Merge tag 'mlx5e-pedit' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux

Or Gerlitz says:

====================
mlx5e-pedit 2017-03-28

This series adds support for offloading modifications of packet headers using
ConnectX-5 HW header re-write as an action applied during packet steering.

The offloaded SW mechanism is TC's pedit action. The offloading is
supported for E-Switch steering of VF traffic in the SRIOV
switchdev mode and for NIC (non eswitch) RX.

One use-case for this offload on virtual networks, is when the hypervisor
implements flow based router such as Open-Stack's DVR, where L2 headers
of guest packets re-written with routers' MAC addresses and the IP TTL
is decremented.

Another use case (which can be applied in parallel with routing) is
stateless NAT where guest L3/L4 headers are re-written.

The series is built as follows: the 1st six patches are preperations which
don't yet add new functionality, patches 7-8 add the FW APIs (data-structures
and commands) for header re-write, and patch nine allows offloading driver
to access pedit keys.

The 10th patch is somehow the core of the series, where we translate from
the pedit way to represent set of header modification elements to the FW
API for that same matter.

Once a set of HW modification is established, we register it with the FW
and get a modify header ID. When this ID is used with an action during
packet steering, the HW applies the header modification on the packet.

Patches 11 and 12 implement the above logic as an offload for pedit action
for the NIC and E-Switch use-cases.

I'd like to thanks Elijah Shakkour <elijahs@mellanox.com> for implementing
and helping me testing this functionality on HW simulator, before it could
be done with FW.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

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

@@ -279,6 +279,8 @@ static int mlx5_internal_err_ret_value(struct mlx5_core_dev *dev, u16 op,
 	case MLX5_CMD_OP_DESTROY_XRC_SRQ:
 	case MLX5_CMD_OP_DESTROY_DCT:
 	case MLX5_CMD_OP_DEALLOC_Q_COUNTER:
+	case MLX5_CMD_OP_DESTROY_SCHEDULING_ELEMENT:
+	case MLX5_CMD_OP_DESTROY_QOS_PARA_VPORT:
 	case MLX5_CMD_OP_DEALLOC_PD:
 	case MLX5_CMD_OP_DEALLOC_UAR:
 	case MLX5_CMD_OP_DETACH_FROM_MCG:
@@ -305,8 +307,7 @@ static int mlx5_internal_err_ret_value(struct mlx5_core_dev *dev, u16 op,
 	case MLX5_CMD_OP_SET_FLOW_TABLE_ENTRY:
 	case MLX5_CMD_OP_SET_FLOW_TABLE_ROOT:
 	case MLX5_CMD_OP_DEALLOC_ENCAP_HEADER:
-	case MLX5_CMD_OP_DESTROY_SCHEDULING_ELEMENT:
-	case MLX5_CMD_OP_DESTROY_QOS_PARA_VPORT:
+	case MLX5_CMD_OP_DEALLOC_MODIFY_HEADER_CONTEXT:
 		return MLX5_CMD_STAT_OK;
 
 	case MLX5_CMD_OP_QUERY_HCA_CAP:
@@ -363,6 +364,10 @@ static int mlx5_internal_err_ret_value(struct mlx5_core_dev *dev, u16 op,
 	case MLX5_CMD_OP_QUERY_Q_COUNTER:
 	case MLX5_CMD_OP_SET_RATE_LIMIT:
 	case MLX5_CMD_OP_QUERY_RATE_LIMIT:
+	case MLX5_CMD_OP_CREATE_SCHEDULING_ELEMENT:
+	case MLX5_CMD_OP_QUERY_SCHEDULING_ELEMENT:
+	case MLX5_CMD_OP_MODIFY_SCHEDULING_ELEMENT:
+	case MLX5_CMD_OP_CREATE_QOS_PARA_VPORT:
 	case MLX5_CMD_OP_ALLOC_PD:
 	case MLX5_CMD_OP_ALLOC_UAR:
 	case MLX5_CMD_OP_CONFIG_INT_MODERATION:
@@ -414,10 +419,7 @@ static int mlx5_internal_err_ret_value(struct mlx5_core_dev *dev, u16 op,
 	case MLX5_CMD_OP_ALLOC_FLOW_COUNTER:
 	case MLX5_CMD_OP_QUERY_FLOW_COUNTER:
 	case MLX5_CMD_OP_ALLOC_ENCAP_HEADER:
-	case MLX5_CMD_OP_CREATE_SCHEDULING_ELEMENT:
-	case MLX5_CMD_OP_QUERY_SCHEDULING_ELEMENT:
-	case MLX5_CMD_OP_MODIFY_SCHEDULING_ELEMENT:
-	case MLX5_CMD_OP_CREATE_QOS_PARA_VPORT:
+	case MLX5_CMD_OP_ALLOC_MODIFY_HEADER_CONTEXT:
 		*status = MLX5_DRIVER_STATUS_ABORTED;
 		*synd = MLX5_DRIVER_SYND;
 		return -EIO;
@@ -501,6 +503,12 @@ const char *mlx5_command_str(int command)
 	MLX5_COMMAND_STR_CASE(QUERY_Q_COUNTER);
 	MLX5_COMMAND_STR_CASE(SET_RATE_LIMIT);
 	MLX5_COMMAND_STR_CASE(QUERY_RATE_LIMIT);
+	MLX5_COMMAND_STR_CASE(CREATE_SCHEDULING_ELEMENT);
+	MLX5_COMMAND_STR_CASE(DESTROY_SCHEDULING_ELEMENT);
+	MLX5_COMMAND_STR_CASE(QUERY_SCHEDULING_ELEMENT);
+	MLX5_COMMAND_STR_CASE(MODIFY_SCHEDULING_ELEMENT);
+	MLX5_COMMAND_STR_CASE(CREATE_QOS_PARA_VPORT);
+	MLX5_COMMAND_STR_CASE(DESTROY_QOS_PARA_VPORT);
 	MLX5_COMMAND_STR_CASE(ALLOC_PD);
 	MLX5_COMMAND_STR_CASE(DEALLOC_PD);
 	MLX5_COMMAND_STR_CASE(ALLOC_UAR);
@@ -576,12 +584,8 @@ const char *mlx5_command_str(int command)
 	MLX5_COMMAND_STR_CASE(MODIFY_FLOW_TABLE);
 	MLX5_COMMAND_STR_CASE(ALLOC_ENCAP_HEADER);
 	MLX5_COMMAND_STR_CASE(DEALLOC_ENCAP_HEADER);
-	MLX5_COMMAND_STR_CASE(CREATE_SCHEDULING_ELEMENT);
-	MLX5_COMMAND_STR_CASE(DESTROY_SCHEDULING_ELEMENT);
-	MLX5_COMMAND_STR_CASE(QUERY_SCHEDULING_ELEMENT);
-	MLX5_COMMAND_STR_CASE(MODIFY_SCHEDULING_ELEMENT);
-	MLX5_COMMAND_STR_CASE(CREATE_QOS_PARA_VPORT);
-	MLX5_COMMAND_STR_CASE(DESTROY_QOS_PARA_VPORT);
+	MLX5_COMMAND_STR_CASE(ALLOC_MODIFY_HEADER_CONTEXT);
+	MLX5_COMMAND_STR_CASE(DEALLOC_MODIFY_HEADER_CONTEXT);
 	default: return "unknown command opcode";
 	}
 }

