Merge branch 'mlxsw-Introduce-TC-Flower-offload-using-TCAM'

Jiri Pirko says:

====================
mlxsw: Introduce TC Flower offload using TCAM

This patchset introduces support for offloading TC cls_flower and actions
to Spectrum TCAM-base policy engine.

The patchset contains patches to allow work with flexible keys and actions
which are used in Spectrum TCAM.

It also contains in-driver infrastructure for offloading TC rules to TCAM HW.
The TCAM management code is simple and limited for now. It is going to be
extended as a follow-up work.

The last patch uses the previously introduced infra to allow to implement
cls_flower offloading. Initially, only limited set of match-keys and only
a drop and forward actions are supported.

As a dependency, this patchset introduces parman - priority array
area manager - as a library.

v1->v2:
- patch11:
  - use __set_bit and __test_and_clear_bit as suggested by DaveM
- patch16:
  - Added documentation to the API functions as suggested by Tom Herbert
- patch17:
  - use __set_bit and __clear_bit as suggested by DaveM
====================

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

MAINTAINERS

@@ -9382,6 +9382,14 @@ F:	drivers/video/fbdev/sti*
 F:	drivers/video/console/sti*
 F:	drivers/video/logo/logo_parisc*
 
+PARMAN
+M:	Jiri Pirko <jiri@mellanox.com>
+L:	netdev@vger.kernel.org
+S:	Supported
+F:	lib/parman.c
+F:	lib/test_parman.c
+F:	include/linux/parman.h
+
 PC87360 HARDWARE MONITORING DRIVER
 M:	Jim Cromie <jim.cromie@gmail.com>
 L:	linux-hwmon@vger.kernel.org

drivers/net/ethernet/mellanox/mlxsw/Kconfig

@@ -73,6 +73,7 @@ config MLXSW_SWITCHX2
 config MLXSW_SPECTRUM
 	tristate "Mellanox Technologies Spectrum support"
 	depends on MLXSW_CORE && MLXSW_PCI && NET_SWITCHDEV && VLAN_8021Q
+	select PARMAN
 	default m
 	---help---
 	  This driver supports Mellanox Technologies Spectrum Ethernet

drivers/net/ethernet/mellanox/mlxsw/Makefile

@@ -1,5 +1,6 @@
 obj-$(CONFIG_MLXSW_CORE)	+= mlxsw_core.o
-mlxsw_core-objs			:= core.o
+mlxsw_core-objs			:= core.o core_acl_flex_keys.o \
+				   core_acl_flex_actions.o
 mlxsw_core-$(CONFIG_MLXSW_CORE_HWMON) += core_hwmon.o
 mlxsw_core-$(CONFIG_MLXSW_CORE_THERMAL) += core_thermal.o
 obj-$(CONFIG_MLXSW_PCI)		+= mlxsw_pci.o
@@ -13,7 +14,8 @@ mlxsw_switchx2-objs		:= switchx2.o
 obj-$(CONFIG_MLXSW_SPECTRUM)	+= mlxsw_spectrum.o
 mlxsw_spectrum-objs		:= spectrum.o spectrum_buffers.o \
 				   spectrum_switchdev.o spectrum_router.o \
-				   spectrum_kvdl.o
+				   spectrum_kvdl.o spectrum_acl_tcam.o \
+				   spectrum_acl.o spectrum_flower.o
 mlxsw_spectrum-$(CONFIG_MLXSW_SPECTRUM_DCB)	+= spectrum_dcb.o
 obj-$(CONFIG_MLXSW_MINIMAL)	+= mlxsw_minimal.o
 mlxsw_minimal-objs		:= minimal.o

drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.c

@@ -0,0 +1,685 @@
+/*
+ * drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.c
+ * Copyright (c) 2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2017 Jiri Pirko <jiri@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/errno.h>
+#include <linux/rhashtable.h>
+#include <linux/list.h>
+
+#include "item.h"
+#include "core_acl_flex_actions.h"
+
+enum mlxsw_afa_set_type {
+	MLXSW_AFA_SET_TYPE_NEXT,
+	MLXSW_AFA_SET_TYPE_GOTO,
+};
+
+/* afa_set_type
+ * Type of the record at the end of the action set.
+ */
+MLXSW_ITEM32(afa, set, type, 0xA0, 28, 4);
+
+/* afa_set_next_action_set_ptr
+ * A pointer to the next action set in the KVD Centralized database.
+ */
+MLXSW_ITEM32(afa, set, next_action_set_ptr, 0xA4, 0, 24);
+
+/* afa_set_goto_g
+ * group - When set, the binding is of an ACL group. When cleared,
+ * the binding is of an ACL.
+ * Must be set to 1 for Spectrum.
+ */
+MLXSW_ITEM32(afa, set, goto_g, 0xA4, 29, 1);
+
+enum mlxsw_afa_set_goto_binding_cmd {
+	/* continue go the next binding point */
+	MLXSW_AFA_SET_GOTO_BINDING_CMD_NONE,
+	/* jump to the next binding point no return */
+	MLXSW_AFA_SET_GOTO_BINDING_CMD_JUMP,
+	/* terminate the acl binding */
+	MLXSW_AFA_SET_GOTO_BINDING_CMD_TERM = 4,
+};
+
+/* afa_set_goto_binding_cmd */
+MLXSW_ITEM32(afa, set, goto_binding_cmd, 0xA4, 24, 3);
+
+/* afa_set_goto_next_binding
+ * ACL/ACL group identifier. If the g bit is set, this field should hold
+ * the acl_group_id, else it should hold the acl_id.
+ */
+MLXSW_ITEM32(afa, set, goto_next_binding, 0xA4, 0, 16);
+
+/* afa_all_action_type
+ * Action Type.
+ */
+MLXSW_ITEM32(afa, all, action_type, 0x00, 24, 6);
+
+struct mlxsw_afa {
+	unsigned int max_acts_per_set;
+	const struct mlxsw_afa_ops *ops;
+	void *ops_priv;
+	struct rhashtable set_ht;
+	struct rhashtable fwd_entry_ht;
+};
+
+#define MLXSW_AFA_SET_LEN 0xA8
+
+struct mlxsw_afa_set_ht_key {
+	char enc_actions[MLXSW_AFA_SET_LEN]; /* Encoded set */
+	bool is_first;
+};
+
+/* Set structure holds one action set record. It contains up to three
+ * actions (depends on size of particular actions). The set is either
+ * put directly to a rule, or it is stored in KVD linear area.
+ * To prevent duplicate entries in KVD linear area, a hashtable is
+ * used to track sets that were previously inserted and may be shared.
+ */
+
+struct mlxsw_afa_set {
+	struct rhash_head ht_node;
+	struct mlxsw_afa_set_ht_key ht_key;
+	u32 kvdl_index;
+	bool shared; /* Inserted in hashtable (doesn't mean that
+		      * kvdl_index is valid).
+		      */
+	unsigned int ref_count;
+	struct mlxsw_afa_set *next; /* Pointer to the next set. */
+	struct mlxsw_afa_set *prev; /* Pointer to the previous set,
+				     * note that set may have multiple
+				     * sets from multiple blocks
+				     * pointing at it. This is only
+				     * usable until commit.
+				     */
+};
+
+static const struct rhashtable_params mlxsw_afa_set_ht_params = {
+	.key_len = sizeof(struct mlxsw_afa_set_ht_key),
+	.key_offset = offsetof(struct mlxsw_afa_set, ht_key),
+	.head_offset = offsetof(struct mlxsw_afa_set, ht_node),
+	.automatic_shrinking = true,
+};
+
+struct mlxsw_afa_fwd_entry_ht_key {
+	u8 local_port;
+};
+
+struct mlxsw_afa_fwd_entry {
+	struct rhash_head ht_node;
+	struct mlxsw_afa_fwd_entry_ht_key ht_key;
+	u32 kvdl_index;
+	unsigned int ref_count;
+};
+
+static const struct rhashtable_params mlxsw_afa_fwd_entry_ht_params = {
+	.key_len = sizeof(struct mlxsw_afa_fwd_entry_ht_key),
+	.key_offset = offsetof(struct mlxsw_afa_fwd_entry, ht_key),
+	.head_offset = offsetof(struct mlxsw_afa_fwd_entry, ht_node),
+	.automatic_shrinking = true,
+};
+
+struct mlxsw_afa *mlxsw_afa_create(unsigned int max_acts_per_set,
+				   const struct mlxsw_afa_ops *ops,
+				   void *ops_priv)
+{
+	struct mlxsw_afa *mlxsw_afa;
+	int err;
+
+	mlxsw_afa = kzalloc(sizeof(*mlxsw_afa), GFP_KERNEL);
+	if (!mlxsw_afa)
+		return ERR_PTR(-ENOMEM);
+	err = rhashtable_init(&mlxsw_afa->set_ht, &mlxsw_afa_set_ht_params);
+	if (err)
+		goto err_set_rhashtable_init;
+	err = rhashtable_init(&mlxsw_afa->fwd_entry_ht,
+			      &mlxsw_afa_fwd_entry_ht_params);
+	if (err)
+		goto err_fwd_entry_rhashtable_init;
+	mlxsw_afa->max_acts_per_set = max_acts_per_set;
+	mlxsw_afa->ops = ops;
+	mlxsw_afa->ops_priv = ops_priv;
+	return mlxsw_afa;
+
+err_fwd_entry_rhashtable_init:
+	rhashtable_destroy(&mlxsw_afa->set_ht);
+err_set_rhashtable_init:
+	kfree(mlxsw_afa);
+	return ERR_PTR(err);
+}
+EXPORT_SYMBOL(mlxsw_afa_create);
+
+void mlxsw_afa_destroy(struct mlxsw_afa *mlxsw_afa)
+{
+	rhashtable_destroy(&mlxsw_afa->fwd_entry_ht);
+	rhashtable_destroy(&mlxsw_afa->set_ht);
+	kfree(mlxsw_afa);
+}
+EXPORT_SYMBOL(mlxsw_afa_destroy);
+
+static void mlxsw_afa_set_goto_set(struct mlxsw_afa_set *set,
+				   enum mlxsw_afa_set_goto_binding_cmd cmd,
+				   u16 group_id)
+{
+	char *actions = set->ht_key.enc_actions;
+
+	mlxsw_afa_set_type_set(actions, MLXSW_AFA_SET_TYPE_GOTO);
+	mlxsw_afa_set_goto_g_set(actions, true);
+	mlxsw_afa_set_goto_binding_cmd_set(actions, cmd);
+	mlxsw_afa_set_goto_next_binding_set(actions, group_id);
+}
+
+static void mlxsw_afa_set_next_set(struct mlxsw_afa_set *set,
+				   u32 next_set_kvdl_index)
+{
+	char *actions = set->ht_key.enc_actions;
+
+	mlxsw_afa_set_type_set(actions, MLXSW_AFA_SET_TYPE_NEXT);
+	mlxsw_afa_set_next_action_set_ptr_set(actions, next_set_kvdl_index);
+}
+
+static struct mlxsw_afa_set *mlxsw_afa_set_create(bool is_first)
+{
+	struct mlxsw_afa_set *set;
+
+	set = kzalloc(sizeof(*set), GFP_KERNEL);
+	if (!set)
+		return NULL;
+	/* Need to initialize the set to pass by default */
+	mlxsw_afa_set_goto_set(set, MLXSW_AFA_SET_GOTO_BINDING_CMD_TERM, 0);
+	set->ht_key.is_first = is_first;
+	set->ref_count = 1;
+	return set;
+}
+
+static void mlxsw_afa_set_destroy(struct mlxsw_afa_set *set)
+{
+	kfree(set);
+}
+
+static int mlxsw_afa_set_share(struct mlxsw_afa *mlxsw_afa,
+			       struct mlxsw_afa_set *set)
+{
+	int err;
+
+	err = rhashtable_insert_fast(&mlxsw_afa->set_ht, &set->ht_node,
+				     mlxsw_afa_set_ht_params);
+	if (err)
+		return err;
+	err = mlxsw_afa->ops->kvdl_set_add(mlxsw_afa->ops_priv,
+					   &set->kvdl_index,
+					   set->ht_key.enc_actions,
+					   set->ht_key.is_first);
+	if (err)
+		goto err_kvdl_set_add;
+	set->shared = true;
+	set->prev = NULL;
+	return 0;
+
+err_kvdl_set_add:
+	rhashtable_remove_fast(&mlxsw_afa->set_ht, &set->ht_node,
+			       mlxsw_afa_set_ht_params);
+	return err;
+}
+
+static void mlxsw_afa_set_unshare(struct mlxsw_afa *mlxsw_afa,
+				  struct mlxsw_afa_set *set)
+{
+	mlxsw_afa->ops->kvdl_set_del(mlxsw_afa->ops_priv,
+				     set->kvdl_index,
+				     set->ht_key.is_first);
+	rhashtable_remove_fast(&mlxsw_afa->set_ht, &set->ht_node,
+			       mlxsw_afa_set_ht_params);
+	set->shared = false;
+}
+
+static void mlxsw_afa_set_put(struct mlxsw_afa *mlxsw_afa,
+			      struct mlxsw_afa_set *set)
+{
+	if (--set->ref_count)
+		return;
+	if (set->shared)
+		mlxsw_afa_set_unshare(mlxsw_afa, set);
+	mlxsw_afa_set_destroy(set);
+}
+
+static struct mlxsw_afa_set *mlxsw_afa_set_get(struct mlxsw_afa *mlxsw_afa,
+					       struct mlxsw_afa_set *orig_set)
+{
+	struct mlxsw_afa_set *set;
+	int err;
+
+	/* There is a hashtable of sets maintained. If a set with the exact
+	 * same encoding exists, we reuse it. Otherwise, the current set
+	 * is shared by making it available to others using the hash table.
+	 */
+	set = rhashtable_lookup_fast(&mlxsw_afa->set_ht, &orig_set->ht_key,
+				     mlxsw_afa_set_ht_params);
+	if (set) {
+		set->ref_count++;
+		mlxsw_afa_set_put(mlxsw_afa, orig_set);
+	} else {
+		set = orig_set;
+		err = mlxsw_afa_set_share(mlxsw_afa, set);
+		if (err)
+			return ERR_PTR(err);
+	}
+	return set;
+}
+
+/* Block structure holds a list of action sets. One action block
+ * represents one chain of actions executed upon match of a rule.
+ */
+
+struct mlxsw_afa_block {
+	struct mlxsw_afa *afa;
+	bool finished;
+	struct mlxsw_afa_set *first_set;
+	struct mlxsw_afa_set *cur_set;
+	unsigned int cur_act_index; /* In current set. */
+	struct list_head fwd_entry_ref_list;
+};
+
+struct mlxsw_afa_block *mlxsw_afa_block_create(struct mlxsw_afa *mlxsw_afa)
+{
+	struct mlxsw_afa_block *block;
+
+	block = kzalloc(sizeof(*block), GFP_KERNEL);
+	if (!block)
+		return NULL;
+	INIT_LIST_HEAD(&block->fwd_entry_ref_list);
+	block->afa = mlxsw_afa;
+
+	/* At least one action set is always present, so just create it here */
+	block->first_set = mlxsw_afa_set_create(true);
+	if (!block->first_set)
+		goto err_first_set_create;
+	block->cur_set = block->first_set;
+	return block;
+
+err_first_set_create:
+	kfree(block);
+	return NULL;
+}
+EXPORT_SYMBOL(mlxsw_afa_block_create);
+
+static void mlxsw_afa_fwd_entry_refs_destroy(struct mlxsw_afa_block *block);
+
+void mlxsw_afa_block_destroy(struct mlxsw_afa_block *block)
+{
+	struct mlxsw_afa_set *set = block->first_set;
+	struct mlxsw_afa_set *next_set;
+
+	do {
+		next_set = set->next;
+		mlxsw_afa_set_put(block->afa, set);
+		set = next_set;
+	} while (set);
+	mlxsw_afa_fwd_entry_refs_destroy(block);
+	kfree(block);
+}
+EXPORT_SYMBOL(mlxsw_afa_block_destroy);
+
+int mlxsw_afa_block_commit(struct mlxsw_afa_block *block)
+{
+	struct mlxsw_afa_set *set = block->cur_set;
+	struct mlxsw_afa_set *prev_set;
+	int err;
+
+	block->cur_set = NULL;
+
+	/* Go over all linked sets starting from last
+	 * and try to find existing set in the hash table.
+	 * In case it is not there, assign a KVD linear index
+	 * and insert it.
+	 */
+	do {
+		prev_set = set->prev;
+		set = mlxsw_afa_set_get(block->afa, set);
+		if (IS_ERR(set)) {
+			err = PTR_ERR(set);
+			goto rollback;
+		}
+		if (prev_set) {
+			prev_set->next = set;
+			mlxsw_afa_set_next_set(prev_set, set->kvdl_index);
+			set = prev_set;
+		}
+	} while (prev_set);
+
+	block->first_set = set;
+	block->finished = true;
+	return 0;
+
+rollback:
+	while ((set = set->next))
+		mlxsw_afa_set_put(block->afa, set);
+	return err;
+}
+EXPORT_SYMBOL(mlxsw_afa_block_commit);
+
+char *mlxsw_afa_block_first_set(struct mlxsw_afa_block *block)
+{
+	return block->first_set->ht_key.enc_actions;
+}
+EXPORT_SYMBOL(mlxsw_afa_block_first_set);
+
+u32 mlxsw_afa_block_first_set_kvdl_index(struct mlxsw_afa_block *block)
+{
+	return block->first_set->kvdl_index;
+}
+EXPORT_SYMBOL(mlxsw_afa_block_first_set_kvdl_index);
+
+void mlxsw_afa_block_continue(struct mlxsw_afa_block *block)
+{
+	if (WARN_ON(block->finished))
+		return;
+	mlxsw_afa_set_goto_set(block->cur_set,
+			       MLXSW_AFA_SET_GOTO_BINDING_CMD_NONE, 0);
+	block->finished = true;
+}
+EXPORT_SYMBOL(mlxsw_afa_block_continue);
+
+void mlxsw_afa_block_jump(struct mlxsw_afa_block *block, u16 group_id)
+{
+	if (WARN_ON(block->finished))
+		return;
+	mlxsw_afa_set_goto_set(block->cur_set,
+			       MLXSW_AFA_SET_GOTO_BINDING_CMD_JUMP, group_id);
+	block->finished = true;
+}
+EXPORT_SYMBOL(mlxsw_afa_block_jump);
+
+static struct mlxsw_afa_fwd_entry *
+mlxsw_afa_fwd_entry_create(struct mlxsw_afa *mlxsw_afa, u8 local_port)
+{
+	struct mlxsw_afa_fwd_entry *fwd_entry;
+	int err;
+
+	fwd_entry = kzalloc(sizeof(*fwd_entry), GFP_KERNEL);
+	if (!fwd_entry)
+		return ERR_PTR(-ENOMEM);
+	fwd_entry->ht_key.local_port = local_port;
+	fwd_entry->ref_count = 1;
+
+	err = rhashtable_insert_fast(&mlxsw_afa->fwd_entry_ht,
+				     &fwd_entry->ht_node,
+				     mlxsw_afa_fwd_entry_ht_params);
+	if (err)
+		goto err_rhashtable_insert;
+
+	err = mlxsw_afa->ops->kvdl_fwd_entry_add(mlxsw_afa->ops_priv,
+						 &fwd_entry->kvdl_index,
+						 local_port);
+	if (err)
+		goto err_kvdl_fwd_entry_add;
+	return fwd_entry;
+
+err_kvdl_fwd_entry_add:
+	rhashtable_remove_fast(&mlxsw_afa->fwd_entry_ht, &fwd_entry->ht_node,
+			       mlxsw_afa_fwd_entry_ht_params);
+err_rhashtable_insert:
+	kfree(fwd_entry);
+	return ERR_PTR(err);
+}
+
+static void mlxsw_afa_fwd_entry_destroy(struct mlxsw_afa *mlxsw_afa,
+					struct mlxsw_afa_fwd_entry *fwd_entry)
+{
+	mlxsw_afa->ops->kvdl_fwd_entry_del(mlxsw_afa->ops_priv,
+					   fwd_entry->kvdl_index);
+	rhashtable_remove_fast(&mlxsw_afa->fwd_entry_ht, &fwd_entry->ht_node,
+			       mlxsw_afa_fwd_entry_ht_params);
+	kfree(fwd_entry);
+}
+
+static struct mlxsw_afa_fwd_entry *
+mlxsw_afa_fwd_entry_get(struct mlxsw_afa *mlxsw_afa, u8 local_port)
+{
+	struct mlxsw_afa_fwd_entry_ht_key ht_key = {0};
+	struct mlxsw_afa_fwd_entry *fwd_entry;
+
+	ht_key.local_port = local_port;
+	fwd_entry = rhashtable_lookup_fast(&mlxsw_afa->fwd_entry_ht, &ht_key,
+					   mlxsw_afa_fwd_entry_ht_params);
+	if (fwd_entry) {
+		fwd_entry->ref_count++;
+		return fwd_entry;
+	}
+	return mlxsw_afa_fwd_entry_create(mlxsw_afa, local_port);
+}
+
+static void mlxsw_afa_fwd_entry_put(struct mlxsw_afa *mlxsw_afa,
+				    struct mlxsw_afa_fwd_entry *fwd_entry)
+{
+	if (--fwd_entry->ref_count)
+		return;
+	mlxsw_afa_fwd_entry_destroy(mlxsw_afa, fwd_entry);
+}
+
+struct mlxsw_afa_fwd_entry_ref {
+	struct list_head list;
+	struct mlxsw_afa_fwd_entry *fwd_entry;
+};
+
+static struct mlxsw_afa_fwd_entry_ref *
+mlxsw_afa_fwd_entry_ref_create(struct mlxsw_afa_block *block, u8 local_port)
+{
+	struct mlxsw_afa_fwd_entry_ref *fwd_entry_ref;
+	struct mlxsw_afa_fwd_entry *fwd_entry;
+	int err;
+
+	fwd_entry_ref = kzalloc(sizeof(*fwd_entry_ref), GFP_KERNEL);
+	if (!fwd_entry_ref)
+		return ERR_PTR(-ENOMEM);
+	fwd_entry = mlxsw_afa_fwd_entry_get(block->afa, local_port);
+	if (IS_ERR(fwd_entry)) {
+		err = PTR_ERR(fwd_entry);
+		goto err_fwd_entry_get;
+	}
+	fwd_entry_ref->fwd_entry = fwd_entry;
+	list_add(&fwd_entry_ref->list, &block->fwd_entry_ref_list);
+	return fwd_entry_ref;
+
+err_fwd_entry_get:
+	kfree(fwd_entry_ref);
+	return ERR_PTR(err);
+}
+
+static void
+mlxsw_afa_fwd_entry_ref_destroy(struct mlxsw_afa_block *block,
+				struct mlxsw_afa_fwd_entry_ref *fwd_entry_ref)
+{
+	list_del(&fwd_entry_ref->list);
+	mlxsw_afa_fwd_entry_put(block->afa, fwd_entry_ref->fwd_entry);
+	kfree(fwd_entry_ref);
+}
+
+static void mlxsw_afa_fwd_entry_refs_destroy(struct mlxsw_afa_block *block)
+{
+	struct mlxsw_afa_fwd_entry_ref *fwd_entry_ref;
+	struct mlxsw_afa_fwd_entry_ref *tmp;
+
+	list_for_each_entry_safe(fwd_entry_ref, tmp,
+				 &block->fwd_entry_ref_list, list)
+		mlxsw_afa_fwd_entry_ref_destroy(block, fwd_entry_ref);
+}
+
+#define MLXSW_AFA_ONE_ACTION_LEN 32
+#define MLXSW_AFA_PAYLOAD_OFFSET 4
+
+static char *mlxsw_afa_block_append_action(struct mlxsw_afa_block *block,
+					   u8 action_code, u8 action_size)
+{
+	char *oneact;
+	char *actions;
+
+	if (WARN_ON(block->finished))
+		return NULL;
+	if (block->cur_act_index + action_size >
+	    block->afa->max_acts_per_set) {
+		struct mlxsw_afa_set *set;
+
+		/* The appended action won't fit into the current action set,
+		 * so create a new set.
+		 */
+		set = mlxsw_afa_set_create(false);
+		if (!set)
+			return NULL;
+		set->prev = block->cur_set;
+		block->cur_act_index = 0;
+		block->cur_set->next = set;
+		block->cur_set = set;
+	}
+
+	actions = block->cur_set->ht_key.enc_actions;
+	oneact = actions + block->cur_act_index * MLXSW_AFA_ONE_ACTION_LEN;
+	block->cur_act_index += action_size;
+	mlxsw_afa_all_action_type_set(oneact, action_code);
+	return oneact + MLXSW_AFA_PAYLOAD_OFFSET;
+}
+
+/* Trap / Discard Action
+ * ---------------------
+ * The Trap / Discard action enables trapping / mirroring packets to the CPU
+ * as well as discarding packets.
+ * The ACL Trap / Discard separates the forward/discard control from CPU
+ * trap control. In addition, the Trap / Discard action enables activating
+ * SPAN (port mirroring).
+ */
+
+#define MLXSW_AFA_TRAPDISC_CODE 0x03
+#define MLXSW_AFA_TRAPDISC_SIZE 1
+
+enum mlxsw_afa_trapdisc_forward_action {
+	MLXSW_AFA_TRAPDISC_FORWARD_ACTION_DISCARD = 3,
+};
+
+/* afa_trapdisc_forward_action
+ * Forward Action.
+ */
+MLXSW_ITEM32(afa, trapdisc, forward_action, 0x00, 0, 4);
+
+static inline void
+mlxsw_afa_trapdisc_pack(char *payload,
+			enum mlxsw_afa_trapdisc_forward_action forward_action)
+{
+	mlxsw_afa_trapdisc_forward_action_set(payload, forward_action);
+}
+
+int mlxsw_afa_block_append_drop(struct mlxsw_afa_block *block)
+{
+	char *act = mlxsw_afa_block_append_action(block,
+						  MLXSW_AFA_TRAPDISC_CODE,
+						  MLXSW_AFA_TRAPDISC_SIZE);
+
+	if (!act)
+		return -ENOBUFS;
+	mlxsw_afa_trapdisc_pack(act, MLXSW_AFA_TRAPDISC_FORWARD_ACTION_DISCARD);
+	return 0;
+}
+EXPORT_SYMBOL(mlxsw_afa_block_append_drop);
+
+/* Forwarding Action
+ * -----------------
+ * Forwarding Action can be used to implement Policy Based Switching (PBS)
+ * as well as OpenFlow related "Output" action.
+ */
+
+#define MLXSW_AFA_FORWARD_CODE 0x07
+#define MLXSW_AFA_FORWARD_SIZE 1
+
+enum mlxsw_afa_forward_type {
+	/* PBS, Policy Based Switching */
+	MLXSW_AFA_FORWARD_TYPE_PBS,
+	/* Output, OpenFlow output type */
+	MLXSW_AFA_FORWARD_TYPE_OUTPUT,
+};
+
+/* afa_forward_type */
+MLXSW_ITEM32(afa, forward, type, 0x00, 24, 2);
+
+/* afa_forward_pbs_ptr
+ * A pointer to the PBS entry configured by PPBS register.
+ * Reserved when in_port is set.
+ */
+MLXSW_ITEM32(afa, forward, pbs_ptr, 0x08, 0, 24);
+
+/* afa_forward_in_port
+ * Packet is forwarded back to the ingress port.
+ */
+MLXSW_ITEM32(afa, forward, in_port, 0x0C, 0, 1);
+
+static inline void
+mlxsw_afa_forward_pack(char *payload, enum mlxsw_afa_forward_type type,
+		       u32 pbs_ptr, bool in_port)
+{
+	mlxsw_afa_forward_type_set(payload, type);
+	mlxsw_afa_forward_pbs_ptr_set(payload, pbs_ptr);
+	mlxsw_afa_forward_in_port_set(payload, in_port);
+}
+
+int mlxsw_afa_block_append_fwd(struct mlxsw_afa_block *block,
+			       u8 local_port, bool in_port)
+{
+	struct mlxsw_afa_fwd_entry_ref *fwd_entry_ref;
+	u32 kvdl_index = 0;
+	char *act;
+	int err;
+
+	if (!in_port) {
+		fwd_entry_ref = mlxsw_afa_fwd_entry_ref_create(block,
+							       local_port);
+		if (IS_ERR(fwd_entry_ref))
+			return PTR_ERR(fwd_entry_ref);
+		kvdl_index = fwd_entry_ref->fwd_entry->kvdl_index;
+	}
+
+	act = mlxsw_afa_block_append_action(block, MLXSW_AFA_FORWARD_CODE,
+					    MLXSW_AFA_FORWARD_SIZE);
+	if (!act) {
+		err = -ENOBUFS;
+		goto err_append_action;
+	}
+	mlxsw_afa_forward_pack(act, MLXSW_AFA_FORWARD_TYPE_OUTPUT,
+			       kvdl_index, in_port);
+	return 0;
+
+err_append_action:
+	if (!in_port)
+		mlxsw_afa_fwd_entry_ref_destroy(block, fwd_entry_ref);
+	return err;
+}
+EXPORT_SYMBOL(mlxsw_afa_block_append_fwd);

drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.h

@@ -0,0 +1,66 @@
+/*
+ * drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.h
+ * Copyright (c) 2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2017 Jiri Pirko <jiri@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _MLXSW_CORE_ACL_FLEX_ACTIONS_H
+#define _MLXSW_CORE_ACL_FLEX_ACTIONS_H
+
+#include <linux/types.h>
+
+struct mlxsw_afa;
+struct mlxsw_afa_block;
+
+struct mlxsw_afa_ops {
+	int (*kvdl_set_add)(void *priv, u32 *p_kvdl_index,
+			    char *enc_actions, bool is_first);
+	void (*kvdl_set_del)(void *priv, u32 kvdl_index, bool is_first);
+	int (*kvdl_fwd_entry_add)(void *priv, u32 *p_kvdl_index, u8 local_port);
+	void (*kvdl_fwd_entry_del)(void *priv, u32 kvdl_index);
+};
+
+struct mlxsw_afa *mlxsw_afa_create(unsigned int max_acts_per_set,
+				   const struct mlxsw_afa_ops *ops,
+				   void *ops_priv);
+void mlxsw_afa_destroy(struct mlxsw_afa *mlxsw_afa);
+struct mlxsw_afa_block *mlxsw_afa_block_create(struct mlxsw_afa *mlxsw_afa);
+void mlxsw_afa_block_destroy(struct mlxsw_afa_block *block);
+int mlxsw_afa_block_commit(struct mlxsw_afa_block *block);
+char *mlxsw_afa_block_first_set(struct mlxsw_afa_block *block);
+u32 mlxsw_afa_block_first_set_kvdl_index(struct mlxsw_afa_block *block);
+void mlxsw_afa_block_continue(struct mlxsw_afa_block *block);
+void mlxsw_afa_block_jump(struct mlxsw_afa_block *block, u16 group_id);
+int mlxsw_afa_block_append_drop(struct mlxsw_afa_block *block);
+int mlxsw_afa_block_append_fwd(struct mlxsw_afa_block *block,
+			       u8 local_port, bool in_port);
+
+#endif

drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.c

@@ -0,0 +1,475 @@
+/*
+ * drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.c
+ * Copyright (c) 2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2017 Jiri Pirko <jiri@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/list.h>
+#include <linux/errno.h>
+
+#include "item.h"
+#include "core_acl_flex_keys.h"
+
+struct mlxsw_afk {
+	struct list_head key_info_list;
+	unsigned int max_blocks;
+	const struct mlxsw_afk_block *blocks;
+	unsigned int blocks_count;
+};
+
+static bool mlxsw_afk_blocks_check(struct mlxsw_afk *mlxsw_afk)
+{
+	int i;
+	int j;
+
+	for (i = 0; i < mlxsw_afk->blocks_count; i++) {
+		const struct mlxsw_afk_block *block = &mlxsw_afk->blocks[i];
+
+		for (j = 0; j < block->instances_count; j++) {
+			struct mlxsw_afk_element_inst *elinst;
+
+			elinst = &block->instances[j];
+			if (elinst->type != elinst->info->type ||
+			    elinst->item.size.bits !=
+			    elinst->info->item.size.bits)
+				return false;
+		}
+	}
+	return true;
+}
+
+struct mlxsw_afk *mlxsw_afk_create(unsigned int max_blocks,
+				   const struct mlxsw_afk_block *blocks,
+				   unsigned int blocks_count)
+{
+	struct mlxsw_afk *mlxsw_afk;
+
+	mlxsw_afk = kzalloc(sizeof(*mlxsw_afk), GFP_KERNEL);
+	if (!mlxsw_afk)
+		return NULL;
+	INIT_LIST_HEAD(&mlxsw_afk->key_info_list);
+	mlxsw_afk->max_blocks = max_blocks;
+	mlxsw_afk->blocks = blocks;
+	mlxsw_afk->blocks_count = blocks_count;
+	WARN_ON(!mlxsw_afk_blocks_check(mlxsw_afk));
+	return mlxsw_afk;
+}
+EXPORT_SYMBOL(mlxsw_afk_create);
+
+void mlxsw_afk_destroy(struct mlxsw_afk *mlxsw_afk)
+{
+	WARN_ON(!list_empty(&mlxsw_afk->key_info_list));
+	kfree(mlxsw_afk);
+}
+EXPORT_SYMBOL(mlxsw_afk_destroy);
+
+struct mlxsw_afk_key_info {
+	struct list_head list;
+	unsigned int ref_count;
+	unsigned int blocks_count;
+	int element_to_block[MLXSW_AFK_ELEMENT_MAX]; /* index is element, value
+						      * is index inside "blocks"
+						      */
+	struct mlxsw_afk_element_usage elusage;
+	const struct mlxsw_afk_block *blocks[0];
+};
+
+static bool
+mlxsw_afk_key_info_elements_eq(struct mlxsw_afk_key_info *key_info,
+			       struct mlxsw_afk_element_usage *elusage)
+{
+	return memcmp(&key_info->elusage, elusage, sizeof(*elusage)) == 0;
+}
+
+static struct mlxsw_afk_key_info *
+mlxsw_afk_key_info_find(struct mlxsw_afk *mlxsw_afk,
+			struct mlxsw_afk_element_usage *elusage)
+{
+	struct mlxsw_afk_key_info *key_info;
+
+	list_for_each_entry(key_info, &mlxsw_afk->key_info_list, list) {
+		if (mlxsw_afk_key_info_elements_eq(key_info, elusage))
+			return key_info;
+	}
+	return NULL;
+}
+
+struct mlxsw_afk_picker {
+	struct {
+		DECLARE_BITMAP(element, MLXSW_AFK_ELEMENT_MAX);
+		unsigned int total;
+	} hits[0];
+};
+
+static void mlxsw_afk_picker_count_hits(struct mlxsw_afk *mlxsw_afk,
+					struct mlxsw_afk_picker *picker,
+					enum mlxsw_afk_element element)
+{
+	int i;
+	int j;
+
+	for (i = 0; i < mlxsw_afk->blocks_count; i++) {
+		const struct mlxsw_afk_block *block = &mlxsw_afk->blocks[i];
+
+		for (j = 0; j < block->instances_count; j++) {
+			struct mlxsw_afk_element_inst *elinst;
+
+			elinst = &block->instances[j];
+			if (elinst->info->element == element) {
+				__set_bit(element, picker->hits[i].element);
+				picker->hits[i].total++;
+			}
+		}
+	}
+}
+
+static void mlxsw_afk_picker_subtract_hits(struct mlxsw_afk *mlxsw_afk,
+					   struct mlxsw_afk_picker *picker,
+					   int block_index)
+{
+	DECLARE_BITMAP(hits_element, MLXSW_AFK_ELEMENT_MAX);
+	int i;
+	int j;
+
+	memcpy(&hits_element, &picker->hits[block_index].element,
+	       sizeof(hits_element));
+
+	for (i = 0; i < mlxsw_afk->blocks_count; i++) {
+		for_each_set_bit(j, hits_element, MLXSW_AFK_ELEMENT_MAX) {
+			if (__test_and_clear_bit(j, picker->hits[i].element))
+				picker->hits[i].total--;
+		}
+	}
+}
+
+static int mlxsw_afk_picker_most_hits_get(struct mlxsw_afk *mlxsw_afk,
+					  struct mlxsw_afk_picker *picker)
+{
+	int most_index = -EINVAL; /* Should never happen to return this */
+	int most_hits = 0;
+	int i;
+
+	for (i = 0; i < mlxsw_afk->blocks_count; i++) {
+		if (picker->hits[i].total > most_hits) {
+			most_hits = picker->hits[i].total;
+			most_index = i;
+		}
+	}
+	return most_index;
+}
+
+static int mlxsw_afk_picker_key_info_add(struct mlxsw_afk *mlxsw_afk,
+					 struct mlxsw_afk_picker *picker,
+					 int block_index,
+					 struct mlxsw_afk_key_info *key_info)
+{
+	enum mlxsw_afk_element element;
+
+	if (key_info->blocks_count == mlxsw_afk->max_blocks)
+		return -EINVAL;
+
+	for_each_set_bit(element, picker->hits[block_index].element,
+			 MLXSW_AFK_ELEMENT_MAX) {
+		key_info->element_to_block[element] = key_info->blocks_count;
+		mlxsw_afk_element_usage_add(&key_info->elusage, element);
+	}
+
+	key_info->blocks[key_info->blocks_count] =
+					&mlxsw_afk->blocks[block_index];
+	key_info->blocks_count++;
+	return 0;
+}
+
+static int mlxsw_afk_picker(struct mlxsw_afk *mlxsw_afk,
+			    struct mlxsw_afk_key_info *key_info,
+			    struct mlxsw_afk_element_usage *elusage)
+{
+	struct mlxsw_afk_picker *picker;
+	enum mlxsw_afk_element element;
+	size_t alloc_size;
+	int err;
+
+	alloc_size = sizeof(picker->hits[0]) * mlxsw_afk->blocks_count;
+	picker = kzalloc(alloc_size, GFP_KERNEL);
+	if (!picker)
+		return -ENOMEM;
+
+	/* Since the same elements could be present in multiple blocks,
+	 * we must find out optimal block list in order to make the
+	 * block count as low as possible.
+	 *
+	 * First, we count hits. We go over all available blocks and count
+	 * how many of requested elements are covered by each.
+	 *
+	 * Then in loop, we find block with most hits and add it to
+	 * output key_info. Then we have to subtract this block hits so
+	 * the next iteration will find most suitable block for
+	 * the rest of requested elements.
+	 */
+
+	mlxsw_afk_element_usage_for_each(element, elusage)
+		mlxsw_afk_picker_count_hits(mlxsw_afk, picker, element);
+
+	do {
+		int block_index;
+
+		block_index = mlxsw_afk_picker_most_hits_get(mlxsw_afk, picker);
+		if (block_index < 0) {
+			err = block_index;
+			goto out;
+		}
+		err = mlxsw_afk_picker_key_info_add(mlxsw_afk, picker,
+						    block_index, key_info);
+		if (err)
+			goto out;
+		mlxsw_afk_picker_subtract_hits(mlxsw_afk, picker, block_index);
+	} while (!mlxsw_afk_key_info_elements_eq(key_info, elusage));
+
+	err = 0;
+out:
+	kfree(picker);
+	return err;
+}
+
+static struct mlxsw_afk_key_info *
+mlxsw_afk_key_info_create(struct mlxsw_afk *mlxsw_afk,
+			  struct mlxsw_afk_element_usage *elusage)
+{
+	struct mlxsw_afk_key_info *key_info;
+	size_t alloc_size;
+	int err;
+
+	alloc_size = sizeof(*key_info) +
+		     sizeof(key_info->blocks[0]) * mlxsw_afk->max_blocks;
+	key_info = kzalloc(alloc_size, GFP_KERNEL);
+	if (!key_info)
+		return ERR_PTR(-ENOMEM);
+	err = mlxsw_afk_picker(mlxsw_afk, key_info, elusage);
+	if (err)
+		goto err_picker;
+	list_add(&key_info->list, &mlxsw_afk->key_info_list);
+	key_info->ref_count = 1;
+	return key_info;
+
+err_picker:
+	kfree(key_info);
+	return ERR_PTR(err);
+}
+
+static void mlxsw_afk_key_info_destroy(struct mlxsw_afk_key_info *key_info)
+{
+	list_del(&key_info->list);
+	kfree(key_info);
+}
+
+struct mlxsw_afk_key_info *
+mlxsw_afk_key_info_get(struct mlxsw_afk *mlxsw_afk,
+		       struct mlxsw_afk_element_usage *elusage)
+{
+	struct mlxsw_afk_key_info *key_info;
+
+	key_info = mlxsw_afk_key_info_find(mlxsw_afk, elusage);
+	if (key_info) {
+		key_info->ref_count++;
+		return key_info;
+	}
+	return mlxsw_afk_key_info_create(mlxsw_afk, elusage);
+}
+EXPORT_SYMBOL(mlxsw_afk_key_info_get);
+
+void mlxsw_afk_key_info_put(struct mlxsw_afk_key_info *key_info)
+{
+	if (--key_info->ref_count)
+		return;
+	mlxsw_afk_key_info_destroy(key_info);
+}
+EXPORT_SYMBOL(mlxsw_afk_key_info_put);
+
+bool mlxsw_afk_key_info_subset(struct mlxsw_afk_key_info *key_info,
+			       struct mlxsw_afk_element_usage *elusage)
+{
+	return mlxsw_afk_element_usage_subset(elusage, &key_info->elusage);
+}
+EXPORT_SYMBOL(mlxsw_afk_key_info_subset);
+
+static const struct mlxsw_afk_element_inst *
+mlxsw_afk_block_elinst_get(const struct mlxsw_afk_block *block,
+			   enum mlxsw_afk_element element)
+{
+	int i;
+
+	for (i = 0; i < block->instances_count; i++) {
+		struct mlxsw_afk_element_inst *elinst;
+
+		elinst = &block->instances[i];
+		if (elinst->info->element == element)
+			return elinst;
+	}
+	return NULL;
+}
+
+static const struct mlxsw_afk_element_inst *
+mlxsw_afk_key_info_elinst_get(struct mlxsw_afk_key_info *key_info,
+			      enum mlxsw_afk_element element,
+			      int *p_block_index)
+{
+	const struct mlxsw_afk_element_inst *elinst;
+	const struct mlxsw_afk_block *block;
+	int block_index;
+
+	if (WARN_ON(!test_bit(element, key_info->elusage.usage)))
+		return NULL;
+	block_index = key_info->element_to_block[element];
+	block = key_info->blocks[block_index];
+
+	elinst = mlxsw_afk_block_elinst_get(block, element);
+	if (WARN_ON(!elinst))
+		return NULL;
+
+	*p_block_index = block_index;
+	return elinst;
+}
+
+u16
+mlxsw_afk_key_info_block_encoding_get(const struct mlxsw_afk_key_info *key_info,
+				      int block_index)
+{
+	return key_info->blocks[block_index]->encoding;
+}
+EXPORT_SYMBOL(mlxsw_afk_key_info_block_encoding_get);
+
+unsigned int
+mlxsw_afk_key_info_blocks_count_get(const struct mlxsw_afk_key_info *key_info)
+{
+	return key_info->blocks_count;
+}
+EXPORT_SYMBOL(mlxsw_afk_key_info_blocks_count_get);
+
+void mlxsw_afk_values_add_u32(struct mlxsw_afk_element_values *values,
+			      enum mlxsw_afk_element element,
+			      u32 key_value, u32 mask_value)
+{
+	const struct mlxsw_afk_element_info *elinfo =
+				&mlxsw_afk_element_infos[element];
+	const struct mlxsw_item *storage_item = &elinfo->item;
+
+	if (!mask_value)
+		return;
+	if (WARN_ON(elinfo->type != MLXSW_AFK_ELEMENT_TYPE_U32))
+		return;
+	__mlxsw_item_set32(values->storage.key, storage_item, 0, key_value);
+	__mlxsw_item_set32(values->storage.mask, storage_item, 0, mask_value);
+	mlxsw_afk_element_usage_add(&values->elusage, element);
+}
+EXPORT_SYMBOL(mlxsw_afk_values_add_u32);
+
+void mlxsw_afk_values_add_buf(struct mlxsw_afk_element_values *values,
+			      enum mlxsw_afk_element element,
+			      const char *key_value, const char *mask_value,
+			      unsigned int len)
+{
+	const struct mlxsw_afk_element_info *elinfo =
+				&mlxsw_afk_element_infos[element];
+	const struct mlxsw_item *storage_item = &elinfo->item;
+
+	if (!memchr_inv(mask_value, 0, len)) /* If mask is zero */
+		return;
+	if (WARN_ON(elinfo->type != MLXSW_AFK_ELEMENT_TYPE_BUF) ||
+	    WARN_ON(elinfo->item.size.bytes != len))
+		return;
+	__mlxsw_item_memcpy_to(values->storage.key, key_value,
+			       storage_item, 0);
+	__mlxsw_item_memcpy_to(values->storage.mask, mask_value,
+			       storage_item, 0);
+	mlxsw_afk_element_usage_add(&values->elusage, element);
+}
+EXPORT_SYMBOL(mlxsw_afk_values_add_buf);
+
+static void mlxsw_afk_encode_u32(const struct mlxsw_item *storage_item,
+				 const struct mlxsw_item *output_item,
+				 char *storage, char *output_indexed)
+{
+	u32 value;
+
+	value = __mlxsw_item_get32(storage, storage_item, 0);
+	__mlxsw_item_set32(output_indexed, output_item, 0, value);
+}
+
+static void mlxsw_afk_encode_buf(const struct mlxsw_item *storage_item,
+				 const struct mlxsw_item *output_item,
+				 char *storage, char *output_indexed)
+{
+	char *storage_data = __mlxsw_item_data(storage, storage_item, 0);
+	char *output_data = __mlxsw_item_data(output_indexed, output_item, 0);
+	size_t len = output_item->size.bytes;
+
+	memcpy(output_data, storage_data, len);
+}
+
+#define MLXSW_AFK_KEY_BLOCK_SIZE 16
+
+static void mlxsw_afk_encode_one(const struct mlxsw_afk_element_inst *elinst,
+				 int block_index, char *storage, char *output)
+{
+	char *output_indexed = output + block_index * MLXSW_AFK_KEY_BLOCK_SIZE;
+	const struct mlxsw_item *storage_item = &elinst->info->item;
+	const struct mlxsw_item *output_item = &elinst->item;
+
+	if (elinst->type == MLXSW_AFK_ELEMENT_TYPE_U32)
+		mlxsw_afk_encode_u32(storage_item, output_item,
+				     storage, output_indexed);
+	else if (elinst->type == MLXSW_AFK_ELEMENT_TYPE_BUF)
+		mlxsw_afk_encode_buf(storage_item, output_item,
+				     storage, output_indexed);
+}
+
+void mlxsw_afk_encode(struct mlxsw_afk_key_info *key_info,
+		      struct mlxsw_afk_element_values *values,
+		      char *key, char *mask)
+{
+	const struct mlxsw_afk_element_inst *elinst;
+	enum mlxsw_afk_element element;
+	int block_index;
+
+	mlxsw_afk_element_usage_for_each(element, &values->elusage) {
+		elinst = mlxsw_afk_key_info_elinst_get(key_info, element,
+						       &block_index);
+		if (!elinst)
+			continue;
+		mlxsw_afk_encode_one(elinst, block_index,
+				     values->storage.key, key);
+		mlxsw_afk_encode_one(elinst, block_index,
+				     values->storage.mask, mask);
+	}
+}
+EXPORT_SYMBOL(mlxsw_afk_encode);

drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.h

@@ -0,0 +1,238 @@
+/*
+ * drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.h
+ * Copyright (c) 2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2017 Jiri Pirko <jiri@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _MLXSW_CORE_ACL_FLEX_KEYS_H
+#define _MLXSW_CORE_ACL_FLEX_KEYS_H
+
+#include <linux/types.h>
+#include <linux/bitmap.h>
+
+#include "item.h"
+
+enum mlxsw_afk_element {
+	MLXSW_AFK_ELEMENT_SRC_SYS_PORT,
+	MLXSW_AFK_ELEMENT_DMAC,
+	MLXSW_AFK_ELEMENT_SMAC,
+	MLXSW_AFK_ELEMENT_ETHERTYPE,
+	MLXSW_AFK_ELEMENT_IP_PROTO,
+	MLXSW_AFK_ELEMENT_SRC_IP4,
+	MLXSW_AFK_ELEMENT_DST_IP4,
+	MLXSW_AFK_ELEMENT_SRC_IP6_HI,
+	MLXSW_AFK_ELEMENT_SRC_IP6_LO,
+	MLXSW_AFK_ELEMENT_DST_IP6_HI,
+	MLXSW_AFK_ELEMENT_DST_IP6_LO,
+	MLXSW_AFK_ELEMENT_DST_L4_PORT,
+	MLXSW_AFK_ELEMENT_SRC_L4_PORT,
+	MLXSW_AFK_ELEMENT_MAX,
+};
+
+enum mlxsw_afk_element_type {
+	MLXSW_AFK_ELEMENT_TYPE_U32,
+	MLXSW_AFK_ELEMENT_TYPE_BUF,
+};
+
+struct mlxsw_afk_element_info {
+	enum mlxsw_afk_element element; /* element ID */
+	enum mlxsw_afk_element_type type;
+	struct mlxsw_item item; /* element geometry in internal storage */
+};
+
+#define MLXSW_AFK_ELEMENT_INFO(_type, _element, _offset, _shift, _size)		\
+	[MLXSW_AFK_ELEMENT_##_element] = {					\
+		.element = MLXSW_AFK_ELEMENT_##_element,			\
+		.type = _type,							\
+		.item = {							\
+			.offset = _offset,					\
+			.shift = _shift,					\
+			.size = {.bits = _size},				\
+			.name = #_element,					\
+		},								\
+	}
+
+#define MLXSW_AFK_ELEMENT_INFO_U32(_element, _offset, _shift, _size)		\
+	MLXSW_AFK_ELEMENT_INFO(MLXSW_AFK_ELEMENT_TYPE_U32,			\
+			       _element, _offset, _shift, _size)
+
+#define MLXSW_AFK_ELEMENT_INFO_BUF(_element, _offset, _size)			\
+	MLXSW_AFK_ELEMENT_INFO(MLXSW_AFK_ELEMENT_TYPE_BUF,			\
+			       _element, _offset, 0, _size)
+
+/* For the purpose of the driver, define a internal storage scratchpad
+ * that will be used to store key/mask values. For each defined element type
+ * define an internal storage geometry.
+ */
+static const struct mlxsw_afk_element_info mlxsw_afk_element_infos[] = {
+	MLXSW_AFK_ELEMENT_INFO_U32(SRC_SYS_PORT, 0x00, 16, 16),
+	MLXSW_AFK_ELEMENT_INFO_BUF(DMAC, 0x04, 6),
+	MLXSW_AFK_ELEMENT_INFO_BUF(SMAC, 0x0A, 6),
+	MLXSW_AFK_ELEMENT_INFO_U32(ETHERTYPE, 0x00, 0, 16),
+	MLXSW_AFK_ELEMENT_INFO_U32(IP_PROTO, 0x10, 0, 8),
+	MLXSW_AFK_ELEMENT_INFO_U32(SRC_IP4, 0x18, 0, 32),
+	MLXSW_AFK_ELEMENT_INFO_U32(DST_IP4, 0x1C, 0, 32),
+	MLXSW_AFK_ELEMENT_INFO_BUF(SRC_IP6_HI, 0x18, 8),
+	MLXSW_AFK_ELEMENT_INFO_BUF(SRC_IP6_LO, 0x20, 8),
+	MLXSW_AFK_ELEMENT_INFO_BUF(DST_IP6_HI, 0x28, 8),
+	MLXSW_AFK_ELEMENT_INFO_BUF(DST_IP6_LO, 0x30, 8),
+	MLXSW_AFK_ELEMENT_INFO_U32(DST_L4_PORT, 0x14, 0, 16),
+	MLXSW_AFK_ELEMENT_INFO_U32(SRC_L4_PORT, 0x14, 16, 16),
+};
+
+#define MLXSW_AFK_ELEMENT_STORAGE_SIZE 0x38
+
+struct mlxsw_afk_element_inst { /* element instance in actual block */
+	const struct mlxsw_afk_element_info *info;
+	enum mlxsw_afk_element_type type;
+	struct mlxsw_item item; /* element geometry in block */
+};
+
+#define MLXSW_AFK_ELEMENT_INST(_type, _element, _offset, _shift, _size)		\
+	{									\
+		.info = &mlxsw_afk_element_infos[MLXSW_AFK_ELEMENT_##_element],	\
+		.type = _type,							\
+		.item = {							\
+			.offset = _offset,					\
+			.shift = _shift,					\
+			.size = {.bits = _size},				\
+			.name = #_element,					\
+		},								\
+	}
+
+#define MLXSW_AFK_ELEMENT_INST_U32(_element, _offset, _shift, _size)		\
+	MLXSW_AFK_ELEMENT_INST(MLXSW_AFK_ELEMENT_TYPE_U32,			\
+			       _element, _offset, _shift, _size)
+
+#define MLXSW_AFK_ELEMENT_INST_BUF(_element, _offset, _size)			\
+	MLXSW_AFK_ELEMENT_INST(MLXSW_AFK_ELEMENT_TYPE_BUF,			\
+			       _element, _offset, 0, _size)
+
+struct mlxsw_afk_block {
+	u16 encoding; /* block ID */
+	struct mlxsw_afk_element_inst *instances;
+	unsigned int instances_count;
+};
+
+#define MLXSW_AFK_BLOCK(_encoding, _instances)					\
+	{									\
+		.encoding = _encoding,						\
+		.instances = _instances,					\
+		.instances_count = ARRAY_SIZE(_instances),			\
+	}
+
+struct mlxsw_afk_element_usage {
+	DECLARE_BITMAP(usage, MLXSW_AFK_ELEMENT_MAX);
+};
+
+#define mlxsw_afk_element_usage_for_each(element, elusage)			\
+	for_each_set_bit(element, (elusage)->usage, MLXSW_AFK_ELEMENT_MAX)
+
+static inline void
+mlxsw_afk_element_usage_add(struct mlxsw_afk_element_usage *elusage,
+			    enum mlxsw_afk_element element)
+{
+	__set_bit(element, elusage->usage);
+}
+
+static inline void
+mlxsw_afk_element_usage_zero(struct mlxsw_afk_element_usage *elusage)
+{
+	bitmap_zero(elusage->usage, MLXSW_AFK_ELEMENT_MAX);
+}
+
+static inline void
+mlxsw_afk_element_usage_fill(struct mlxsw_afk_element_usage *elusage,
+			     const enum mlxsw_afk_element *elements,
+			     unsigned int elements_count)
+{
+	int i;
+
+	mlxsw_afk_element_usage_zero(elusage);
+	for (i = 0; i < elements_count; i++)
+		mlxsw_afk_element_usage_add(elusage, elements[i]);
+}
+
+static inline bool
+mlxsw_afk_element_usage_subset(struct mlxsw_afk_element_usage *elusage_small,
+			       struct mlxsw_afk_element_usage *elusage_big)
+{
+	int i;
+
+	for (i = 0; i < MLXSW_AFK_ELEMENT_MAX; i++)
+		if (test_bit(i, elusage_small->usage) &&
+		    !test_bit(i, elusage_big->usage))
+			return false;
+	return true;
+}
+
+struct mlxsw_afk;
+
+struct mlxsw_afk *mlxsw_afk_create(unsigned int max_blocks,
+				   const struct mlxsw_afk_block *blocks,
+				   unsigned int blocks_count);
+void mlxsw_afk_destroy(struct mlxsw_afk *mlxsw_afk);
+
+struct mlxsw_afk_key_info;
+
+struct mlxsw_afk_key_info *
+mlxsw_afk_key_info_get(struct mlxsw_afk *mlxsw_afk,
+		       struct mlxsw_afk_element_usage *elusage);
+void mlxsw_afk_key_info_put(struct mlxsw_afk_key_info *key_info);
+bool mlxsw_afk_key_info_subset(struct mlxsw_afk_key_info *key_info,
+			       struct mlxsw_afk_element_usage *elusage);
+
+u16
+mlxsw_afk_key_info_block_encoding_get(const struct mlxsw_afk_key_info *key_info,
+				      int block_index);
+unsigned int
+mlxsw_afk_key_info_blocks_count_get(const struct mlxsw_afk_key_info *key_info);
+
+struct mlxsw_afk_element_values {
+	struct mlxsw_afk_element_usage elusage;
+	struct {
+		char key[MLXSW_AFK_ELEMENT_STORAGE_SIZE];
+		char mask[MLXSW_AFK_ELEMENT_STORAGE_SIZE];
+	} storage;
+};
+
+void mlxsw_afk_values_add_u32(struct mlxsw_afk_element_values *values,
+			      enum mlxsw_afk_element element,
+			      u32 key_value, u32 mask_value);
+void mlxsw_afk_values_add_buf(struct mlxsw_afk_element_values *values,
+			      enum mlxsw_afk_element element,
+			      const char *key_value, const char *mask_value,
+			      unsigned int len);
+void mlxsw_afk_encode(struct mlxsw_afk_key_info *key_info,
+		      struct mlxsw_afk_element_values *values,
+		      char *key, char *mask);
+
+#endif

drivers/net/ethernet/mellanox/mlxsw/item.h

@@ -1,7 +1,7 @@
 /*
  * drivers/net/ethernet/mellanox/mlxsw/item.h
- * Copyright (c) 2015 Mellanox Technologies. All rights reserved.
- * Copyright (c) 2015 Jiri Pirko <jiri@mellanox.com>
+ * Copyright (c) 2015-2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2015-2017 Jiri Pirko <jiri@mellanox.com>
  * Copyright (c) 2015 Ido Schimmel <idosch@mellanox.com>
  *
  * Redistribution and use in source and binary forms, with or without
@@ -72,6 +72,40 @@ __mlxsw_item_offset(const struct mlxsw_item *item, unsigned short index,
 		typesize);
 }
 
+static inline u8 __mlxsw_item_get8(const char *buf,
+				   const struct mlxsw_item *item,
+				   unsigned short index)
+{
+	unsigned int offset = __mlxsw_item_offset(item, index, sizeof(u8));
+	u8 *b = (u8 *) buf;
+	u8 tmp;
+
+	tmp = b[offset];
+	tmp >>= item->shift;
+	tmp &= GENMASK(item->size.bits - 1, 0);
+	if (item->no_real_shift)
+		tmp <<= item->shift;
+	return tmp;
+}
+
+static inline void __mlxsw_item_set8(char *buf, const struct mlxsw_item *item,
+				     unsigned short index, u8 val)
+{
+	unsigned int offset = __mlxsw_item_offset(item, index,
+						  sizeof(u8));
+	u8 *b = (u8 *) buf;
+	u8 mask = GENMASK(item->size.bits - 1, 0) << item->shift;
+	u8 tmp;
+
+	if (!item->no_real_shift)
+		val <<= item->shift;
+	val &= mask;
+	tmp = b[offset];
+	tmp &= ~mask;
+	tmp |= val;
+	b[offset] = tmp;
+}
+
 static inline u16 __mlxsw_item_get16(const char *buf,
 				     const struct mlxsw_item *item,
 				     unsigned short index)
@@ -191,6 +225,14 @@ static inline void __mlxsw_item_memcpy_to(char *buf, const char *src,
 	memcpy(&buf[offset], src, item->size.bytes);
 }
 
+static inline char *__mlxsw_item_data(char *buf, const struct mlxsw_item *item,
+				      unsigned short index)
+{
+	unsigned int offset = __mlxsw_item_offset(item, index, sizeof(char));
+
+	return &buf[offset];
+}
+
 static inline u16
 __mlxsw_item_bit_array_offset(const struct mlxsw_item *item,
 			      u16 index, u8 *shift)
@@ -253,6 +295,47 @@ static inline void __mlxsw_item_bit_array_set(char *buf,
  * _iname: item name within the container
  */
 
+#define MLXSW_ITEM8(_type, _cname, _iname, _offset, _shift, _sizebits)		\
+static struct mlxsw_item __ITEM_NAME(_type, _cname, _iname) = {			\
+	.offset = _offset,							\
+	.shift = _shift,							\
+	.size = {.bits = _sizebits,},						\
+	.name = #_type "_" #_cname "_" #_iname,					\
+};										\
+static inline u8 mlxsw_##_type##_##_cname##_##_iname##_get(const char *buf)	\
+{										\
+	return __mlxsw_item_get8(buf, &__ITEM_NAME(_type, _cname, _iname), 0);	\
+}										\
+static inline void mlxsw_##_type##_##_cname##_##_iname##_set(char *buf, u8 val)\
+{										\
+	__mlxsw_item_set8(buf, &__ITEM_NAME(_type, _cname, _iname), 0, val);	\
+}
+
+#define MLXSW_ITEM8_INDEXED(_type, _cname, _iname, _offset, _shift, _sizebits,	\
+			    _step, _instepoffset, _norealshift)			\
+static struct mlxsw_item __ITEM_NAME(_type, _cname, _iname) = {			\
+	.offset = _offset,							\
+	.step = _step,								\
+	.in_step_offset = _instepoffset,					\
+	.shift = _shift,							\
+	.no_real_shift = _norealshift,						\
+	.size = {.bits = _sizebits,},						\
+	.name = #_type "_" #_cname "_" #_iname,					\
+};										\
+static inline u8								\
+mlxsw_##_type##_##_cname##_##_iname##_get(const char *buf, unsigned short index)\
+{										\
+	return __mlxsw_item_get8(buf, &__ITEM_NAME(_type, _cname, _iname),	\
+				 index);					\
+}										\
+static inline void								\
+mlxsw_##_type##_##_cname##_##_iname##_set(char *buf, unsigned short index,	\
+					  u8 val)				\
+{										\
+	__mlxsw_item_set8(buf, &__ITEM_NAME(_type, _cname, _iname),		\
+			  index, val);						\
+}
+
 #define MLXSW_ITEM16(_type, _cname, _iname, _offset, _shift, _sizebits)		\
 static struct mlxsw_item __ITEM_NAME(_type, _cname, _iname) = {			\
 	.offset = _offset,							\
@@ -393,6 +476,11 @@ mlxsw_##_type##_##_cname##_##_iname##_memcpy_to(char *buf, const char *src)	\
 {										\
 	__mlxsw_item_memcpy_to(buf, src,					\
 			       &__ITEM_NAME(_type, _cname, _iname), 0);		\
+}										\
+static inline char *								\
+mlxsw_##_type##_##_cname##_##_iname##_data(char *buf)				\
+{										\
+	return __mlxsw_item_data(buf, &__ITEM_NAME(_type, _cname, _iname), 0);	\
 }
 
 #define MLXSW_ITEM_BUF_INDEXED(_type, _cname, _iname, _offset, _sizebytes,	\
@@ -419,6 +507,12 @@ mlxsw_##_type##_##_cname##_##_iname##_memcpy_to(char *buf,			\
 {										\
 	__mlxsw_item_memcpy_to(buf, src,					\
 			       &__ITEM_NAME(_type, _cname, _iname), index);	\
+}										\
+static inline char *								\
+mlxsw_##_type##_##_cname##_##_iname##_data(char *buf, unsigned short index)	\
+{										\
+	return __mlxsw_item_data(buf,						\
+				 &__ITEM_NAME(_type, _cname, _iname), index);	\
 }
 
 #define MLXSW_ITEM_BIT_ARRAY(_type, _cname, _iname, _offset, _sizebytes,	\

drivers/net/ethernet/mellanox/mlxsw/reg.h

@@ -1,9 +1,9 @@
 /*
  * drivers/net/ethernet/mellanox/mlxsw/reg.h
- * Copyright (c) 2015 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2015-2017 Mellanox Technologies. All rights reserved.
  * Copyright (c) 2015-2016 Ido Schimmel <idosch@mellanox.com>
  * Copyright (c) 2015 Elad Raz <eladr@mellanox.com>
- * Copyright (c) 2015-2016 Jiri Pirko <jiri@mellanox.com>
+ * Copyright (c) 2015-2017 Jiri Pirko <jiri@mellanox.com>
  * Copyright (c) 2016 Yotam Gigi <yotamg@mellanox.com>
  *
  * Redistribution and use in source and binary forms, with or without
@@ -1757,6 +1757,505 @@ static inline void mlxsw_reg_spvmlr_pack(char *payload, u8 local_port,
 	}
 }
 
