar 2022 15:02:55 +0200 Roi Dayan wrote:
> +	case FLOW_ACTION_VLAN_PUSH_ETH:
> +		if (!flow_flag_test(parse_state->flow, L3_TO_L2_DECAP))
> +			return -EOPNOTSUPP;
> +		parse_state->eth_push = true;
> +		memcpy(attr->eth.h_dest, act->vlan_push_eth_dst, ETH_ALEN);
> +		memcpy(attr->eth.h_source, act->vlan_push_eth_src, ETH_ALEN);
> +		break;

How does the device know the proto? I kind of expected this code will
require some form of a match on proto.

Re: [PATCH net-next 2/3] net/mlx5e: MPLSoUDP decap, use vlan push_eth instead of pedit

[Roi Dayan]

On 2022-03-15 7:02 AM, Jakub Kicinski wrote:
> On Wed, 9 Mar 2022 15:02:55 +0200 Roi Dayan wrote:
>> +	case FLOW_ACTION_VLAN_PUSH_ETH:
>> +		if (!flow_flag_test(parse_state->flow, L3_TO_L2_DECAP))
>> +			return -EOPNOTSUPP;
>> +		parse_state->eth_push = true;
>> +		memcpy(attr->eth.h_dest, act->vlan_push_eth_dst, ETH_ALEN);
>> +		memcpy(attr->eth.h_source, act->vlan_push_eth_src, ETH_ALEN);
>> +		break;
> 
> How does the device know the proto? I kind of expected this code will
> require some form of a match on proto.

Vlan push_eth doesn't handle the protocol, it only push src/dst mac.

 From tc-vlan man page:
        PUSH_ETH := push_eth dst_mac LLADDR src_mac LLADDR

In the case of MPLSoUDP the protocol is set by mpls pop action,
prior to the vlan push_eth action.

Example:
         action order 2: mpls  pop protocol ip pipe
          index 2 ref 1 bind 1
         used_hw_stats delayed

         action order 3: vlan  push_eth dst_mac de:a2:ec:d6:69:c8 
src_mac de:a2:ec:d6:69:c8 pipe
          index 2 ref 1 bind 1
         used_hw_stats delayed

Re: [PATCH net-next 1/3] net/sched: add vlan push_eth and pop_eth action to the hardware IR

[Roi Dayan]

On 2022-03-15 7:02 AM, Jakub Kicinski wrote:
> Sorry for the late review.
> 
> On Wed, 9 Mar 2022 15:02:54 +0200 Roi Dayan wrote:
>> @@ -211,6 +213,8 @@ struct flow_action_entry {
>>   			__be16		proto;
>>   			u8		prio;
>>   		} vlan;
>> +		unsigned char vlan_push_eth_dst[ETH_ALEN];
>> +		unsigned char vlan_push_eth_src[ETH_ALEN];
> 
> Let's wrap these two in a struct, like all other members here,
> and add the customary comment indicating which action its for.
> 

ok.

>>   		struct {				/* FLOW_ACTION_MANGLE */
>>   							/* FLOW_ACTION_ADD */
>>   			enum flow_action_mangle_base htype;
>> diff --git a/include/net/tc_act/tc_vlan.h b/include/net/tc_act/tc_vlan.h
>> index f94b8bc26f9e..8a3422c70f9f 100644
>> --- a/include/net/tc_act/tc_vlan.h
>> +++ b/include/net/tc_act/tc_vlan.h
>> @@ -78,4 +78,18 @@ static inline u8 tcf_vlan_push_prio(const struct tc_action *a)
>>   
>>   	return tcfv_push_prio;
>>   }
>> +
>> +static inline void tcf_vlan_push_dst(unsigned char *dest, const struct tc_action *a)
>> +{
>> +	rcu_read_lock();
>> +	memcpy(dest, rcu_dereference(to_vlan(a)->vlan_p)->tcfv_push_dst, ETH_ALEN);
>> +	rcu_read_unlock();
>> +}
>> +
>> +static inline void tcf_vlan_push_src(unsigned char *dest, const struct tc_action *a)
>> +{
>> +	rcu_read_lock();
>> +	memcpy(dest, rcu_dereference(to_vlan(a)->vlan_p)->tcfv_push_src, ETH_ALEN);
>> +	rcu_read_unlock();
>> +}
> 
> The use of these two helpers separately makes no sense, we can't push
> half a header. It should be one helper populating both src and dst, IMO.

ok.

Re: [PATCH net-next 2/3] net/mlx5e: MPLSoUDP decap, use vlan push_eth instead of pedit

[Jakub Kicinski]

On Tue, 15 Mar 2022 11:50:46 +0200 Roi Dayan wrote:
> On 2022-03-15 7:02 AM, Jakub Kicinski wrote:
> > On Wed, 9 Mar 2022 15:02:55 +0200 Roi Dayan wrote:  
> >> +	case FLOW_ACTION_VLAN_PUSH_ETH:
> >> +		if (!flow_flag_test(parse_state->flow, L3_TO_L2_DECAP))
> >> +			return -EOPNOTSUPP;
> >> +		parse_state->eth_push = true;
> >> +		memcpy(attr->eth.h_dest, act->vlan_push_eth_dst, ETH_ALEN);
> >> +		memcpy(attr->eth.h_source, act->vlan_push_eth_src, ETH_ALEN);
> >> +		break;  
> > 
> > How does the device know the proto? I kind of expected this code will
> > require some form of a match on proto.  
> 
> Vlan push_eth doesn't handle the protocol, it only push src/dst mac.
> 
>  From tc-vlan man page:
>         PUSH_ETH := push_eth dst_mac LLADDR src_mac LLADDR

It uses the protocol from skb.

> In the case of MPLSoUDP the protocol is set by mpls pop action,
> prior to the vlan push_eth action.
> 
> Example:
>          action order 2: mpls  pop protocol ip pipe
>           index 2 ref 1 bind 1
>          used_hw_stats delayed
> 
>          action order 3: vlan  push_eth dst_mac de:a2:ec:d6:69:c8 
> src_mac de:a2:ec:d6:69:c8 pipe
>           index 2 ref 1 bind 1
>          used_hw_stats delayed

Ack, I was basically asking how the existence of an earlier mpls pop
gets enforced by the driver.