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

@@ -174,13 +174,9 @@ static int arfs_add_default_rule(struct mlx5e_priv *priv,
 				 enum arfs_type type)
 {
 	struct arfs_table *arfs_t = &priv->fs.arfs.arfs_tables[type];
-	struct mlx5_flow_act flow_act = {
-		.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST,
-		.flow_tag = MLX5_FS_DEFAULT_FLOW_TAG,
-		.encap_id = 0,
-	};
-	struct mlx5_flow_destination dest;
 	struct mlx5e_tir *tir = priv->indir_tir;
+	struct mlx5_flow_destination dest;
+	MLX5_DECLARE_FLOW_ACT(flow_act);
 	struct mlx5_flow_spec *spec;
 	int err = 0;
 
@@ -469,15 +465,11 @@ static struct arfs_table *arfs_get_table(struct mlx5e_arfs_tables *arfs,
 static struct mlx5_flow_handle *arfs_add_rule(struct mlx5e_priv *priv,
 					      struct arfs_rule *arfs_rule)
 {
-	struct mlx5_flow_act flow_act = {
-		.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST,
-		.flow_tag = MLX5_FS_DEFAULT_FLOW_TAG,
-		.encap_id = 0,
-	};
 	struct mlx5e_arfs_tables *arfs = &priv->fs.arfs;
 	struct arfs_tuple *tuple = &arfs_rule->tuple;
 	struct mlx5_flow_handle *rule = NULL;
 	struct mlx5_flow_destination dest;
+	MLX5_DECLARE_FLOW_ACT(flow_act);
 	struct arfs_table *arfs_table;
 	struct mlx5_flow_spec *spec;
 	struct mlx5_flow_table *ft;

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

@@ -159,14 +159,10 @@ static int __mlx5e_add_vlan_rule(struct mlx5e_priv *priv,
 				 enum mlx5e_vlan_rule_type rule_type,
 				 u16 vid, struct mlx5_flow_spec *spec)
 {
-	struct mlx5_flow_act flow_act = {
-		.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST,
-		.flow_tag = MLX5_FS_DEFAULT_FLOW_TAG,
-		.encap_id = 0,
-	};
 	struct mlx5_flow_table *ft = priv->fs.vlan.ft.t;
 	struct mlx5_flow_destination dest;
 	struct mlx5_flow_handle **rule_p;
+	MLX5_DECLARE_FLOW_ACT(flow_act);
 	int err = 0;
 
 	dest.type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
@@ -659,11 +655,7 @@ mlx5e_generate_ttc_rule(struct mlx5e_priv *priv,
 			u16 etype,
 			u8 proto)
 {
-	struct mlx5_flow_act flow_act = {
-		.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST,
-		.flow_tag = MLX5_FS_DEFAULT_FLOW_TAG,
-		.encap_id = 0,
-	};
+	MLX5_DECLARE_FLOW_ACT(flow_act);
 	struct mlx5_flow_handle *rule;
 	struct mlx5_flow_spec *spec;
 	int err = 0;
@@ -848,13 +840,9 @@ static void mlx5e_del_l2_flow_rule(struct mlx5e_priv *priv,
 static int mlx5e_add_l2_flow_rule(struct mlx5e_priv *priv,
 				  struct mlx5e_l2_rule *ai, int type)
 {
-	struct mlx5_flow_act flow_act = {
-		.action = MLX5_FLOW_CONTEXT_ACTION_FWD_DEST,
-		.flow_tag = MLX5_FS_DEFAULT_FLOW_TAG,
-		.encap_id = 0,
-	};
 	struct mlx5_flow_table *ft = priv->fs.l2.ft.t;
 	struct mlx5_flow_destination dest;
+	MLX5_DECLARE_FLOW_ACT(flow_act);
 	struct mlx5_flow_spec *spec;
 	int err = 0;
 	u8 *mc_dmac;

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

@@ -42,14 +42,22 @@
 #include <net/tc_act/tc_mirred.h>
 #include <net/tc_act/tc_vlan.h>
 #include <net/tc_act/tc_tunnel_key.h>
+#include <net/tc_act/tc_pedit.h>
 #include <net/vxlan.h>
 #include "en.h"
 #include "en_tc.h"
 #include "eswitch.h"
 #include "vxlan.h"
 
+struct mlx5_nic_flow_attr {
+	u32 action;
+	u32 flow_tag;
+	u32 mod_hdr_id;
+};
+
 enum {
 	MLX5E_TC_FLOW_ESWITCH	= BIT(0),
+	MLX5E_TC_FLOW_NIC	= BIT(1),
 };
 
 struct mlx5e_tc_flow {
@@ -58,7 +66,16 @@ struct mlx5e_tc_flow {
 	u8			flags;
 	struct mlx5_flow_handle *rule;
 	struct list_head	encap; /* flows sharing the same encap */
-	struct mlx5_esw_flow_attr *attr;
+	union {
+		struct mlx5_esw_flow_attr esw_attr[0];
+		struct mlx5_nic_flow_attr nic_attr[0];
+	};
+};
+
+struct mlx5e_tc_flow_parse_attr {
+	struct mlx5_flow_spec spec;
+	int num_mod_hdr_actions;
+	void *mod_hdr_actions;
 };
 
 enum {
@@ -71,24 +88,26 @@ enum {
 
 static struct mlx5_flow_handle *
 mlx5e_tc_add_nic_flow(struct mlx5e_priv *priv,
-		      struct mlx5_flow_spec *spec,
-		      u32 action, u32 flow_tag)
+		      struct mlx5e_tc_flow_parse_attr *parse_attr,
+		      struct mlx5e_tc_flow *flow)
 {
+	struct mlx5_nic_flow_attr *attr = flow->nic_attr;
 	struct mlx5_core_dev *dev = priv->mdev;
-	struct mlx5_flow_destination dest = { 0 };
+	struct mlx5_flow_destination dest = {};
 	struct mlx5_flow_act flow_act = {
-		.action = action,
-		.flow_tag = flow_tag,
+		.action = attr->action,
+		.flow_tag = attr->flow_tag,
 		.encap_id = 0,
 	};
 	struct mlx5_fc *counter = NULL;
 	struct mlx5_flow_handle *rule;
 	bool table_created = false;
+	int err;
 
-	if (action & MLX5_FLOW_CONTEXT_ACTION_FWD_DEST) {
+	if (attr->action & MLX5_FLOW_CONTEXT_ACTION_FWD_DEST) {
 		dest.type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE;
 		dest.ft = priv->fs.vlan.ft.t;
-	} else if (action & MLX5_FLOW_CONTEXT_ACTION_COUNT) {
+	} else if (attr->action & MLX5_FLOW_CONTEXT_ACTION_COUNT) {
 		counter = mlx5_fc_create(dev, true);
 		if (IS_ERR(counter))
 			return ERR_CAST(counter);
@@ -97,6 +116,19 @@ mlx5e_tc_add_nic_flow(struct mlx5e_priv *priv,
 		dest.counter = counter;
 	}
 
+	if (attr->action & MLX5_FLOW_CONTEXT_ACTION_MOD_HDR) {
+		err = mlx5_modify_header_alloc(dev, MLX5_FLOW_NAMESPACE_KERNEL,
+					       parse_attr->num_mod_hdr_actions,
+					       parse_attr->mod_hdr_actions,
+					       &attr->mod_hdr_id);
+		flow_act.modify_id = attr->mod_hdr_id;
+		kfree(parse_attr->mod_hdr_actions);
+		if (err) {
+			rule = ERR_PTR(err);
+			goto err_create_mod_hdr_id;
+		}
+	}
+
 	if (IS_ERR_OR_NULL(priv->fs.tc.t)) {
 		priv->fs.tc.t =
 			mlx5_create_auto_grouped_flow_table(priv->fs.ns,
@@ -114,8 +146,9 @@ mlx5e_tc_add_nic_flow(struct mlx5e_priv *priv,
 		table_created = true;
 	}
 