+/* PPBT - Policy-Engine Port Binding Table
+ * ---------------------------------------
+ * This register is used for configuration of the Port Binding Table.
+ */
+#define MLXSW_REG_PPBT_ID 0x3002
+#define MLXSW_REG_PPBT_LEN 0x14
+
+MLXSW_REG_DEFINE(ppbt, MLXSW_REG_PPBT_ID, MLXSW_REG_PPBT_LEN);
+
+enum mlxsw_reg_pxbt_e {
+	MLXSW_REG_PXBT_E_IACL,
+	MLXSW_REG_PXBT_E_EACL,
+};
+
+/* reg_ppbt_e
+ * Access: Index
+ */
+MLXSW_ITEM32(reg, ppbt, e, 0x00, 31, 1);
+
+enum mlxsw_reg_pxbt_op {
+	MLXSW_REG_PXBT_OP_BIND,
+	MLXSW_REG_PXBT_OP_UNBIND,
+};
+
+/* reg_ppbt_op
+ * Access: RW
+ */
+MLXSW_ITEM32(reg, ppbt, op, 0x00, 28, 3);
+
+/* reg_ppbt_local_port
+ * Local port. Not including CPU port.
+ * Access: Index
+ */
+MLXSW_ITEM32(reg, ppbt, local_port, 0x00, 16, 8);
+
+/* reg_ppbt_g
+ * group - When set, the binding is of an ACL group. When cleared,
+ * the binding is of an ACL.
+ * Must be set to 1 for Spectrum.
+ * Access: RW
+ */
+MLXSW_ITEM32(reg, ppbt, g, 0x10, 31, 1);
+
+/* reg_ppbt_acl_info
+ * ACL/ACL group identifier. If the g bit is set, this field should hold
+ * the acl_group_id, else it should hold the acl_id.
+ * Access: RW
+ */
+MLXSW_ITEM32(reg, ppbt, acl_info, 0x10, 0, 16);
+
+static inline void mlxsw_reg_ppbt_pack(char *payload, enum mlxsw_reg_pxbt_e e,
+				       enum mlxsw_reg_pxbt_op op,
+				       u8 local_port, u16 acl_info)
+{
+	MLXSW_REG_ZERO(ppbt, payload);
+	mlxsw_reg_ppbt_e_set(payload, e);
+	mlxsw_reg_ppbt_op_set(payload, op);
+	mlxsw_reg_ppbt_local_port_set(payload, local_port);
+	mlxsw_reg_ppbt_g_set(payload, true);
+	mlxsw_reg_ppbt_acl_info_set(payload, acl_info);
+}
+
+/* PACL - Policy-Engine ACL Register
+ * ---------------------------------
+ * This register is used for configuration of the ACL.
+ */
+#define MLXSW_REG_PACL_ID 0x3004
+#define MLXSW_REG_PACL_LEN 0x70
+
+MLXSW_REG_DEFINE(pacl, MLXSW_REG_PACL_ID, MLXSW_REG_PACL_LEN);
+
+/* reg_pacl_v
+ * Valid. Setting the v bit makes the ACL valid. It should not be cleared
+ * while the ACL is bounded to either a port, VLAN or ACL rule.
+ * Access: RW
+ */
+MLXSW_ITEM32(reg, pacl, v, 0x00, 24, 1);
+
+/* reg_pacl_acl_id
+ * An identifier representing the ACL (managed by software)
+ * Range 0 .. cap_max_acl_regions - 1
+ * Access: Index
+ */
+MLXSW_ITEM32(reg, pacl, acl_id, 0x08, 0, 16);
+
+#define MLXSW_REG_PXXX_TCAM_REGION_INFO_LEN 16
+
+/* reg_pacl_tcam_region_info
+ * Opaque object that represents a TCAM region.
+ * Obtained through PTAR register.
+ * Access: RW
+ */
+MLXSW_ITEM_BUF(reg, pacl, tcam_region_info, 0x30,
+	       MLXSW_REG_PXXX_TCAM_REGION_INFO_LEN);
+
+static inline void mlxsw_reg_pacl_pack(char *payload, u16 acl_id,
+				       bool valid, const char *tcam_region_info)
+{
+	MLXSW_REG_ZERO(pacl, payload);
+	mlxsw_reg_pacl_acl_id_set(payload, acl_id);
+	mlxsw_reg_pacl_v_set(payload, valid);
+	mlxsw_reg_pacl_tcam_region_info_memcpy_to(payload, tcam_region_info);
+}
+
+/* PAGT - Policy-Engine ACL Group Table
+ * ------------------------------------
+ * This register is used for configuration of the ACL Group Table.
+ */
+#define MLXSW_REG_PAGT_ID 0x3005
+#define MLXSW_REG_PAGT_BASE_LEN 0x30
+#define MLXSW_REG_PAGT_ACL_LEN 4
+#define MLXSW_REG_PAGT_ACL_MAX_NUM 16
+#define MLXSW_REG_PAGT_LEN (MLXSW_REG_PAGT_BASE_LEN + \
+		MLXSW_REG_PAGT_ACL_MAX_NUM * MLXSW_REG_PAGT_ACL_LEN)
+
+MLXSW_REG_DEFINE(pagt, MLXSW_REG_PAGT_ID, MLXSW_REG_PAGT_LEN);
+
+/* reg_pagt_size
+ * Number of ACLs in the group.
+ * Size 0 invalidates a group.
+ * Range 0 .. cap_max_acl_group_size (hard coded to 16 for now)
+ * Total number of ACLs in all groups must be lower or equal
+ * to cap_max_acl_tot_groups
+ * Note: a group which is binded must not be invalidated
+ * Access: Index
+ */
+MLXSW_ITEM32(reg, pagt, size, 0x00, 0, 8);
+
+/* reg_pagt_acl_group_id
+ * An identifier (numbered from 0..cap_max_acl_groups-1) representing
+ * the ACL Group identifier (managed by software).
+ * Access: Index
+ */
+MLXSW_ITEM32(reg, pagt, acl_group_id, 0x08, 0, 16);
+
+/* reg_pagt_acl_id
+ * ACL identifier
+ * Access: RW
+ */
+MLXSW_ITEM32_INDEXED(reg, pagt, acl_id, 0x30, 0, 16, 0x04, 0x00, false);
+
+static inline void mlxsw_reg_pagt_pack(char *payload, u16 acl_group_id)
+{
+	MLXSW_REG_ZERO(pagt, payload);
+	mlxsw_reg_pagt_acl_group_id_set(payload, acl_group_id);
+}
+
+static inline void mlxsw_reg_pagt_acl_id_pack(char *payload, int index,
+					      u16 acl_id)
+{
+	u8 size = mlxsw_reg_pagt_size_get(payload);
+
+	if (index >= size)
+		mlxsw_reg_pagt_size_set(payload, index + 1);
+	mlxsw_reg_pagt_acl_id_set(payload, index, acl_id);
+}
+
+/* PTAR - Policy-Engine TCAM Allocation Register
+ * ---------------------------------------------
+ * This register is used for allocation of regions in the TCAM.
+ * Note: Query method is not supported on this register.
+ */
+#define MLXSW_REG_PTAR_ID 0x3006
+#define MLXSW_REG_PTAR_BASE_LEN 0x20
+#define MLXSW_REG_PTAR_KEY_ID_LEN 1
+#define MLXSW_REG_PTAR_KEY_ID_MAX_NUM 16
+#define MLXSW_REG_PTAR_LEN (MLXSW_REG_PTAR_BASE_LEN + \
+		MLXSW_REG_PTAR_KEY_ID_MAX_NUM * MLXSW_REG_PTAR_KEY_ID_LEN)
+
+MLXSW_REG_DEFINE(ptar, MLXSW_REG_PTAR_ID, MLXSW_REG_PTAR_LEN);
+
+enum mlxsw_reg_ptar_op {
+	/* allocate a TCAM region */
+	MLXSW_REG_PTAR_OP_ALLOC,
+	/* resize a TCAM region */
+	MLXSW_REG_PTAR_OP_RESIZE,
+	/* deallocate TCAM region */
+	MLXSW_REG_PTAR_OP_FREE,
+	/* test allocation */
+	MLXSW_REG_PTAR_OP_TEST,
+};
+
+/* reg_ptar_op
+ * Access: OP
+ */
+MLXSW_ITEM32(reg, ptar, op, 0x00, 28, 4);
+
+/* reg_ptar_action_set_type
+ * Type of action set to be used on this region.
+ * For Spectrum, this is always type 2 - "flexible"
+ * Access: WO
+ */
+MLXSW_ITEM32(reg, ptar, action_set_type, 0x00, 16, 8);
+
+/* reg_ptar_key_type
+ * TCAM key type for the region.
+ * For Spectrum, this is always type 0x50 - "FLEX_KEY"
+ * Access: WO
+ */
+MLXSW_ITEM32(reg, ptar, key_type, 0x00, 0, 8);
+
+/* reg_ptar_region_size
+ * TCAM region size. When allocating/resizing this is the requested size,
+ * the response is the actual size. Note that actual size may be
+ * larger than requested.
+ * Allowed range 1 .. cap_max_rules-1
+ * Reserved during op deallocate.
+ * Access: WO
+ */
+MLXSW_ITEM32(reg, ptar, region_size, 0x04, 0, 16);
+
+/* reg_ptar_region_id
+ * Region identifier
+ * Range 0 .. cap_max_regions-1
+ * Access: Index
+ */
+MLXSW_ITEM32(reg, ptar, region_id, 0x08, 0, 16);
+
+/* reg_ptar_tcam_region_info
+ * Opaque object that represents the TCAM region.
+ * Returned when allocating a region.
+ * Provided by software for ACL generation and region deallocation and resize.
+ * Access: RW
+ */
+MLXSW_ITEM_BUF(reg, ptar, tcam_region_info, 0x10,
+	       MLXSW_REG_PXXX_TCAM_REGION_INFO_LEN);
+
+/* reg_ptar_flexible_key_id
+ * Identifier of the Flexible Key.
+ * Only valid if key_type == "FLEX_KEY"
+ * The key size will be rounded up to one of the following values:
+ * 9B, 18B, 36B, 54B.
+ * This field is reserved for in resize operation.
+ * Access: WO
+ */
+MLXSW_ITEM8_INDEXED(reg, ptar, flexible_key_id, 0x20, 0, 8,
+		    MLXSW_REG_PTAR_KEY_ID_LEN, 0x00, false);
+
+static inline void mlxsw_reg_ptar_pack(char *payload, enum mlxsw_reg_ptar_op op,
+				       u16 region_size, u16 region_id,
+				       const char *tcam_region_info)
+{
+	MLXSW_REG_ZERO(ptar, payload);
+	mlxsw_reg_ptar_op_set(payload, op);
+	mlxsw_reg_ptar_action_set_type_set(payload, 2); /* "flexible" */
+	mlxsw_reg_ptar_key_type_set(payload, 0x50); /* "FLEX_KEY" */
+	mlxsw_reg_ptar_region_size_set(payload, region_size);
+	mlxsw_reg_ptar_region_id_set(payload, region_id);
+	mlxsw_reg_ptar_tcam_region_info_memcpy_to(payload, tcam_region_info);
+}
+
+static inline void mlxsw_reg_ptar_key_id_pack(char *payload, int index,
+					      u16 key_id)
+{
+	mlxsw_reg_ptar_flexible_key_id_set(payload, index, key_id);
+}
+
+static inline void mlxsw_reg_ptar_unpack(char *payload, char *tcam_region_info)
+{
+	mlxsw_reg_ptar_tcam_region_info_memcpy_from(payload, tcam_region_info);
+}
+
+/* PPBS - Policy-Engine Policy Based Switching Register
+ * ----------------------------------------------------
+ * This register retrieves and sets Policy Based Switching Table entries.
+ */
+#define MLXSW_REG_PPBS_ID 0x300C
+#define MLXSW_REG_PPBS_LEN 0x14
+
+MLXSW_REG_DEFINE(ppbs, MLXSW_REG_PPBS_ID, MLXSW_REG_PPBS_LEN);
+
+/* reg_ppbs_pbs_ptr
+ * Index into the PBS table.
+ * For Spectrum, the index points to the KVD Linear.
+ * Access: Index
+ */
+MLXSW_ITEM32(reg, ppbs, pbs_ptr, 0x08, 0, 24);
+
+/* reg_ppbs_system_port
+ * Unique port identifier for the final destination of the packet.
+ * Access: RW
+ */
+MLXSW_ITEM32(reg, ppbs, system_port, 0x10, 0, 16);
+
+static inline void mlxsw_reg_ppbs_pack(char *payload, u32 pbs_ptr,
+				       u16 system_port)
+{
+	MLXSW_REG_ZERO(ppbs, payload);
+	mlxsw_reg_ppbs_pbs_ptr_set(payload, pbs_ptr);
+	mlxsw_reg_ppbs_system_port_set(payload, system_port);
+}
+
+/* PRCR - Policy-Engine Rules Copy Register
+ * ----------------------------------------
+ * This register is used for accessing rules within a TCAM region.
+ */
+#define MLXSW_REG_PRCR_ID 0x300D
+#define MLXSW_REG_PRCR_LEN 0x40
+
+MLXSW_REG_DEFINE(prcr, MLXSW_REG_PRCR_ID, MLXSW_REG_PRCR_LEN);
+
+enum mlxsw_reg_prcr_op {
+	/* Move rules. Moves the rules from "tcam_region_info" starting
+	 * at offset "offset" to "dest_tcam_region_info"
+	 * at offset "dest_offset."
+	 */
+	MLXSW_REG_PRCR_OP_MOVE,
+	/* Copy rules. Copies the rules from "tcam_region_info" starting
+	 * at offset "offset" to "dest_tcam_region_info"
+	 * at offset "dest_offset."
+	 */
+	MLXSW_REG_PRCR_OP_COPY,
+};
+
+/* reg_prcr_op
+ * Access: OP
+ */
+MLXSW_ITEM32(reg, prcr, op, 0x00, 28, 4);
+
+/* reg_prcr_offset
+ * Offset within the source region to copy/move from.
+ * Access: Index
+ */
+MLXSW_ITEM32(reg, prcr, offset, 0x00, 0, 16);
+
+/* reg_prcr_size
+ * The number of rules to copy/move.
+ * Access: WO
+ */
+MLXSW_ITEM32(reg, prcr, size, 0x04, 0, 16);
+
+/* reg_prcr_tcam_region_info
+ * Opaque object that represents the source TCAM region.
+ * Access: Index
+ */
+MLXSW_ITEM_BUF(reg, prcr, tcam_region_info, 0x10,
+	       MLXSW_REG_PXXX_TCAM_REGION_INFO_LEN);
+
+/* reg_prcr_dest_offset
+ * Offset within the source region to copy/move to.
+ * Access: Index
+ */
+MLXSW_ITEM32(reg, prcr, dest_offset, 0x20, 0, 16);
+
+/* reg_prcr_dest_tcam_region_info
+ * Opaque object that represents the destination TCAM region.
+ * Access: Index
+ */
+MLXSW_ITEM_BUF(reg, prcr, dest_tcam_region_info, 0x30,
+	       MLXSW_REG_PXXX_TCAM_REGION_INFO_LEN);
+
+static inline void mlxsw_reg_prcr_pack(char *payload, enum mlxsw_reg_prcr_op op,
+				       const char *src_tcam_region_info,
+				       u16 src_offset,
+				       const char *dest_tcam_region_info,
+				       u16 dest_offset, u16 size)
+{
+	MLXSW_REG_ZERO(prcr, payload);
+	mlxsw_reg_prcr_op_set(payload, op);
+	mlxsw_reg_prcr_offset_set(payload, src_offset);
+	mlxsw_reg_prcr_size_set(payload, size);
+	mlxsw_reg_prcr_tcam_region_info_memcpy_to(payload,
+						  src_tcam_region_info);
+	mlxsw_reg_prcr_dest_offset_set(payload, dest_offset);
+	mlxsw_reg_prcr_dest_tcam_region_info_memcpy_to(payload,
+						       dest_tcam_region_info);
+}
+
+/* PEFA - Policy-Engine Extended Flexible Action Register
+ * ------------------------------------------------------
+ * This register is used for accessing an extended flexible action entry
+ * in the central KVD Linear Database.
+ */
+#define MLXSW_REG_PEFA_ID 0x300F
+#define MLXSW_REG_PEFA_LEN 0xB0
+
+MLXSW_REG_DEFINE(pefa, MLXSW_REG_PEFA_ID, MLXSW_REG_PEFA_LEN);
+
+/* reg_pefa_index
+ * Index in the KVD Linear Centralized Database.
+ * Access: Index
+ */
+MLXSW_ITEM32(reg, pefa, index, 0x00, 0, 24);
+
+#define MLXSW_REG_PXXX_FLEX_ACTION_SET_LEN 0xA8
+
+/* reg_pefa_flex_action_set
+ * Action-set to perform when rule is matched.
+ * Must be zero padded if action set is shorter.
+ * Access: RW
+ */
+MLXSW_ITEM_BUF(reg, pefa, flex_action_set, 0x08,
+	       MLXSW_REG_PXXX_FLEX_ACTION_SET_LEN);
+
+static inline void mlxsw_reg_pefa_pack(char *payload, u32 index,
+				       const char *flex_action_set)
+{
+	MLXSW_REG_ZERO(pefa, payload);
+	mlxsw_reg_pefa_index_set(payload, index);
+	mlxsw_reg_pefa_flex_action_set_memcpy_to(payload, flex_action_set);
+}
+
+/* PTCE-V2 - Policy-Engine TCAM Entry Register Version 2
+ * -----------------------------------------------------
+ * This register is used for accessing rules within a TCAM region.
+ * It is a new version of PTCE in order to support wider key,
+ * mask and action within a TCAM region. This register is not supported
+ * by SwitchX and SwitchX-2.
+ */
+#define MLXSW_REG_PTCE2_ID 0x3017
+#define MLXSW_REG_PTCE2_LEN 0x1D8
+
+MLXSW_REG_DEFINE(ptce2, MLXSW_REG_PTCE2_ID, MLXSW_REG_PTCE2_LEN);
+
+/* reg_ptce2_v
+ * Valid.
+ * Access: RW
+ */
+MLXSW_ITEM32(reg, ptce2, v, 0x00, 31, 1);
+
+/* reg_ptce2_a
+ * Activity. Set if a packet lookup has hit on the specific entry.
+ * To clear the "a" bit, use "clear activity" op or "clear on read" op.
+ * Access: RO
+ */
+MLXSW_ITEM32(reg, ptce2, a, 0x00, 30, 1);
+
+enum mlxsw_reg_ptce2_op {
+	/* Read operation. */
+	MLXSW_REG_PTCE2_OP_QUERY_READ = 0,
+	/* clear on read operation. Used to read entry
+	 * and clear Activity bit.
+	 */
+	MLXSW_REG_PTCE2_OP_QUERY_CLEAR_ON_READ = 1,
+	/* Write operation. Used to write a new entry to the table.
+	 * All R/W fields are relevant for new entry. Activity bit is set
+	 * for new entries - Note write with v = 0 will delete the entry.
+	 */
+	MLXSW_REG_PTCE2_OP_WRITE_WRITE = 0,
+	/* Update action. Only action set will be updated. */
+	MLXSW_REG_PTCE2_OP_WRITE_UPDATE = 1,
+	/* Clear activity. A bit is cleared for the entry. */
+	MLXSW_REG_PTCE2_OP_WRITE_CLEAR_ACTIVITY = 2,
+};
+
+/* reg_ptce2_op
+ * Access: OP
+ */
+MLXSW_ITEM32(reg, ptce2, op, 0x00, 20, 3);
+
+/* reg_ptce2_offset
+ * Access: Index
+ */
+MLXSW_ITEM32(reg, ptce2, offset, 0x00, 0, 16);
+
+/* reg_ptce2_tcam_region_info
+ * Opaque object that represents the TCAM region.
+ * Access: Index
+ */
+MLXSW_ITEM_BUF(reg, ptce2, tcam_region_info, 0x10,
+	       MLXSW_REG_PXXX_TCAM_REGION_INFO_LEN);
+
+#define MLXSW_REG_PTCE2_FLEX_KEY_BLOCKS_LEN 96
+
+/* reg_ptce2_flex_key_blocks
+ * ACL Key.
+ * Access: RW
+ */
+MLXSW_ITEM_BUF(reg, ptce2, flex_key_blocks, 0x20,
+	       MLXSW_REG_PTCE2_FLEX_KEY_BLOCKS_LEN);
+
+/* reg_ptce2_mask
+ * mask- in the same size as key. A bit that is set directs the TCAM
+ * to compare the corresponding bit in key. A bit that is clear directs
+ * the TCAM to ignore the corresponding bit in key.
+ * Access: RW
+ */
+MLXSW_ITEM_BUF(reg, ptce2, mask, 0x80,
+	       MLXSW_REG_PTCE2_FLEX_KEY_BLOCKS_LEN);
+
+/* reg_ptce2_flex_action_set
+ * ACL action set.
+ * Access: RW
+ */
+MLXSW_ITEM_BUF(reg, ptce2, flex_action_set, 0xE0,
+	       MLXSW_REG_PXXX_FLEX_ACTION_SET_LEN);
+
+static inline void mlxsw_reg_ptce2_pack(char *payload, bool valid,
+					enum mlxsw_reg_ptce2_op op,
+					const char *tcam_region_info,
+					u16 offset)
+{
+	MLXSW_REG_ZERO(ptce2, payload);
+	mlxsw_reg_ptce2_v_set(payload, valid);
+	mlxsw_reg_ptce2_op_set(payload, op);
+	mlxsw_reg_ptce2_offset_set(payload, offset);
+	mlxsw_reg_ptce2_tcam_region_info_memcpy_to(payload, tcam_region_info);
+}
+
 /* QPCR - QoS Policer Configuration Register
  * -----------------------------------------
  * The QPCR register is used to create policers - that limit
@@ -5434,6 +5933,14 @@ static const struct mlxsw_reg_info *mlxsw_reg_infos[] = {
 	MLXSW_REG(svpe),
 	MLXSW_REG(sfmr),
 	MLXSW_REG(spvmlr),
+	MLXSW_REG(ppbt),
+	MLXSW_REG(pacl),
+	MLXSW_REG(pagt),
+	MLXSW_REG(ptar),
+	MLXSW_REG(ppbs),
+	MLXSW_REG(prcr),
+	MLXSW_REG(pefa),
+	MLXSW_REG(ptce2),
 	MLXSW_REG(qpcr),
 	MLXSW_REG(qtct),
 	MLXSW_REG(qeec),