-	spec->match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
-	rule = mlx5_add_flow_rules(priv->fs.tc.t, spec, &flow_act, &dest, 1);
+	parse_attr->spec.match_criteria_enable = MLX5_MATCH_OUTER_HEADERS;
+	rule = mlx5_add_flow_rules(priv->fs.tc.t, &parse_attr->spec,
+				   &flow_act, &dest, 1);
 
 	if (IS_ERR(rule))
 		goto err_add_rule;
@@ -128,6 +161,10 @@ err_add_rule:
 		priv->fs.tc.t = NULL;
 	}
 err_create_ft:
+	if (attr->action & MLX5_FLOW_CONTEXT_ACTION_MOD_HDR)
+		mlx5_modify_header_dealloc(priv->mdev,
+					   attr->mod_hdr_id);
+err_create_mod_hdr_id:
 	mlx5_fc_destroy(dev, counter);
 
 	return rule;
@@ -138,47 +175,85 @@ static void mlx5e_tc_del_nic_flow(struct mlx5e_priv *priv,
 {
 	struct mlx5_fc *counter = NULL;
 
-	if (!IS_ERR(flow->rule)) {
-		counter = mlx5_flow_rule_counter(flow->rule);
-		mlx5_del_flow_rules(flow->rule);
-		mlx5_fc_destroy(priv->mdev, counter);
-	}
+	counter = mlx5_flow_rule_counter(flow->rule);
+	mlx5_del_flow_rules(flow->rule);
+	mlx5_fc_destroy(priv->mdev, counter);
 
 	if (!mlx5e_tc_num_filters(priv) && (priv->fs.tc.t)) {
 		mlx5_destroy_flow_table(priv->fs.tc.t);
 		priv->fs.tc.t = NULL;
 	}
+
+	if (flow->nic_attr->action & MLX5_FLOW_CONTEXT_ACTION_MOD_HDR)
+		mlx5_modify_header_dealloc(priv->mdev,
+					   flow->nic_attr->mod_hdr_id);
 }
 
+static void mlx5e_detach_encap(struct mlx5e_priv *priv,
+			       struct mlx5e_tc_flow *flow);
+
 static struct mlx5_flow_handle *
 mlx5e_tc_add_fdb_flow(struct mlx5e_priv *priv,
-		      struct mlx5_flow_spec *spec,
-		      struct mlx5_esw_flow_attr *attr)
+		      struct mlx5e_tc_flow_parse_attr *parse_attr,
+		      struct mlx5e_tc_flow *flow)
 {
 	struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
+	struct mlx5_esw_flow_attr *attr = flow->esw_attr;
+	struct mlx5_flow_handle *rule;
 	int err;
 
 	err = mlx5_eswitch_add_vlan_action(esw, attr);
-	if (err)
-		return ERR_PTR(err);
+	if (err) {
+		rule = ERR_PTR(err);
+		goto err_add_vlan;
+	}
 
-	return mlx5_eswitch_add_offloaded_rule(esw, spec, attr);
-}
+	if (attr->action & MLX5_FLOW_CONTEXT_ACTION_MOD_HDR) {
+		err = mlx5_modify_header_alloc(priv->mdev, MLX5_FLOW_NAMESPACE_FDB,
+					       parse_attr->num_mod_hdr_actions,
+					       parse_attr->mod_hdr_actions,
+					       &attr->mod_hdr_id);
+		kfree(parse_attr->mod_hdr_actions);
+		if (err) {
+			rule = ERR_PTR(err);
+			goto err_mod_hdr;
+		}
+	}
 
-static void mlx5e_detach_encap(struct mlx5e_priv *priv,
-			       struct mlx5e_tc_flow *flow);
+	rule = mlx5_eswitch_add_offloaded_rule(esw, &parse_attr->spec, attr);
+	if (IS_ERR(rule))
+		goto err_add_rule;
+
+	return rule;
+
+err_add_rule:
+	if (flow->esw_attr->action & MLX5_FLOW_CONTEXT_ACTION_MOD_HDR)
+		mlx5_modify_header_dealloc(priv->mdev,
+					   attr->mod_hdr_id);
+err_mod_hdr:
+	mlx5_eswitch_del_vlan_action(esw, attr);
+err_add_vlan:
+	if (attr->action & MLX5_FLOW_CONTEXT_ACTION_ENCAP)
+		mlx5e_detach_encap(priv, flow);
+	return rule;
+}
 
 static void mlx5e_tc_del_fdb_flow(struct mlx5e_priv *priv,
 				  struct mlx5e_tc_flow *flow)
 {
 	struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
+	struct mlx5_esw_flow_attr *attr = flow->esw_attr;
 
-	mlx5_eswitch_del_offloaded_rule(esw, flow->rule, flow->attr);
+	mlx5_eswitch_del_offloaded_rule(esw, flow->rule, flow->esw_attr);
 
-	mlx5_eswitch_del_vlan_action(esw, flow->attr);
+	mlx5_eswitch_del_vlan_action(esw, flow->esw_attr);
 
-	if (flow->attr->action & MLX5_FLOW_CONTEXT_ACTION_ENCAP)
+	if (flow->esw_attr->action & MLX5_FLOW_CONTEXT_ACTION_ENCAP)
 		mlx5e_detach_encap(priv, flow);
+
+	if (flow->esw_attr->action & MLX5_FLOW_CONTEXT_ACTION_MOD_HDR)
+		mlx5_modify_header_dealloc(priv->mdev,
+					   attr->mod_hdr_id);
 }
 
 static void mlx5e_detach_encap(struct mlx5e_priv *priv,
@@ -200,10 +275,6 @@ static void mlx5e_detach_encap(struct mlx5e_priv *priv,
 	}
 }
 
-/* we get here also when setting rule to the FW failed, etc. It means that the
- * flow rule itself might not exist, but some offloading related to the actions
- * should be cleaned.
- */
 static void mlx5e_tc_del_flow(struct mlx5e_priv *priv,
 			      struct mlx5e_tc_flow *flow)
 {
@@ -650,29 +721,313 @@ static int parse_cls_flower(struct mlx5e_priv *priv,
 	return err;
 }
 
+struct pedit_headers {
+	struct ethhdr  eth;
+	struct iphdr   ip4;
+	struct ipv6hdr ip6;
+	struct tcphdr  tcp;
+	struct udphdr  udp;
+};
+
+static int pedit_header_offsets[] = {
+	[TCA_PEDIT_KEY_EX_HDR_TYPE_ETH] = offsetof(struct pedit_headers, eth),
+	[TCA_PEDIT_KEY_EX_HDR_TYPE_IP4] = offsetof(struct pedit_headers, ip4),
+	[TCA_PEDIT_KEY_EX_HDR_TYPE_IP6] = offsetof(struct pedit_headers, ip6),
+	[TCA_PEDIT_KEY_EX_HDR_TYPE_TCP] = offsetof(struct pedit_headers, tcp),
+	[TCA_PEDIT_KEY_EX_HDR_TYPE_UDP] = offsetof(struct pedit_headers, udp),
+};
+
+#define pedit_header(_ph, _htype) ((void *)(_ph) + pedit_header_offsets[_htype])
+
+static int set_pedit_val(u8 hdr_type, u32 mask, u32 val, u32 offset,
+			 struct pedit_headers *masks,
+			 struct pedit_headers *vals)
+{
+	u32 *curr_pmask, *curr_pval;
+
+	if (hdr_type >= __PEDIT_HDR_TYPE_MAX)
+		goto out_err;
+
+	curr_pmask = (u32 *)(pedit_header(masks, hdr_type) + offset);
+	curr_pval  = (u32 *)(pedit_header(vals, hdr_type) + offset);
+
+	if (*curr_pmask & mask)  /* disallow acting twice on the same location */
+		goto out_err;
+
+	*curr_pmask |= mask;
+	*curr_pval  |= (val & mask);
+
+	return 0;
+
+out_err:
+	return -EOPNOTSUPP;
+}
+
+struct mlx5_fields {
+	u8  field;
+	u8  size;
+	u32 offset;
+};
+
+static struct mlx5_fields fields[] = {
+	{MLX5_ACTION_IN_FIELD_OUT_DMAC_47_16, 4, offsetof(struct pedit_headers, eth.h_dest[0])},
+	{MLX5_ACTION_IN_FIELD_OUT_DMAC_15_0,  2, offsetof(struct pedit_headers, eth.h_dest[4])},
+	{MLX5_ACTION_IN_FIELD_OUT_SMAC_47_16, 4, offsetof(struct pedit_headers, eth.h_source[0])},
+	{MLX5_ACTION_IN_FIELD_OUT_SMAC_15_0,  2, offsetof(struct pedit_headers, eth.h_source[4])},
+	{MLX5_ACTION_IN_FIELD_OUT_ETHERTYPE,  2, offsetof(struct pedit_headers, eth.h_proto)},
+
+	{MLX5_ACTION_IN_FIELD_OUT_IP_DSCP, 1, offsetof(struct pedit_headers, ip4.tos)},
+	{MLX5_ACTION_IN_FIELD_OUT_IP_TTL,  1, offsetof(struct pedit_headers, ip4.ttl)},
+	{MLX5_ACTION_IN_FIELD_OUT_SIPV4,   4, offsetof(struct pedit_headers, ip4.saddr)},
+	{MLX5_ACTION_IN_FIELD_OUT_DIPV4,   4, offsetof(struct pedit_headers, ip4.daddr)},
+
+	{MLX5_ACTION_IN_FIELD_OUT_SIPV6_127_96, 4, offsetof(struct pedit_headers, ip6.saddr.s6_addr32[0])},
+	{MLX5_ACTION_IN_FIELD_OUT_SIPV6_95_64,  4, offsetof(struct pedit_headers, ip6.saddr.s6_addr32[1])},
+	{MLX5_ACTION_IN_FIELD_OUT_SIPV6_63_32,  4, offsetof(struct pedit_headers, ip6.saddr.s6_addr32[2])},
+	{MLX5_ACTION_IN_FIELD_OUT_SIPV6_31_0,   4, offsetof(struct pedit_headers, ip6.saddr.s6_addr32[3])},
+	{MLX5_ACTION_IN_FIELD_OUT_DIPV6_127_96, 4, offsetof(struct pedit_headers, ip6.daddr.s6_addr32[0])},
+	{MLX5_ACTION_IN_FIELD_OUT_DIPV6_95_64,  4, offsetof(struct pedit_headers, ip6.daddr.s6_addr32[1])},
+	{MLX5_ACTION_IN_FIELD_OUT_DIPV6_63_32,  4, offsetof(struct pedit_headers, ip6.daddr.s6_addr32[2])},
+	{MLX5_ACTION_IN_FIELD_OUT_DIPV6_31_0,   4, offsetof(struct pedit_headers, ip6.daddr.s6_addr32[3])},
+
+	{MLX5_ACTION_IN_FIELD_OUT_TCP_SPORT, 2, offsetof(struct pedit_headers, tcp.source)},
+	{MLX5_ACTION_IN_FIELD_OUT_TCP_DPORT, 2, offsetof(struct pedit_headers, tcp.dest)},
+	{MLX5_ACTION_IN_FIELD_OUT_TCP_FLAGS, 1, offsetof(struct pedit_headers, tcp.ack_seq) + 5},
+
+	{MLX5_ACTION_IN_FIELD_OUT_UDP_SPORT, 2, offsetof(struct pedit_headers, udp.source)},
+	{MLX5_ACTION_IN_FIELD_OUT_UDP_DPORT, 2, offsetof(struct pedit_headers, udp.dest)},
+};
+
+/* On input attr->num_mod_hdr_actions tells how many HW actions can be parsed at
+ * max from the SW pedit action. On success, it says how many HW actions were
+ * actually parsed.
+ */
+static int offload_pedit_fields(struct pedit_headers *masks,
+				struct pedit_headers *vals,
+				struct mlx5e_tc_flow_parse_attr *parse_attr)
+{
+	struct pedit_headers *set_masks, *add_masks, *set_vals, *add_vals;
+	int i, action_size, nactions, max_actions, first, last;
+	void *s_masks_p, *a_masks_p, *vals_p;
+	u32 s_mask, a_mask, val;
+	struct mlx5_fields *f;
+	u8 cmd, field_bsize;
+	unsigned long mask;
+	void *action;
+
+	set_masks = &masks[TCA_PEDIT_KEY_EX_CMD_SET];
+	add_masks = &masks[TCA_PEDIT_KEY_EX_CMD_ADD];
+	set_vals = &vals[TCA_PEDIT_KEY_EX_CMD_SET];
+	add_vals = &vals[TCA_PEDIT_KEY_EX_CMD_ADD];
+
+	action_size = MLX5_UN_SZ_BYTES(set_action_in_add_action_in_auto);
+	action = parse_attr->mod_hdr_actions;
+	max_actions = parse_attr->num_mod_hdr_actions;
+	nactions = 0;
+
+	for (i = 0; i < ARRAY_SIZE(fields); i++) {
+		f = &fields[i];
+		/* avoid seeing bits set from previous iterations */
+		s_mask = a_mask = mask = val = 0;
+
+		s_masks_p = (void *)set_masks + f->offset;
+		a_masks_p = (void *)add_masks + f->offset;
+
+		memcpy(&s_mask, s_masks_p, f->size);
+		memcpy(&a_mask, a_masks_p, f->size);
+
+		if (!s_mask && !a_mask) /* nothing to offload here */
+			continue;
+
+		if (s_mask && a_mask) {
+			printk(KERN_WARNING "mlx5: can't set and add to the same HW field (%x)\n", f->field);
+			return -EOPNOTSUPP;
+		}
+
+		if (nactions == max_actions) {
+			printk(KERN_WARNING "mlx5: parsed %d pedit actions, can't do more\n", nactions);
+			return -EOPNOTSUPP;
+		}
+
+		if (s_mask) {
+			cmd  = MLX5_ACTION_TYPE_SET;
+			mask = s_mask;
+			vals_p = (void *)set_vals + f->offset;
+			/* clear to denote we consumed this field */
+			memset(s_masks_p, 0, f->size);
+		} else {
+			cmd  = MLX5_ACTION_TYPE_ADD;
+			mask = a_mask;
+			vals_p = (void *)add_vals + f->offset;
+			/* clear to denote we consumed this field */
+			memset(a_masks_p, 0, f->size);
+		}
+
+		memcpy(&val, vals_p, f->size);
+
+		field_bsize = f->size * BITS_PER_BYTE;
+		first = find_first_bit(&mask, field_bsize);
+		last  = find_last_bit(&mask, field_bsize);
+		if (first > 0 || last != (field_bsize - 1)) {
+			printk(KERN_WARNING "mlx5: partial rewrite (mask %lx) is currently not offloaded\n",
+			       mask);
+			return -EOPNOTSUPP;
+		}
+
+		MLX5_SET(set_action_in, action, action_type, cmd);
+		MLX5_SET(set_action_in, action, field, f->field);
+
+		if (cmd == MLX5_ACTION_TYPE_SET) {
+			MLX5_SET(set_action_in, action, offset, 0);
+			/* length is num of bits to be written, zero means length of 32 */
+			MLX5_SET(set_action_in, action, length, field_bsize);
+		}
+
+		if (field_bsize == 32)
+			MLX5_SET(set_action_in, action, data, ntohl(val));
+		else if (field_bsize == 16)
+			MLX5_SET(set_action_in, action, data, ntohs(val));
+		else if (field_bsize == 8)
+			MLX5_SET(set_action_in, action, data, val);
+
+		action += action_size;
+		nactions++;
+	}
+
+	parse_attr->num_mod_hdr_actions = nactions;
+	return 0;
+}
+
+static int alloc_mod_hdr_actions(struct mlx5e_priv *priv,
+				 const struct tc_action *a, int namespace,
+				 struct mlx5e_tc_flow_parse_attr *parse_attr)
+{
+	int nkeys, action_size, max_actions;
+
+	nkeys = tcf_pedit_nkeys(a);
+	action_size = MLX5_UN_SZ_BYTES(set_action_in_add_action_in_auto);
+
+	if (namespace == MLX5_FLOW_NAMESPACE_FDB) /* FDB offloading */
+		max_actions = MLX5_CAP_ESW_FLOWTABLE_FDB(priv->mdev, max_modify_header_actions);
+	else /* namespace is MLX5_FLOW_NAMESPACE_KERNEL - NIC offloading */
+		max_actions = MLX5_CAP_FLOWTABLE_NIC_RX(priv->mdev, max_modify_header_actions);
+
+	/* can get up to crazingly 16 HW actions in 32 bits pedit SW key */
+	max_actions = min(max_actions, nkeys * 16);
+
+	parse_attr->mod_hdr_actions = kcalloc(max_actions, action_size, GFP_KERNEL);
+	if (!parse_attr->mod_hdr_actions)
+		return -ENOMEM;
+
+	parse_attr->num_mod_hdr_actions = max_actions;
+	return 0;
+}
+
+static const struct pedit_headers zero_masks = {};
+
+static int parse_tc_pedit_action(struct mlx5e_priv *priv,
+				 const struct tc_action *a, int namespace,
+				 struct mlx5e_tc_flow_parse_attr *parse_attr)
+{
+	struct pedit_headers masks[__PEDIT_CMD_MAX], vals[__PEDIT_CMD_MAX], *cmd_masks;
+	int nkeys, i, err = -EOPNOTSUPP;
+	u32 mask, val, offset;
+	u8 cmd, htype;
+
+	nkeys = tcf_pedit_nkeys(a);
+
+	memset(masks, 0, sizeof(struct pedit_headers) * __PEDIT_CMD_MAX);
+	memset(vals,  0, sizeof(struct pedit_headers) * __PEDIT_CMD_MAX);
+
+	for (i = 0; i < nkeys; i++) {
+		htype = tcf_pedit_htype(a, i);
+		cmd = tcf_pedit_cmd(a, i);
+		err = -EOPNOTSUPP; /* can't be all optimistic */
+
+		if (htype == TCA_PEDIT_KEY_EX_HDR_TYPE_NETWORK) {
+			printk(KERN_WARNING "mlx5: legacy pedit isn't offloaded\n");
+			goto out_err;
+		}
+
+		if (cmd != TCA_PEDIT_KEY_EX_CMD_SET && cmd != TCA_PEDIT_KEY_EX_CMD_ADD) {
+			printk(KERN_WARNING "mlx5: pedit cmd %d isn't offloaded\n", cmd);
+			goto out_err;
+		}
+
+		mask = tcf_pedit_mask(a, i);
+		val = tcf_pedit_val(a, i);
+		offset = tcf_pedit_offset(a, i);
+
+		err = set_pedit_val(htype, ~mask, val, offset, &masks[cmd], &vals[cmd]);
+		if (err)
+			goto out_err;
+	}
+
+	err = alloc_mod_hdr_actions(priv, a, namespace, parse_attr);
+	if (err)
+		goto out_err;
+
+	err = offload_pedit_fields(masks, vals, parse_attr);
+	if (err < 0)
+		goto out_dealloc_parsed_actions;
+
+	for (cmd = 0; cmd < __PEDIT_CMD_MAX; cmd++) {
+		cmd_masks = &masks[cmd];
+		if (memcmp(cmd_masks, &zero_masks, sizeof(zero_masks))) {
+			printk(KERN_WARNING "mlx5: attempt to offload an unsupported field (cmd %d)\n",
+			       cmd);
+			print_hex_dump(KERN_WARNING, "mask: ", DUMP_PREFIX_ADDRESS,
+				       16, 1, cmd_masks, sizeof(zero_masks), true);
+			err = -EOPNOTSUPP;
+			goto out_dealloc_parsed_actions;
+		}
+	}
+
+	return 0;
+
+out_dealloc_parsed_actions:
+	kfree(parse_attr->mod_hdr_actions);
+out_err:
+	return err;
+}
+
 static int parse_tc_nic_actions(struct mlx5e_priv *priv, struct tcf_exts *exts,
-				u32 *action, u32 *flow_tag)
+				struct mlx5e_tc_flow_parse_attr *parse_attr,
+				struct mlx5e_tc_flow *flow)
 {
+	struct mlx5_nic_flow_attr *attr = flow->nic_attr;
 	const struct tc_action *a;
 	LIST_HEAD(actions);
+	int err;
 
 	if (tc_no_actions(exts))
 		return -EINVAL;
 
-	*flow_tag = MLX5_FS_DEFAULT_FLOW_TAG;
-	*action = 0;
+	attr->flow_tag = MLX5_FS_DEFAULT_FLOW_TAG;
+	attr->action = 0;
 
 	tcf_exts_to_list(exts, &actions);
 	list_for_each_entry(a, &actions, list) {
 		/* Only support a single action per rule */
-		if (*action)
+		if (attr->action)
 			return -EINVAL;
 
 		if (is_tcf_gact_shot(a)) {
-			*action |= MLX5_FLOW_CONTEXT_ACTION_DROP;
+			attr->action |= MLX5_FLOW_CONTEXT_ACTION_DROP;
 			if (MLX5_CAP_FLOWTABLE(priv->mdev,
 					       flow_table_properties_nic_receive.flow_counter))
-				*action |= MLX5_FLOW_CONTEXT_ACTION_COUNT;
+				attr->action |= MLX5_FLOW_CONTEXT_ACTION_COUNT;
+			continue;
+		}
+
+		if (is_tcf_pedit(a)) {
+			err = parse_tc_pedit_action(priv, a, MLX5_FLOW_NAMESPACE_KERNEL,
+						    parse_attr);
+			if (err)
+				return err;
+
+			attr->action |= MLX5_FLOW_CONTEXT_ACTION_MOD_HDR |
+					MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
 			continue;
 		}
 
@@ -685,8 +1040,8 @@ static int parse_tc_nic_actions(struct mlx5e_priv *priv, struct tcf_exts *exts,
 				return -EINVAL;
 			}
 
-			*flow_tag = mark;
-			*action |= MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
+			attr->flow_tag = mark;
+			attr->action |= MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
 			continue;
 		}
 
@@ -1071,9 +1426,10 @@ out_err:
 }
 
 static int parse_tc_fdb_actions(struct mlx5e_priv *priv, struct tcf_exts *exts,
+				struct mlx5e_tc_flow_parse_attr *parse_attr,
 				struct mlx5e_tc_flow *flow)
 {
-	struct mlx5_esw_flow_attr *attr = flow->attr;
+	struct mlx5_esw_flow_attr *attr = flow->esw_attr;
 	struct ip_tunnel_info *info = NULL;
 	const struct tc_action *a;
 	LIST_HEAD(actions);
@@ -1094,6 +1450,16 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv, struct tcf_exts *exts,
 			continue;
 		}
 
+		if (is_tcf_pedit(a)) {
+			err = parse_tc_pedit_action(priv, a, MLX5_FLOW_NAMESPACE_FDB,
+						    parse_attr);
+			if (err)
+				return err;
+
+			attr->action |= MLX5_FLOW_CONTEXT_ACTION_MOD_HDR;
+			continue;
+		}
+
 		if (is_tcf_mirred_egress_redirect(a)) {
 			int ifindex = tcf_mirred_ifindex(a);
 			struct net_device *out_dev;
@@ -1163,22 +1529,24 @@ static int parse_tc_fdb_actions(struct mlx5e_priv *priv, struct tcf_exts *exts,
 int mlx5e_configure_flower(struct mlx5e_priv *priv, __be16 protocol,
 			   struct tc_cls_flower_offload *f)
 {
+	struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
+	struct mlx5e_tc_flow_parse_attr *parse_attr;
 	struct mlx5e_tc_table *tc = &priv->fs.tc;
-	int err, attr_size = 0;
-	u32 flow_tag, action;
 	struct mlx5e_tc_flow *flow;
-	struct mlx5_flow_spec *spec;
-	struct mlx5_eswitch *esw = priv->mdev->priv.eswitch;
+	int attr_size, err = 0;
 	u8 flow_flags = 0;
 
 	if (esw && esw->mode == SRIOV_OFFLOADS) {
 		flow_flags = MLX5E_TC_FLOW_ESWITCH;
 		attr_size  = sizeof(struct mlx5_esw_flow_attr);
+	} else {
+		flow_flags = MLX5E_TC_FLOW_NIC;
+		attr_size  = sizeof(struct mlx5_nic_flow_attr);
 	}
 
 	flow = kzalloc(sizeof(*flow) + attr_size, GFP_KERNEL);
-	spec = mlx5_vzalloc(sizeof(*spec));
-	if (!spec || !flow) {
+	parse_attr = mlx5_vzalloc(sizeof(*parse_attr));
+	if (!parse_attr || !flow) {
 		err = -ENOMEM;
 		goto err_free;
 	}
@@ -1186,26 +1554,25 @@ int mlx5e_configure_flower(struct mlx5e_priv *priv, __be16 protocol,
 	flow->cookie = f->cookie;
 	flow->flags = flow_flags;
 
-	err = parse_cls_flower(priv, flow, spec, f);
+	err = parse_cls_flower(priv, flow, &parse_attr->spec, f);
 	if (err < 0)
 		goto err_free;
 
 	if (flow->flags & MLX5E_TC_FLOW_ESWITCH) {
-		flow->attr  = (struct mlx5_esw_flow_attr *)(flow + 1);
-		err = parse_tc_fdb_actions(priv, f->exts, flow);
+		err = parse_tc_fdb_actions(priv, f->exts, parse_attr, flow);
 		if (err < 0)
 			goto err_free;
-		flow->rule = mlx5e_tc_add_fdb_flow(priv, spec, flow->attr);
+		flow->rule = mlx5e_tc_add_fdb_flow(priv, parse_attr, flow);
 	} else {
-		err = parse_tc_nic_actions(priv, f->exts, &action, &flow_tag);
+		err = parse_tc_nic_actions(priv, f->exts, parse_attr, flow);
 		if (err < 0)
 			goto err_free;
-		flow->rule = mlx5e_tc_add_nic_flow(priv, spec, action, flow_tag);
+		flow->rule = mlx5e_tc_add_nic_flow(priv, parse_attr, flow);
 	}
 
 	if (IS_ERR(flow->rule)) {
 		err = PTR_ERR(flow->rule);
-		goto err_del_rule;
+		goto err_free;
 	}
 
 	err = rhashtable_insert_fast(&tc->ht, &flow->node,
@@ -1221,7 +1588,7 @@ err_del_rule:
 err_free:
 	kfree(flow);
 out:
-	kvfree(spec);
+	kvfree(parse_attr);
 	return err;
 }
 

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

@@ -285,8 +285,8 @@ enum {
 	SET_VLAN_INSERT	= BIT(1)
 };
 
-#define MLX5_FLOW_CONTEXT_ACTION_VLAN_POP  0x40
-#define MLX5_FLOW_CONTEXT_ACTION_VLAN_PUSH 0x80
+#define MLX5_FLOW_CONTEXT_ACTION_VLAN_POP  0x4000
+#define MLX5_FLOW_CONTEXT_ACTION_VLAN_PUSH 0x8000
 
 struct mlx5_encap_entry {
 	struct hlist_node encap_hlist;
@@ -308,6 +308,7 @@ struct mlx5_esw_flow_attr {
 	u16	vlan;
 	bool	vlan_handled;
 	struct mlx5_encap_entry *encap;
+	u32	mod_hdr_id;
 };
 
 int mlx5_eswitch_sqs2vport_start(struct mlx5_eswitch *esw,

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

@@ -68,8 +68,10 @@ mlx5_eswitch_add_offloaded_rule(struct mlx5_eswitch *esw,
 	}
 	if (flow_act.action & MLX5_FLOW_CONTEXT_ACTION_COUNT) {
 		counter = mlx5_fc_create(esw->dev, true);
-		if (IS_ERR(counter))
-			return ERR_CAST(counter);
+		if (IS_ERR(counter)) {
+			rule = ERR_CAST(counter);
+			goto err_counter_alloc;
+		}
 		dest[i].type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
 		dest[i].counter = counter;
 		i++;
@@ -86,17 +88,25 @@ mlx5_eswitch_add_offloaded_rule(struct mlx5_eswitch *esw,
 	if (flow_act.action & MLX5_FLOW_CONTEXT_ACTION_DECAP)
 		spec->match_criteria_enable |= MLX5_MATCH_INNER_HEADERS;
 
-	if (attr->encap)
+	if (attr->action & MLX5_FLOW_CONTEXT_ACTION_MOD_HDR)
+		flow_act.modify_id = attr->mod_hdr_id;
+
+	if (attr->action & MLX5_FLOW_CONTEXT_ACTION_ENCAP)
 		flow_act.encap_id = attr->encap->encap_id;
 
 	rule = mlx5_add_flow_rules((struct mlx5_flow_table *)esw->fdb_table.fdb,
 				   spec, &flow_act, dest, i);
 	if (IS_ERR(rule))
-		mlx5_fc_destroy(esw->dev, counter);
+		goto err_add_rule;
 	else
 		esw->offloads.num_flows++;
 
 	return rule;
+
+err_add_rule:
+	mlx5_fc_destroy(esw->dev, counter);
+err_counter_alloc:
+	return rule;
 }
 
 void
@@ -106,12 +116,10 @@ mlx5_eswitch_del_offloaded_rule(struct mlx5_eswitch *esw,
 {
 	struct mlx5_fc *counter = NULL;
 
-	if (!IS_ERR(rule)) {
-		counter = mlx5_flow_rule_counter(rule);
-		mlx5_del_flow_rules(rule);
-		mlx5_fc_destroy(esw->dev, counter);
-		esw->offloads.num_flows--;
-	}
+	counter = mlx5_flow_rule_counter(rule);
+	mlx5_del_flow_rules(rule);
+	mlx5_fc_destroy(esw->dev, counter);
+	esw->offloads.num_flows--;
 }
 
 static int esw_set_global_vlan_pop(struct mlx5_eswitch *esw, u8 val)

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

@@ -249,6 +249,7 @@ static int mlx5_cmd_set_fte(struct mlx5_core_dev *dev,
 	MLX5_SET(flow_context, in_flow_context, flow_tag, fte->flow_tag);
 	MLX5_SET(flow_context, in_flow_context, action, fte->action);
 	MLX5_SET(flow_context, in_flow_context, encap_id, fte->encap_id);
+	MLX5_SET(flow_context, in_flow_context, modify_header_id, fte->modify_id);
 	in_match_value = MLX5_ADDR_OF(flow_context, in_flow_context,
 				      match_value);
 	memcpy(in_match_value, &fte->val, MLX5_ST_SZ_BYTES(fte_match_param));
@@ -515,3 +516,69 @@ void mlx5_encap_dealloc(struct mlx5_core_dev *dev, u32 encap_id)
 
 	mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
 }
+
+int mlx5_modify_header_alloc(struct mlx5_core_dev *dev,
+			     u8 namespace, u8 num_actions,
+			     void *modify_actions, u32 *modify_header_id)
+{
+	u32 out[MLX5_ST_SZ_DW(alloc_modify_header_context_out)];
+	int max_actions, actions_size, inlen, err;
+	void *actions_in;
+	u8 table_type;
+	u32 *in;
+
+	switch (namespace) {
+	case MLX5_FLOW_NAMESPACE_FDB:
+		max_actions = MLX5_CAP_ESW_FLOWTABLE_FDB(dev, max_modify_header_actions);
+		table_type = FS_FT_FDB;
+		break;
+	case MLX5_FLOW_NAMESPACE_KERNEL:
+		max_actions = MLX5_CAP_FLOWTABLE_NIC_RX(dev, max_modify_header_actions);
+		table_type = FS_FT_NIC_RX;
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
+
+	if (num_actions > max_actions) {
+		mlx5_core_warn(dev, "too many modify header actions %d, max supported %d\n",
+			       num_actions, max_actions);
+		return -EOPNOTSUPP;
+	}
+
+	actions_size = MLX5_UN_SZ_BYTES(set_action_in_add_action_in_auto) * num_actions;
+	inlen = MLX5_ST_SZ_BYTES(alloc_modify_header_context_in) + actions_size;
+
+	in = kzalloc(inlen, GFP_KERNEL);
+	if (!in)
+		return -ENOMEM;
+
+	MLX5_SET(alloc_modify_header_context_in, in, opcode,
+		 MLX5_CMD_OP_ALLOC_MODIFY_HEADER_CONTEXT);
+	MLX5_SET(alloc_modify_header_context_in, in, table_type, table_type);
+	MLX5_SET(alloc_modify_header_context_in, in, num_of_actions, num_actions);
+
+	actions_in = MLX5_ADDR_OF(alloc_modify_header_context_in, in, actions);
+	memcpy(actions_in, modify_actions, actions_size);
+
+	memset(out, 0, sizeof(out));
+	err = mlx5_cmd_exec(dev, in, inlen, out, sizeof(out));
+
+	*modify_header_id = MLX5_GET(alloc_modify_header_context_out, out, modify_header_id);
+	kfree(in);
+	return err;
+}
+
+void mlx5_modify_header_dealloc(struct mlx5_core_dev *dev, u32 modify_header_id)
+{
+	u32 in[MLX5_ST_SZ_DW(dealloc_modify_header_context_in)];
+	u32 out[MLX5_ST_SZ_DW(dealloc_modify_header_context_out)];
+
+	memset(in, 0, sizeof(in));
+	MLX5_SET(dealloc_modify_header_context_in, in, opcode,
+		 MLX5_CMD_OP_DEALLOC_MODIFY_HEADER_CONTEXT);
+	MLX5_SET(dealloc_modify_header_context_in, in, modify_header_id,
+		 modify_header_id);
+
+	mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
+}

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

@@ -476,6 +476,7 @@ static struct fs_fte *alloc_fte(struct mlx5_flow_act *flow_act,
 	fte->index = index;
 	fte->action = flow_act->action;
 	fte->encap_id = flow_act->encap_id;
+	fte->modify_id = flow_act->modify_id;
 
 	return fte;
 }

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

@@ -152,6 +152,7 @@ struct fs_fte {
 	u32				index;
 	u32				action;
 	u32				encap_id;
+	u32				modify_id;
 	enum fs_fte_status		status;
 	struct mlx5_fc			*counter;
 };

drivers/net/ethernet/mellanox/mlx5/core/mlx5_core.h

@@ -141,6 +141,11 @@ int mlx5_encap_alloc(struct mlx5_core_dev *dev,
 		     u32 *encap_id);
 void mlx5_encap_dealloc(struct mlx5_core_dev *dev, u32 encap_id);
 
+int mlx5_modify_header_alloc(struct mlx5_core_dev *dev,
+			     u8 namespace, u8 num_actions,
+			     void *modify_actions, u32 *modify_header_id);
+void mlx5_modify_header_dealloc(struct mlx5_core_dev *dev, u32 modify_header_id);
+
 bool mlx5_lag_intf_add(struct mlx5_interface *intf, struct mlx5_priv *priv);
 
 int mlx5_query_mtpps(struct mlx5_core_dev *dev, u32 *mtpps, u32 mtpps_size);

include/linux/mlx5/fs.h

@@ -134,8 +134,13 @@ struct mlx5_flow_act {
 	u32 action;
 	u32 flow_tag;
 	u32 encap_id;
+	u32 modify_id;
 };
 
+#define MLX5_DECLARE_FLOW_ACT(name) \
+	struct mlx5_flow_act name = {MLX5_FLOW_CONTEXT_ACTION_FWD_DEST,\
+				     MLX5_FS_DEFAULT_FLOW_TAG, 0, 0}
+
 /* Single destination per rule.
  * Group ID is implied by the match criteria.
  */
@@ -156,5 +161,4 @@ struct mlx5_fc *mlx5_fc_create(struct mlx5_core_dev *dev, bool aging);
 void mlx5_fc_destroy(struct mlx5_core_dev *dev, struct mlx5_fc *counter);
 void mlx5_fc_query_cached(struct mlx5_fc *counter,
 			  u64 *bytes, u64 *packets, u64 *lastuse);
-
 #endif

include/linux/mlx5/mlx5_ifc.h

@@ -227,6 +227,8 @@ enum {
 	MLX5_CMD_OP_MODIFY_FLOW_TABLE             = 0x93c,
 	MLX5_CMD_OP_ALLOC_ENCAP_HEADER            = 0x93d,
 	MLX5_CMD_OP_DEALLOC_ENCAP_HEADER          = 0x93e,
+	MLX5_CMD_OP_ALLOC_MODIFY_HEADER_CONTEXT   = 0x940,
+	MLX5_CMD_OP_DEALLOC_MODIFY_HEADER_CONTEXT = 0x941,
 	MLX5_CMD_OP_MAX
 };
 
@@ -302,7 +304,8 @@ struct mlx5_ifc_flow_table_prop_layout_bits {
 
 	u8         reserved_at_20[0x2];
 	u8         log_max_ft_size[0x6];
-	u8         reserved_at_28[0x10];
+	u8         log_max_modify_header_context[0x8];
+	u8         max_modify_header_actions[0x8];
 	u8         max_ft_level[0x8];
 
 	u8         reserved_at_40[0x20];
@@ -2190,6 +2193,7 @@ enum {
 	MLX5_FLOW_CONTEXT_ACTION_COUNT     = 0x8,
 	MLX5_FLOW_CONTEXT_ACTION_ENCAP     = 0x10,
 	MLX5_FLOW_CONTEXT_ACTION_DECAP     = 0x20,
+	MLX5_FLOW_CONTEXT_ACTION_MOD_HDR   = 0x40,
 };
 
 struct mlx5_ifc_flow_context_bits {
@@ -2211,7 +2215,9 @@ struct mlx5_ifc_flow_context_bits {
 
 	u8         encap_id[0x20];
 
-	u8         reserved_at_e0[0x120];
+	u8         modify_header_id[0x20];
+
+	u8         reserved_at_100[0x100];
 
 	struct mlx5_ifc_fte_match_param_bits match_value;
 
@@ -4534,6 +4540,109 @@ struct mlx5_ifc_dealloc_encap_header_in_bits {
 	u8         reserved_60[0x20];
 };
 
+struct mlx5_ifc_set_action_in_bits {
+	u8         action_type[0x4];
+	u8         field[0xc];
+	u8         reserved_at_10[0x3];
+	u8         offset[0x5];
+	u8         reserved_at_18[0x3];
+	u8         length[0x5];
+
+	u8         data[0x20];
+};
+
+struct mlx5_ifc_add_action_in_bits {
+	u8         action_type[0x4];
+	u8         field[0xc];
+	u8         reserved_at_10[0x10];
+
+	u8         data[0x20];
+};
+
+union mlx5_ifc_set_action_in_add_action_in_auto_bits {
+	struct mlx5_ifc_set_action_in_bits set_action_in;
+	struct mlx5_ifc_add_action_in_bits add_action_in;
+	u8         reserved_at_0[0x40];
+};
+
+enum {
+	MLX5_ACTION_TYPE_SET   = 0x1,
+	MLX5_ACTION_TYPE_ADD   = 0x2,
+};
+
+enum {
+	MLX5_ACTION_IN_FIELD_OUT_SMAC_47_16    = 0x1,
+	MLX5_ACTION_IN_FIELD_OUT_SMAC_15_0     = 0x2,
+	MLX5_ACTION_IN_FIELD_OUT_ETHERTYPE     = 0x3,
+	MLX5_ACTION_IN_FIELD_OUT_DMAC_47_16    = 0x4,
+	MLX5_ACTION_IN_FIELD_OUT_DMAC_15_0     = 0x5,
+	MLX5_ACTION_IN_FIELD_OUT_IP_DSCP       = 0x6,
+	MLX5_ACTION_IN_FIELD_OUT_TCP_FLAGS     = 0x7,
+	MLX5_ACTION_IN_FIELD_OUT_TCP_SPORT     = 0x8,
+	MLX5_ACTION_IN_FIELD_OUT_TCP_DPORT     = 0x9,
+	MLX5_ACTION_IN_FIELD_OUT_IP_TTL        = 0xa,
+	MLX5_ACTION_IN_FIELD_OUT_UDP_SPORT     = 0xb,
+	MLX5_ACTION_IN_FIELD_OUT_UDP_DPORT     = 0xc,
+	MLX5_ACTION_IN_FIELD_OUT_SIPV6_127_96  = 0xd,
+	MLX5_ACTION_IN_FIELD_OUT_SIPV6_95_64   = 0xe,
+	MLX5_ACTION_IN_FIELD_OUT_SIPV6_63_32   = 0xf,
+	MLX5_ACTION_IN_FIELD_OUT_SIPV6_31_0    = 0x10,
+	MLX5_ACTION_IN_FIELD_OUT_DIPV6_127_96  = 0x11,
+	MLX5_ACTION_IN_FIELD_OUT_DIPV6_95_64   = 0x12,
+	MLX5_ACTION_IN_FIELD_OUT_DIPV6_63_32   = 0x13,
+	MLX5_ACTION_IN_FIELD_OUT_DIPV6_31_0    = 0x14,
+	MLX5_ACTION_IN_FIELD_OUT_SIPV4         = 0x15,
+	MLX5_ACTION_IN_FIELD_OUT_DIPV4         = 0x16,
+};
+
+struct mlx5_ifc_alloc_modify_header_context_out_bits {
+	u8         status[0x8];
+	u8         reserved_at_8[0x18];
+
+	u8         syndrome[0x20];
+
+	u8         modify_header_id[0x20];
+
+	u8         reserved_at_60[0x20];
+};
+
+struct mlx5_ifc_alloc_modify_header_context_in_bits {
+	u8         opcode[0x10];
+	u8         reserved_at_10[0x10];
+
+	u8         reserved_at_20[0x10];
+	u8         op_mod[0x10];
+
+	u8         reserved_at_40[0x20];
+
+	u8         table_type[0x8];
+	u8         reserved_at_68[0x10];
+	u8         num_of_actions[0x8];
+
+	union mlx5_ifc_set_action_in_add_action_in_auto_bits actions[0];
+};
+
+struct mlx5_ifc_dealloc_modify_header_context_out_bits {
+	u8         status[0x8];
+	u8         reserved_at_8[0x18];
+
+	u8         syndrome[0x20];
+
+	u8         reserved_at_40[0x40];
+};
+
+struct mlx5_ifc_dealloc_modify_header_context_in_bits {
+	u8         opcode[0x10];
+	u8         reserved_at_10[0x10];
+
+	u8         reserved_at_20[0x10];
+	u8         op_mod[0x10];
+
+	u8         modify_header_id[0x20];
+
+	u8         reserved_at_60[0x20];
+};
+
 struct mlx5_ifc_query_dct_out_bits {
 	u8         status[0x8];
 	u8         reserved_at_8[0x18];

include/net/tc_act/tc_pedit.h

@@ -2,6 +2,7 @@
 #define __NET_TC_PED_H
 
 #include <net/act_api.h>
+#include <linux/tc_act/tc_pedit.h>
 
 struct tcf_pedit_key_ex {
 	enum pedit_header_type htype;
@@ -17,4 +18,48 @@ struct tcf_pedit {
 };
 #define to_pedit(a) ((struct tcf_pedit *)a)
 
+static inline bool is_tcf_pedit(const struct tc_action *a)
+{
+#ifdef CONFIG_NET_CLS_ACT
+	if (a->ops && a->ops->type == TCA_ACT_PEDIT)
+		return true;
+#endif
+	return false;
+}
+
+static inline int tcf_pedit_nkeys(const struct tc_action *a)
+{
+	return to_pedit(a)->tcfp_nkeys;
+}
+
+static inline u32 tcf_pedit_htype(const struct tc_action *a, int index)
+{
+	if (to_pedit(a)->tcfp_keys_ex)
+		return to_pedit(a)->tcfp_keys_ex[index].htype;
+
+	return TCA_PEDIT_KEY_EX_HDR_TYPE_NETWORK;
+}
+
+static inline u32 tcf_pedit_cmd(const struct tc_action *a, int index)
+{
+	if (to_pedit(a)->tcfp_keys_ex)
+		return to_pedit(a)->tcfp_keys_ex[index].cmd;
+
+	return __PEDIT_CMD_MAX;
+}
+
+static inline u32 tcf_pedit_mask(const struct tc_action *a, int index)
+{
+	return to_pedit(a)->tcfp_keys[index].mask;
+}
+
+static inline u32 tcf_pedit_val(const struct tc_action *a, int index)
+{
+	return to_pedit(a)->tcfp_keys[index].val;
+}
+
+static inline u32 tcf_pedit_offset(const struct tc_action *a, int index)
+{
+	return to_pedit(a)->tcfp_keys[index].off;
+}
 #endif /* __NET_TC_PED_H */