drivers/net/ethernet/mellanox/mlxsw/resources.h

@@ -1,7 +1,7 @@
 /*
  * drivers/net/ethernet/mellanox/mlxsw/resources.h
- * Copyright (c) 2016 Mellanox Technologies. All rights reserved.
- * Copyright (c) 2016 Jiri Pirko <jiri@mellanox.com>
+ * Copyright (c) 2016-2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2016-2017 Jiri Pirko <jiri@mellanox.com>
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are met:
@@ -48,6 +48,14 @@ enum mlxsw_res_id {
 	MLXSW_RES_ID_MAX_LAG,
 	MLXSW_RES_ID_MAX_LAG_MEMBERS,
 	MLXSW_RES_ID_MAX_BUFFER_SIZE,
+	MLXSW_RES_ID_ACL_MAX_TCAM_REGIONS,
+	MLXSW_RES_ID_ACL_MAX_TCAM_RULES,
+	MLXSW_RES_ID_ACL_MAX_REGIONS,
+	MLXSW_RES_ID_ACL_MAX_GROUPS,
+	MLXSW_RES_ID_ACL_MAX_GROUP_SIZE,
+	MLXSW_RES_ID_ACL_FLEX_KEYS,
+	MLXSW_RES_ID_ACL_MAX_ACTION_PER_RULE,
+	MLXSW_RES_ID_ACL_ACTIONS_PER_SET,
 	MLXSW_RES_ID_MAX_CPU_POLICERS,
 	MLXSW_RES_ID_MAX_VRS,
 	MLXSW_RES_ID_MAX_RIFS,
@@ -72,6 +80,14 @@ static u16 mlxsw_res_ids[] = {
 	[MLXSW_RES_ID_MAX_LAG] = 0x2520,
 	[MLXSW_RES_ID_MAX_LAG_MEMBERS] = 0x2521,
 	[MLXSW_RES_ID_MAX_BUFFER_SIZE] = 0x2802,	/* Bytes */
+	[MLXSW_RES_ID_ACL_MAX_TCAM_REGIONS] = 0x2901,
+	[MLXSW_RES_ID_ACL_MAX_TCAM_RULES] = 0x2902,
+	[MLXSW_RES_ID_ACL_MAX_REGIONS] = 0x2903,
+	[MLXSW_RES_ID_ACL_MAX_GROUPS] = 0x2904,
+	[MLXSW_RES_ID_ACL_MAX_GROUP_SIZE] = 0x2905,
+	[MLXSW_RES_ID_ACL_FLEX_KEYS] = 0x2910,
+	[MLXSW_RES_ID_ACL_MAX_ACTION_PER_RULE] = 0x2911,
+	[MLXSW_RES_ID_ACL_ACTIONS_PER_SET] = 0x2912,
 	[MLXSW_RES_ID_MAX_CPU_POLICERS] = 0x2A13,
 	[MLXSW_RES_ID_MAX_VRS] = 0x2C01,
 	[MLXSW_RES_ID_MAX_RIFS] = 0x2C02,

drivers/net/ethernet/mellanox/mlxsw/spectrum.c

@@ -1,7 +1,7 @@
 /*
  * drivers/net/ethernet/mellanox/mlxsw/spectrum.c
- * Copyright (c) 2015 Mellanox Technologies. All rights reserved.
- * Copyright (c) 2015 Jiri Pirko <jiri@mellanox.com>
+ * Copyright (c) 2015-2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2015-2017 Jiri Pirko <jiri@mellanox.com>
  * Copyright (c) 2015 Ido Schimmel <idosch@mellanox.com>
  * Copyright (c) 2015 Elad Raz <eladr@mellanox.com>
  *
@@ -138,8 +138,6 @@ MLXSW_ITEM32(tx, hdr, fid, 0x08, 0, 16);
  */
 MLXSW_ITEM32(tx, hdr, type, 0x0C, 0, 4);
 
-static bool mlxsw_sp_port_dev_check(const struct net_device *dev);
-
 static void mlxsw_sp_txhdr_construct(struct sk_buff *skb,
 				     const struct mlxsw_tx_info *tx_info)
 {
@@ -1357,7 +1355,8 @@ static int mlxsw_sp_setup_tc(struct net_device *dev, u32 handle,
 	struct mlxsw_sp_port *mlxsw_sp_port = netdev_priv(dev);
 	bool ingress = TC_H_MAJ(handle) == TC_H_MAJ(TC_H_INGRESS);
 
-	if (tc->type == TC_SETUP_MATCHALL) {
+	switch (tc->type) {
+	case TC_SETUP_MATCHALL:
 		switch (tc->cls_mall->command) {
 		case TC_CLSMATCHALL_REPLACE:
 			return mlxsw_sp_port_add_cls_matchall(mlxsw_sp_port,
@@ -1371,6 +1370,18 @@ static int mlxsw_sp_setup_tc(struct net_device *dev, u32 handle,
 		default:
 			return -EINVAL;
 		}
+	case TC_SETUP_CLSFLOWER:
+		switch (tc->cls_flower->command) {
+		case TC_CLSFLOWER_REPLACE:
+			return mlxsw_sp_flower_replace(mlxsw_sp_port, ingress,
+						       proto, tc->cls_flower);
+		case TC_CLSFLOWER_DESTROY:
+			mlxsw_sp_flower_destroy(mlxsw_sp_port, ingress,
+						tc->cls_flower);
+			return 0;
+		default:
+			return -EOPNOTSUPP;
+		}
 	}
 
 	return -EOPNOTSUPP;
@@ -3203,6 +3214,12 @@ static int mlxsw_sp_init(struct mlxsw_core *mlxsw_core,
 		goto err_span_init;
 	}
 
+	err = mlxsw_sp_acl_init(mlxsw_sp);
+	if (err) {
+		dev_err(mlxsw_sp->bus_info->dev, "Failed to initialize ACL\n");
+		goto err_acl_init;
+	}
+
 	err = mlxsw_sp_ports_create(mlxsw_sp);
 	if (err) {
 		dev_err(mlxsw_sp->bus_info->dev, "Failed to create ports\n");
@@ -3212,6 +3229,8 @@ static int mlxsw_sp_init(struct mlxsw_core *mlxsw_core,
 	return 0;
 
 err_ports_create:
+	mlxsw_sp_acl_fini(mlxsw_sp);
+err_acl_init:
 	mlxsw_sp_span_fini(mlxsw_sp);
 err_span_init:
 	mlxsw_sp_router_fini(mlxsw_sp);
@@ -3232,6 +3251,7 @@ static void mlxsw_sp_fini(struct mlxsw_core *mlxsw_core)
 	struct mlxsw_sp *mlxsw_sp = mlxsw_core_driver_priv(mlxsw_core);
 
 	mlxsw_sp_ports_remove(mlxsw_sp);
+	mlxsw_sp_acl_fini(mlxsw_sp);
 	mlxsw_sp_span_fini(mlxsw_sp);
 	mlxsw_sp_router_fini(mlxsw_sp);
 	mlxsw_sp_switchdev_fini(mlxsw_sp);
@@ -3297,7 +3317,7 @@ static struct mlxsw_driver mlxsw_sp_driver = {
 	.profile			= &mlxsw_sp_config_profile,
 };
 
-static bool mlxsw_sp_port_dev_check(const struct net_device *dev)
+bool mlxsw_sp_port_dev_check(const struct net_device *dev)
 {
 	return dev->netdev_ops == &mlxsw_sp_port_netdev_ops;
 }

drivers/net/ethernet/mellanox/mlxsw/spectrum.h

@@ -1,7 +1,7 @@
 /*
  * drivers/net/ethernet/mellanox/mlxsw/spectrum.h
- * Copyright (c) 2015 Mellanox Technologies. All rights reserved.
- * Copyright (c) 2015 Jiri Pirko <jiri@mellanox.com>
+ * Copyright (c) 2015-2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2015-2017 Jiri Pirko <jiri@mellanox.com>
  * Copyright (c) 2015 Ido Schimmel <idosch@mellanox.com>
  * Copyright (c) 2015 Elad Raz <eladr@mellanox.com>
  *
@@ -47,9 +47,12 @@
 #include <linux/in6.h>
 #include <linux/notifier.h>
 #include <net/psample.h>
+#include <net/pkt_cls.h>
 
 #include "port.h"
 #include "core.h"
+#include "core_acl_flex_keys.h"
+#include "core_acl_flex_actions.h"
 
 #define MLXSW_SP_VFID_BASE VLAN_N_VID
 #define MLXSW_SP_VFID_MAX 6656	/* Bridged VLAN interfaces */
@@ -262,6 +265,8 @@ struct mlxsw_sp_router {
 	bool aborted;
 };
 
+struct mlxsw_sp_acl;
+
 struct mlxsw_sp {
 	struct {
 		struct list_head list;
@@ -291,6 +296,7 @@ struct mlxsw_sp {
 	u8 port_to_module[MLXSW_PORT_MAX_PORTS];
 	struct mlxsw_sp_sb sb;
 	struct mlxsw_sp_router router;
+	struct mlxsw_sp_acl *acl;
 	struct {
 		DECLARE_BITMAP(usage, MLXSW_SP_KVD_LINEAR_SIZE);
 	} kvdl;
@@ -373,6 +379,7 @@ struct mlxsw_sp_port {
 	struct mlxsw_sp_port_sample *sample;
 };
 
+bool mlxsw_sp_port_dev_check(const struct net_device *dev);
 struct mlxsw_sp_port *mlxsw_sp_port_lower_dev_hold(struct net_device *dev);
 void mlxsw_sp_port_dev_put(struct mlxsw_sp_port *mlxsw_sp_port);
 
@@ -602,4 +609,99 @@ int mlxsw_sp_router_netevent_event(struct notifier_block *unused,
 int mlxsw_sp_kvdl_alloc(struct mlxsw_sp *mlxsw_sp, unsigned int entry_count);
 void mlxsw_sp_kvdl_free(struct mlxsw_sp *mlxsw_sp, int entry_index);
 
+struct mlxsw_afk *mlxsw_sp_acl_afk(struct mlxsw_sp_acl *acl);
+
+struct mlxsw_sp_acl_rule_info {
+	unsigned int priority;
+	struct mlxsw_afk_element_values values;
+	struct mlxsw_afa_block *act_block;
+};
+
+enum mlxsw_sp_acl_profile {
+	MLXSW_SP_ACL_PROFILE_FLOWER,
+};
+
+struct mlxsw_sp_acl_profile_ops {
+	size_t ruleset_priv_size;
+	int (*ruleset_add)(struct mlxsw_sp *mlxsw_sp,
+			   void *priv, void *ruleset_priv);
+	void (*ruleset_del)(struct mlxsw_sp *mlxsw_sp, void *ruleset_priv);
+	int (*ruleset_bind)(struct mlxsw_sp *mlxsw_sp, void *ruleset_priv,
+			    struct net_device *dev, bool ingress);
+	void (*ruleset_unbind)(struct mlxsw_sp *mlxsw_sp, void *ruleset_priv);
+	size_t rule_priv_size;
+	int (*rule_add)(struct mlxsw_sp *mlxsw_sp,
+			void *ruleset_priv, void *rule_priv,
+			struct mlxsw_sp_acl_rule_info *rulei);
+	void (*rule_del)(struct mlxsw_sp *mlxsw_sp, void *rule_priv);
+};
+
+struct mlxsw_sp_acl_ops {
+	size_t priv_size;
+	int (*init)(struct mlxsw_sp *mlxsw_sp, void *priv);
+	void (*fini)(struct mlxsw_sp *mlxsw_sp, void *priv);
+	const struct mlxsw_sp_acl_profile_ops *
+			(*profile_ops)(struct mlxsw_sp *mlxsw_sp,
+				       enum mlxsw_sp_acl_profile profile);
+};
+
+struct mlxsw_sp_acl_ruleset;
+
+struct mlxsw_sp_acl_ruleset *
+mlxsw_sp_acl_ruleset_get(struct mlxsw_sp *mlxsw_sp,
+			 struct net_device *dev, bool ingress,
+			 enum mlxsw_sp_acl_profile profile);
+void mlxsw_sp_acl_ruleset_put(struct mlxsw_sp *mlxsw_sp,
+			      struct mlxsw_sp_acl_ruleset *ruleset);
+
+struct mlxsw_sp_acl_rule_info *
+mlxsw_sp_acl_rulei_create(struct mlxsw_sp_acl *acl);
+void mlxsw_sp_acl_rulei_destroy(struct mlxsw_sp_acl_rule_info *rulei);
+int mlxsw_sp_acl_rulei_commit(struct mlxsw_sp_acl_rule_info *rulei);
+void mlxsw_sp_acl_rulei_priority(struct mlxsw_sp_acl_rule_info *rulei,
+				 unsigned int priority);
+void mlxsw_sp_acl_rulei_keymask_u32(struct mlxsw_sp_acl_rule_info *rulei,
+				    enum mlxsw_afk_element element,
+				    u32 key_value, u32 mask_value);
+void mlxsw_sp_acl_rulei_keymask_buf(struct mlxsw_sp_acl_rule_info *rulei,
+				    enum mlxsw_afk_element element,
+				    const char *key_value,
+				    const char *mask_value, unsigned int len);
+void mlxsw_sp_acl_rulei_act_continue(struct mlxsw_sp_acl_rule_info *rulei);
+void mlxsw_sp_acl_rulei_act_jump(struct mlxsw_sp_acl_rule_info *rulei,
+				 u16 group_id);
+int mlxsw_sp_acl_rulei_act_drop(struct mlxsw_sp_acl_rule_info *rulei);
+int mlxsw_sp_acl_rulei_act_fwd(struct mlxsw_sp *mlxsw_sp,
+			       struct mlxsw_sp_acl_rule_info *rulei,
+			       struct net_device *out_dev);
+
+struct mlxsw_sp_acl_rule;
+
+struct mlxsw_sp_acl_rule *
+mlxsw_sp_acl_rule_create(struct mlxsw_sp *mlxsw_sp,
+			 struct mlxsw_sp_acl_ruleset *ruleset,
+			 unsigned long cookie);
+void mlxsw_sp_acl_rule_destroy(struct mlxsw_sp *mlxsw_sp,
+			       struct mlxsw_sp_acl_rule *rule);
+int mlxsw_sp_acl_rule_add(struct mlxsw_sp *mlxsw_sp,
+			  struct mlxsw_sp_acl_rule *rule);
+void mlxsw_sp_acl_rule_del(struct mlxsw_sp *mlxsw_sp,
+			   struct mlxsw_sp_acl_rule *rule);
+struct mlxsw_sp_acl_rule *
+mlxsw_sp_acl_rule_lookup(struct mlxsw_sp *mlxsw_sp,
+			 struct mlxsw_sp_acl_ruleset *ruleset,
+			 unsigned long cookie);
+struct mlxsw_sp_acl_rule_info *
+mlxsw_sp_acl_rule_rulei(struct mlxsw_sp_acl_rule *rule);
+
+int mlxsw_sp_acl_init(struct mlxsw_sp *mlxsw_sp);
+void mlxsw_sp_acl_fini(struct mlxsw_sp *mlxsw_sp);
+
+extern const struct mlxsw_sp_acl_ops mlxsw_sp_acl_tcam_ops;
+
+int mlxsw_sp_flower_replace(struct mlxsw_sp_port *mlxsw_sp_port, bool ingress,
+			    __be16 protocol, struct tc_cls_flower_offload *f);
+void mlxsw_sp_flower_destroy(struct mlxsw_sp_port *mlxsw_sp_port, bool ingress,
+			     struct tc_cls_flower_offload *f);
+
 #endif

drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c

@@ -0,0 +1,572 @@
+/*
+ * drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c
+ * Copyright (c) 2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2017 Jiri Pirko <jiri@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/errno.h>
+#include <linux/list.h>
+#include <linux/string.h>
+#include <linux/rhashtable.h>
+#include <linux/netdevice.h>
+
+#include "reg.h"
+#include "core.h"
+#include "resources.h"
+#include "spectrum.h"
+#include "core_acl_flex_keys.h"
+#include "core_acl_flex_actions.h"
+#include "spectrum_acl_flex_keys.h"
+
+struct mlxsw_sp_acl {
+	struct mlxsw_afk *afk;
+	struct mlxsw_afa *afa;
+	const struct mlxsw_sp_acl_ops *ops;
+	struct rhashtable ruleset_ht;
+	unsigned long priv[0];
+	/* priv has to be always the last item */
+};
+
+struct mlxsw_afk *mlxsw_sp_acl_afk(struct mlxsw_sp_acl *acl)
+{
+	return acl->afk;
+}
+
+struct mlxsw_sp_acl_ruleset_ht_key {
+	struct net_device *dev; /* dev this ruleset is bound to */
+	bool ingress;
+	const struct mlxsw_sp_acl_profile_ops *ops;
+};
+
+struct mlxsw_sp_acl_ruleset {
+	struct rhash_head ht_node; /* Member of acl HT */
+	struct mlxsw_sp_acl_ruleset_ht_key ht_key;
+	struct rhashtable rule_ht;
+	unsigned int ref_count;
+	unsigned long priv[0];
+	/* priv has to be always the last item */
+};
+
+struct mlxsw_sp_acl_rule {
+	struct rhash_head ht_node; /* Member of rule HT */
+	unsigned long cookie; /* HT key */
+	struct mlxsw_sp_acl_ruleset *ruleset;
+	struct mlxsw_sp_acl_rule_info *rulei;
+	unsigned long priv[0];
+	/* priv has to be always the last item */
+};
+
+static const struct rhashtable_params mlxsw_sp_acl_ruleset_ht_params = {
+	.key_len = sizeof(struct mlxsw_sp_acl_ruleset_ht_key),
+	.key_offset = offsetof(struct mlxsw_sp_acl_ruleset, ht_key),
+	.head_offset = offsetof(struct mlxsw_sp_acl_ruleset, ht_node),
+	.automatic_shrinking = true,
+};
+
+static const struct rhashtable_params mlxsw_sp_acl_rule_ht_params = {
+	.key_len = sizeof(unsigned long),
+	.key_offset = offsetof(struct mlxsw_sp_acl_rule, cookie),
+	.head_offset = offsetof(struct mlxsw_sp_acl_rule, ht_node),
+	.automatic_shrinking = true,
+};
+
+static struct mlxsw_sp_acl_ruleset *
+mlxsw_sp_acl_ruleset_create(struct mlxsw_sp *mlxsw_sp,
+			    const struct mlxsw_sp_acl_profile_ops *ops)
+{
+	struct mlxsw_sp_acl *acl = mlxsw_sp->acl;
+	struct mlxsw_sp_acl_ruleset *ruleset;
+	size_t alloc_size;
+	int err;
+
+	alloc_size = sizeof(*ruleset) + ops->ruleset_priv_size;
+	ruleset = kzalloc(alloc_size, GFP_KERNEL);
+	if (!ruleset)
+		return ERR_PTR(-ENOMEM);
+	ruleset->ref_count = 1;
+	ruleset->ht_key.ops = ops;
+
+	err = rhashtable_init(&ruleset->rule_ht, &mlxsw_sp_acl_rule_ht_params);
+	if (err)
+		goto err_rhashtable_init;
+
+	err = ops->ruleset_add(mlxsw_sp, acl->priv, ruleset->priv);
+	if (err)
+		goto err_ops_ruleset_add;
+
+	return ruleset;
+
+err_ops_ruleset_add:
+	rhashtable_destroy(&ruleset->rule_ht);
+err_rhashtable_init:
+	kfree(ruleset);
+	return ERR_PTR(err);
+}
+
+static void mlxsw_sp_acl_ruleset_destroy(struct mlxsw_sp *mlxsw_sp,
+					 struct mlxsw_sp_acl_ruleset *ruleset)
+{
+	const struct mlxsw_sp_acl_profile_ops *ops = ruleset->ht_key.ops;
+
+	ops->ruleset_del(mlxsw_sp, ruleset->priv);
+	rhashtable_destroy(&ruleset->rule_ht);
+	kfree(ruleset);
+}
+
+static int mlxsw_sp_acl_ruleset_bind(struct mlxsw_sp *mlxsw_sp,
+				     struct mlxsw_sp_acl_ruleset *ruleset,
+				     struct net_device *dev, bool ingress)
+{
+	const struct mlxsw_sp_acl_profile_ops *ops = ruleset->ht_key.ops;
+	struct mlxsw_sp_acl *acl = mlxsw_sp->acl;
+	int err;
+
+	ruleset->ht_key.dev = dev;
+	ruleset->ht_key.ingress = ingress;
+	err = rhashtable_insert_fast(&acl->ruleset_ht, &ruleset->ht_node,
+				     mlxsw_sp_acl_ruleset_ht_params);
+	if (err)
+		return err;
+	err = ops->ruleset_bind(mlxsw_sp, ruleset->priv, dev, ingress);
+	if (err)
+		goto err_ops_ruleset_bind;
+	return 0;
+
+err_ops_ruleset_bind:
+	rhashtable_remove_fast(&acl->ruleset_ht, &ruleset->ht_node,
+			       mlxsw_sp_acl_ruleset_ht_params);
+	return err;
+}
+
+static void mlxsw_sp_acl_ruleset_unbind(struct mlxsw_sp *mlxsw_sp,
+					struct mlxsw_sp_acl_ruleset *ruleset)
+{
+	const struct mlxsw_sp_acl_profile_ops *ops = ruleset->ht_key.ops;
+	struct mlxsw_sp_acl *acl = mlxsw_sp->acl;
+
+	ops->ruleset_unbind(mlxsw_sp, ruleset->priv);
+	rhashtable_remove_fast(&acl->ruleset_ht, &ruleset->ht_node,
+			       mlxsw_sp_acl_ruleset_ht_params);
+}
+
+static void mlxsw_sp_acl_ruleset_ref_inc(struct mlxsw_sp_acl_ruleset *ruleset)
+{
+	ruleset->ref_count++;
+}
+
+static void mlxsw_sp_acl_ruleset_ref_dec(struct mlxsw_sp *mlxsw_sp,
+					 struct mlxsw_sp_acl_ruleset *ruleset)
+{
+	if (--ruleset->ref_count)
+		return;
+	mlxsw_sp_acl_ruleset_unbind(mlxsw_sp, ruleset);
+	mlxsw_sp_acl_ruleset_destroy(mlxsw_sp, ruleset);
+}
+
+struct mlxsw_sp_acl_ruleset *
+mlxsw_sp_acl_ruleset_get(struct mlxsw_sp *mlxsw_sp,
+			 struct net_device *dev, bool ingress,
+			 enum mlxsw_sp_acl_profile profile)
+{
+	const struct mlxsw_sp_acl_profile_ops *ops;
+	struct mlxsw_sp_acl *acl = mlxsw_sp->acl;
+	struct mlxsw_sp_acl_ruleset_ht_key ht_key;
+	struct mlxsw_sp_acl_ruleset *ruleset;
+	int err;
+
+	ops = acl->ops->profile_ops(mlxsw_sp, profile);
+	if (!ops)
+		return ERR_PTR(-EINVAL);
+
+	memset(&ht_key, 0, sizeof(ht_key));
+	ht_key.dev = dev;
+	ht_key.ingress = ingress;
+	ht_key.ops = ops;
+	ruleset = rhashtable_lookup_fast(&acl->ruleset_ht, &ht_key,
+					 mlxsw_sp_acl_ruleset_ht_params);
+	if (ruleset) {
+		mlxsw_sp_acl_ruleset_ref_inc(ruleset);
+		return ruleset;
+	}
+	ruleset = mlxsw_sp_acl_ruleset_create(mlxsw_sp, ops);
+	if (IS_ERR(ruleset))
+		return ruleset;
+	err = mlxsw_sp_acl_ruleset_bind(mlxsw_sp, ruleset, dev, ingress);
+	if (err)
+		goto err_ruleset_bind;
+	return ruleset;
+
+err_ruleset_bind:
+	mlxsw_sp_acl_ruleset_destroy(mlxsw_sp, ruleset);
+	return ERR_PTR(err);
+}
+
+void mlxsw_sp_acl_ruleset_put(struct mlxsw_sp *mlxsw_sp,
+			      struct mlxsw_sp_acl_ruleset *ruleset)
+{
+	mlxsw_sp_acl_ruleset_ref_dec(mlxsw_sp, ruleset);
+}
+
+struct mlxsw_sp_acl_rule_info *
+mlxsw_sp_acl_rulei_create(struct mlxsw_sp_acl *acl)
+{
+	struct mlxsw_sp_acl_rule_info *rulei;
+	int err;
+
+	rulei = kzalloc(sizeof(*rulei), GFP_KERNEL);
+	if (!rulei)
+		return NULL;
+	rulei->act_block = mlxsw_afa_block_create(acl->afa);
+	if (IS_ERR(rulei->act_block)) {
+		err = PTR_ERR(rulei->act_block);
+		goto err_afa_block_create;
+	}
+	return rulei;
+
+err_afa_block_create:
+	kfree(rulei);
+	return ERR_PTR(err);
+}
+
+void mlxsw_sp_acl_rulei_destroy(struct mlxsw_sp_acl_rule_info *rulei)
+{
+	mlxsw_afa_block_destroy(rulei->act_block);
+	kfree(rulei);
+}
+
+int mlxsw_sp_acl_rulei_commit(struct mlxsw_sp_acl_rule_info *rulei)
+{
+	return mlxsw_afa_block_commit(rulei->act_block);
+}
+
+void mlxsw_sp_acl_rulei_priority(struct mlxsw_sp_acl_rule_info *rulei,
+				 unsigned int priority)
+{
+	rulei->priority = priority;
+}
+
+void mlxsw_sp_acl_rulei_keymask_u32(struct mlxsw_sp_acl_rule_info *rulei,
+				    enum mlxsw_afk_element element,
+				    u32 key_value, u32 mask_value)
+{
+	mlxsw_afk_values_add_u32(&rulei->values, element,
+				 key_value, mask_value);
+}
+
+void mlxsw_sp_acl_rulei_keymask_buf(struct mlxsw_sp_acl_rule_info *rulei,
+				    enum mlxsw_afk_element element,
+				    const char *key_value,
+				    const char *mask_value, unsigned int len)
+{
+	mlxsw_afk_values_add_buf(&rulei->values, element,
+				 key_value, mask_value, len);
+}
+
+void mlxsw_sp_acl_rulei_act_continue(struct mlxsw_sp_acl_rule_info *rulei)
+{
+	mlxsw_afa_block_continue(rulei->act_block);
+}
+
+void mlxsw_sp_acl_rulei_act_jump(struct mlxsw_sp_acl_rule_info *rulei,
+				 u16 group_id)
+{
+	mlxsw_afa_block_jump(rulei->act_block, group_id);
+}
+
+int mlxsw_sp_acl_rulei_act_drop(struct mlxsw_sp_acl_rule_info *rulei)
+{
+	return mlxsw_afa_block_append_drop(rulei->act_block);
+}
+
+int mlxsw_sp_acl_rulei_act_fwd(struct mlxsw_sp *mlxsw_sp,
+			       struct mlxsw_sp_acl_rule_info *rulei,
+			       struct net_device *out_dev)
+{
+	struct mlxsw_sp_port *mlxsw_sp_port;
+	u8 local_port;
+	bool in_port;
+
+	if (out_dev) {
+		if (!mlxsw_sp_port_dev_check(out_dev))
+			return -EINVAL;
+		mlxsw_sp_port = netdev_priv(out_dev);
+		if (mlxsw_sp_port->mlxsw_sp != mlxsw_sp)
+			return -EINVAL;
+		local_port = mlxsw_sp_port->local_port;
+		in_port = false;
+	} else {
+		/* If out_dev is NULL, the called wants to
+		 * set forward to ingress port.
+		 */
+		local_port = 0;
+		in_port = true;
+	}
+	return mlxsw_afa_block_append_fwd(rulei->act_block,
+					  local_port, in_port);
+}
+
+struct mlxsw_sp_acl_rule *
+mlxsw_sp_acl_rule_create(struct mlxsw_sp *mlxsw_sp,
+			 struct mlxsw_sp_acl_ruleset *ruleset,
+			 unsigned long cookie)
+{
+	const struct mlxsw_sp_acl_profile_ops *ops = ruleset->ht_key.ops;
+	struct mlxsw_sp_acl_rule *rule;
+	int err;
+
+	mlxsw_sp_acl_ruleset_ref_inc(ruleset);
+	rule = kzalloc(sizeof(*rule) + ops->rule_priv_size, GFP_KERNEL);
+	if (!rule) {
+		err = -ENOMEM;
+		goto err_alloc;
+	}
+	rule->cookie = cookie;
+	rule->ruleset = ruleset;
+
+	rule->rulei = mlxsw_sp_acl_rulei_create(mlxsw_sp->acl);
+	if (IS_ERR(rule->rulei)) {
+		err = PTR_ERR(rule->rulei);
+		goto err_rulei_create;
+	}
+	return rule;
+
+err_rulei_create:
+	kfree(rule);
+err_alloc:
+	mlxsw_sp_acl_ruleset_ref_dec(mlxsw_sp, ruleset);
+	return ERR_PTR(err);
+}
+
+void mlxsw_sp_acl_rule_destroy(struct mlxsw_sp *mlxsw_sp,
+			       struct mlxsw_sp_acl_rule *rule)
+{
+	struct mlxsw_sp_acl_ruleset *ruleset = rule->ruleset;
+
+	mlxsw_sp_acl_rulei_destroy(rule->rulei);
+	kfree(rule);
+	mlxsw_sp_acl_ruleset_ref_dec(mlxsw_sp, ruleset);
+}
+
+int mlxsw_sp_acl_rule_add(struct mlxsw_sp *mlxsw_sp,
+			  struct mlxsw_sp_acl_rule *rule)
+{
+	struct mlxsw_sp_acl_ruleset *ruleset = rule->ruleset;
+	const struct mlxsw_sp_acl_profile_ops *ops = ruleset->ht_key.ops;
+	int err;
+
+	err = ops->rule_add(mlxsw_sp, ruleset->priv, rule->priv, rule->rulei);
+	if (err)
+		return err;
+
+	err = rhashtable_insert_fast(&ruleset->rule_ht, &rule->ht_node,
+				     mlxsw_sp_acl_rule_ht_params);
+	if (err)
+		goto err_rhashtable_insert;
+
+	return 0;
+
+err_rhashtable_insert:
+	ops->rule_del(mlxsw_sp, rule->priv);
+	return err;
+}
+
+void mlxsw_sp_acl_rule_del(struct mlxsw_sp *mlxsw_sp,
+			   struct mlxsw_sp_acl_rule *rule)
+{
+	struct mlxsw_sp_acl_ruleset *ruleset = rule->ruleset;
+	const struct mlxsw_sp_acl_profile_ops *ops = ruleset->ht_key.ops;
+
+	rhashtable_remove_fast(&ruleset->rule_ht, &rule->ht_node,
+			       mlxsw_sp_acl_rule_ht_params);
+	ops->rule_del(mlxsw_sp, rule->priv);
+}
+
+struct mlxsw_sp_acl_rule *
+mlxsw_sp_acl_rule_lookup(struct mlxsw_sp *mlxsw_sp,
+			 struct mlxsw_sp_acl_ruleset *ruleset,
+			 unsigned long cookie)
+{
+	return rhashtable_lookup_fast(&ruleset->rule_ht, &cookie,
+				       mlxsw_sp_acl_rule_ht_params);
+}
+
+struct mlxsw_sp_acl_rule_info *
+mlxsw_sp_acl_rule_rulei(struct mlxsw_sp_acl_rule *rule)
+{
+	return rule->rulei;
+}
+
+#define MLXSW_SP_KDVL_ACT_EXT_SIZE 1
+
+static int mlxsw_sp_act_kvdl_set_add(void *priv, u32 *p_kvdl_index,
+				     char *enc_actions, bool is_first)
+{
+	struct mlxsw_sp *mlxsw_sp = priv;
+	char pefa_pl[MLXSW_REG_PEFA_LEN];
+	u32 kvdl_index;
+	int ret;
+	int err;
+
+	/* The first action set of a TCAM entry is stored directly in TCAM,
+	 * not KVD linear area.
+	 */
+	if (is_first)
+		return 0;
+
+	ret = mlxsw_sp_kvdl_alloc(mlxsw_sp, MLXSW_SP_KDVL_ACT_EXT_SIZE);
+	if (ret < 0)
+		return ret;
+	kvdl_index = ret;
+	mlxsw_reg_pefa_pack(pefa_pl, kvdl_index, enc_actions);
+	err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(pefa), pefa_pl);
+	if (err)
+		goto err_pefa_write;
+	*p_kvdl_index = kvdl_index;
+	return 0;
+
+err_pefa_write:
+	mlxsw_sp_kvdl_free(mlxsw_sp, kvdl_index);
+	return err;
+}
+
+static void mlxsw_sp_act_kvdl_set_del(void *priv, u32 kvdl_index,
+				      bool is_first)
+{
+	struct mlxsw_sp *mlxsw_sp = priv;
+
+	if (is_first)
+		return;
+	mlxsw_sp_kvdl_free(mlxsw_sp, kvdl_index);
+}
+
+static int mlxsw_sp_act_kvdl_fwd_entry_add(void *priv, u32 *p_kvdl_index,
+					   u8 local_port)
+{
+	struct mlxsw_sp *mlxsw_sp = priv;
+	char ppbs_pl[MLXSW_REG_PPBS_LEN];
+	u32 kvdl_index;
+	int ret;
+	int err;
+
+	ret = mlxsw_sp_kvdl_alloc(mlxsw_sp, 1);
+	if (ret < 0)
+		return ret;
+	kvdl_index = ret;
+	mlxsw_reg_ppbs_pack(ppbs_pl, kvdl_index, local_port);
+	err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(ppbs), ppbs_pl);
+	if (err)
+		goto err_ppbs_write;
+	*p_kvdl_index = kvdl_index;
+	return 0;
+
+err_ppbs_write:
+	mlxsw_sp_kvdl_free(mlxsw_sp, kvdl_index);
+	return err;
+}
+
+static void mlxsw_sp_act_kvdl_fwd_entry_del(void *priv, u32 kvdl_index)
+{
+	struct mlxsw_sp *mlxsw_sp = priv;
+
+	mlxsw_sp_kvdl_free(mlxsw_sp, kvdl_index);
+}
+
+static const struct mlxsw_afa_ops mlxsw_sp_act_afa_ops = {
+	.kvdl_set_add		= mlxsw_sp_act_kvdl_set_add,
+	.kvdl_set_del		= mlxsw_sp_act_kvdl_set_del,
+	.kvdl_fwd_entry_add	= mlxsw_sp_act_kvdl_fwd_entry_add,
+	.kvdl_fwd_entry_del	= mlxsw_sp_act_kvdl_fwd_entry_del,
+};
+
+int mlxsw_sp_acl_init(struct mlxsw_sp *mlxsw_sp)
+{
+	const struct mlxsw_sp_acl_ops *acl_ops = &mlxsw_sp_acl_tcam_ops;
+	struct mlxsw_sp_acl *acl;
+	int err;
+
+	acl = kzalloc(sizeof(*acl) + acl_ops->priv_size, GFP_KERNEL);
+	if (!acl)
+		return -ENOMEM;
+	mlxsw_sp->acl = acl;
+
+	acl->afk = mlxsw_afk_create(MLXSW_CORE_RES_GET(mlxsw_sp->core,
+						       ACL_FLEX_KEYS),
+				    mlxsw_sp_afk_blocks,
+				    MLXSW_SP_AFK_BLOCKS_COUNT);
+	if (!acl->afk) {
+		err = -ENOMEM;
+		goto err_afk_create;
+	}
+
+	acl->afa = mlxsw_afa_create(MLXSW_CORE_RES_GET(mlxsw_sp->core,
+						       ACL_ACTIONS_PER_SET),
+				    &mlxsw_sp_act_afa_ops, mlxsw_sp);
+	if (IS_ERR(acl->afa)) {
+		err = PTR_ERR(acl->afa);
+		goto err_afa_create;
+	}
+
+	err = rhashtable_init(&acl->ruleset_ht,
+			      &mlxsw_sp_acl_ruleset_ht_params);
+	if (err)
+		goto err_rhashtable_init;
+
+	err = acl_ops->init(mlxsw_sp, acl->priv);
+	if (err)
+		goto err_acl_ops_init;
+
+	acl->ops = acl_ops;
+	return 0;
+
+err_acl_ops_init:
+	rhashtable_destroy(&acl->ruleset_ht);
+err_rhashtable_init:
+	mlxsw_afa_destroy(acl->afa);
+err_afa_create:
+	mlxsw_afk_destroy(acl->afk);
+err_afk_create:
+	kfree(acl);
+	return err;
+}
+
+void mlxsw_sp_acl_fini(struct mlxsw_sp *mlxsw_sp)
+{
+	struct mlxsw_sp_acl *acl = mlxsw_sp->acl;
+	const struct mlxsw_sp_acl_ops *acl_ops = acl->ops;
+
+	acl_ops->fini(mlxsw_sp, acl->priv);
+	rhashtable_destroy(&acl->ruleset_ht);
+	mlxsw_afa_destroy(acl->afa);
+	mlxsw_afk_destroy(acl->afk);
+	kfree(acl);
+}

drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_flex_keys.h

@@ -0,0 +1,109 @@
+/*
+ * drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_flex_keys.h
+ * Copyright (c) 2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2017 Jiri Pirko <jiri@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _MLXSW_SPECTRUM_ACL_FLEX_KEYS_H
+#define _MLXSW_SPECTRUM_ACL_FLEX_KEYS_H
+
+#include "core_acl_flex_keys.h"
+
+static struct mlxsw_afk_element_inst mlxsw_sp_afk_element_info_l2_dmac[] = {
+	MLXSW_AFK_ELEMENT_INST_BUF(DMAC, 0x00, 6),
+	MLXSW_AFK_ELEMENT_INST_U32(SRC_SYS_PORT, 0x0C, 0, 16),
+};
+
+static struct mlxsw_afk_element_inst mlxsw_sp_afk_element_info_l2_smac[] = {
+	MLXSW_AFK_ELEMENT_INST_BUF(SMAC, 0x00, 6),
+	MLXSW_AFK_ELEMENT_INST_U32(SRC_SYS_PORT, 0x0C, 0, 16),
+};
+
+static struct mlxsw_afk_element_inst mlxsw_sp_afk_element_info_l2_smac_ex[] = {
+	MLXSW_AFK_ELEMENT_INST_BUF(SMAC, 0x02, 6),
+	MLXSW_AFK_ELEMENT_INST_U32(ETHERTYPE, 0x0C, 0, 16),
+};
+
+static struct mlxsw_afk_element_inst mlxsw_sp_afk_element_info_ipv4_sip[] = {
+	MLXSW_AFK_ELEMENT_INST_U32(SRC_IP4, 0x00, 0, 32),
+	MLXSW_AFK_ELEMENT_INST_U32(IP_PROTO, 0x08, 0, 8),
+	MLXSW_AFK_ELEMENT_INST_U32(SRC_SYS_PORT, 0x0C, 0, 16),
+};
+
+static struct mlxsw_afk_element_inst mlxsw_sp_afk_element_info_ipv4_dip[] = {
+	MLXSW_AFK_ELEMENT_INST_U32(DST_IP4, 0x00, 0, 32),
+	MLXSW_AFK_ELEMENT_INST_U32(IP_PROTO, 0x08, 0, 8),
+	MLXSW_AFK_ELEMENT_INST_U32(SRC_SYS_PORT, 0x0C, 0, 16),
+};
+
+static struct mlxsw_afk_element_inst mlxsw_sp_afk_element_info_ipv4_ex[] = {
+	MLXSW_AFK_ELEMENT_INST_U32(SRC_L4_PORT, 0x08, 0, 16),
+	MLXSW_AFK_ELEMENT_INST_U32(DST_L4_PORT, 0x0C, 0, 16),
+};
+
+static struct mlxsw_afk_element_inst mlxsw_sp_afk_element_info_ipv6_dip[] = {
+	MLXSW_AFK_ELEMENT_INST_BUF(DST_IP6_LO, 0x00, 8),
+};
+
+static struct mlxsw_afk_element_inst mlxsw_sp_afk_element_info_ipv6_ex1[] = {
+	MLXSW_AFK_ELEMENT_INST_BUF(DST_IP6_HI, 0x00, 8),
+	MLXSW_AFK_ELEMENT_INST_U32(IP_PROTO, 0x08, 0, 8),
+};
+
+static struct mlxsw_afk_element_inst mlxsw_sp_afk_element_info_ipv6_sip[] = {
+	MLXSW_AFK_ELEMENT_INST_BUF(SRC_IP6_LO, 0x00, 8),
+};
+
+static struct mlxsw_afk_element_inst mlxsw_sp_afk_element_info_ipv6_sip_ex[] = {
+	MLXSW_AFK_ELEMENT_INST_BUF(SRC_IP6_HI, 0x00, 8),
+};
+
+static struct mlxsw_afk_element_inst mlxsw_sp_afk_element_info_packet_type[] = {
+	MLXSW_AFK_ELEMENT_INST_U32(ETHERTYPE, 0x00, 0, 16),
+};
+
+static const struct mlxsw_afk_block mlxsw_sp_afk_blocks[] = {
+	MLXSW_AFK_BLOCK(0x10, mlxsw_sp_afk_element_info_l2_dmac),
+	MLXSW_AFK_BLOCK(0x11, mlxsw_sp_afk_element_info_l2_smac),
+	MLXSW_AFK_BLOCK(0x12, mlxsw_sp_afk_element_info_l2_smac_ex),
+	MLXSW_AFK_BLOCK(0x30, mlxsw_sp_afk_element_info_ipv4_sip),
+	MLXSW_AFK_BLOCK(0x31, mlxsw_sp_afk_element_info_ipv4_dip),
+	MLXSW_AFK_BLOCK(0x33, mlxsw_sp_afk_element_info_ipv4_ex),
+	MLXSW_AFK_BLOCK(0x60, mlxsw_sp_afk_element_info_ipv6_dip),
+	MLXSW_AFK_BLOCK(0x65, mlxsw_sp_afk_element_info_ipv6_ex1),
+	MLXSW_AFK_BLOCK(0x62, mlxsw_sp_afk_element_info_ipv6_sip),
+	MLXSW_AFK_BLOCK(0x63, mlxsw_sp_afk_element_info_ipv6_sip_ex),
+	MLXSW_AFK_BLOCK(0xB0, mlxsw_sp_afk_element_info_packet_type),
+};
+
+#define MLXSW_SP_AFK_BLOCKS_COUNT ARRAY_SIZE(mlxsw_sp_afk_blocks)
+
+#endif

drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_tcam.c

@@ -0,0 +1,1084 @@
+/*
+ * drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_tcam.c
+ * Copyright (c) 2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2017 Jiri Pirko <jiri@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/errno.h>
+#include <linux/bitops.h>
+#include <linux/list.h>
+#include <linux/rhashtable.h>
+#include <linux/netdevice.h>
+#include <linux/parman.h>
+
+#include "reg.h"
+#include "core.h"
+#include "resources.h"
+#include "spectrum.h"
+#include "core_acl_flex_keys.h"
+
+struct mlxsw_sp_acl_tcam {
+	unsigned long *used_regions; /* bit array */
+	unsigned int max_regions;
+	unsigned long *used_groups;  /* bit array */
+	unsigned int max_groups;
+	unsigned int max_group_size;
+};
+
+static int mlxsw_sp_acl_tcam_init(struct mlxsw_sp *mlxsw_sp, void *priv)
+{
+	struct mlxsw_sp_acl_tcam *tcam = priv;
+	u64 max_tcam_regions;
+	u64 max_regions;
+	u64 max_groups;
+	size_t alloc_size;
+	int err;
+
+	max_tcam_regions = MLXSW_CORE_RES_GET(mlxsw_sp->core,
+					      ACL_MAX_TCAM_REGIONS);
+	max_regions = MLXSW_CORE_RES_GET(mlxsw_sp->core, ACL_MAX_REGIONS);
+
+	/* Use 1:1 mapping between ACL region and TCAM region */
+	if (max_tcam_regions < max_regions)
+		max_regions = max_tcam_regions;
+
+	alloc_size = sizeof(tcam->used_regions[0]) * BITS_TO_LONGS(max_regions);
+	tcam->used_regions = kzalloc(alloc_size, GFP_KERNEL);
+	if (!tcam->used_regions)
+		return -ENOMEM;
+	tcam->max_regions = max_regions;
+
+	max_groups = MLXSW_CORE_RES_GET(mlxsw_sp->core, ACL_MAX_GROUPS);
+	alloc_size = sizeof(tcam->used_groups[0]) * BITS_TO_LONGS(max_groups);
+	tcam->used_groups = kzalloc(alloc_size, GFP_KERNEL);
+	if (!tcam->used_groups) {
+		err = -ENOMEM;
+		goto err_alloc_used_groups;
+	}
+	tcam->max_groups = max_groups;
+	tcam->max_group_size = MLXSW_CORE_RES_GET(mlxsw_sp->core,
+						 ACL_MAX_GROUP_SIZE);
+	return 0;
+
+err_alloc_used_groups:
+	kfree(tcam->used_regions);
+	return err;
+}
+
+static void mlxsw_sp_acl_tcam_fini(struct mlxsw_sp *mlxsw_sp, void *priv)
+{
+	struct mlxsw_sp_acl_tcam *tcam = priv;
+
+	kfree(tcam->used_groups);
+	kfree(tcam->used_regions);
+}
+
+static int mlxsw_sp_acl_tcam_region_id_get(struct mlxsw_sp_acl_tcam *tcam,
+					   u16 *p_id)
+{
+	u16 id;
+
+	id = find_first_zero_bit(tcam->used_regions, tcam->max_regions);
+	if (id < tcam->max_regions) {
+		__set_bit(id, tcam->used_regions);
+		*p_id = id;
+		return 0;
+	}
+	return -ENOBUFS;
+}
+
+static void mlxsw_sp_acl_tcam_region_id_put(struct mlxsw_sp_acl_tcam *tcam,
+					    u16 id)
+{
+	__clear_bit(id, tcam->used_regions);
+}
+
+static int mlxsw_sp_acl_tcam_group_id_get(struct mlxsw_sp_acl_tcam *tcam,
+					  u16 *p_id)
+{
+	u16 id;
+
+	id = find_first_zero_bit(tcam->used_groups, tcam->max_groups);
+	if (id < tcam->max_groups) {
+		__set_bit(id, tcam->used_groups);
+		*p_id = id;
+		return 0;
+	}
+	return -ENOBUFS;
+}
+
+static void mlxsw_sp_acl_tcam_group_id_put(struct mlxsw_sp_acl_tcam *tcam,
+					   u16 id)
+{
+	__clear_bit(id, tcam->used_groups);
+}
+
+struct mlxsw_sp_acl_tcam_pattern {
+	const enum mlxsw_afk_element *elements;
+	unsigned int elements_count;
+};
+
+struct mlxsw_sp_acl_tcam_group {
+	struct mlxsw_sp_acl_tcam *tcam;
+	u16 id;
+	struct list_head region_list;
+	unsigned int region_count;
+	struct rhashtable chunk_ht;
+	struct {
+		u16 local_port;
+		bool ingress;
+	} bound;
+	struct mlxsw_sp_acl_tcam_group_ops *ops;
+	const struct mlxsw_sp_acl_tcam_pattern *patterns;
+	unsigned int patterns_count;
+};
+
+struct mlxsw_sp_acl_tcam_region {
+	struct list_head list; /* Member of a TCAM group */
+	struct list_head chunk_list; /* List of chunks under this region */
+	struct parman *parman;
+	struct mlxsw_sp *mlxsw_sp;
+	struct mlxsw_sp_acl_tcam_group *group;
+	u16 id; /* ACL ID and region ID - they are same */
+	char tcam_region_info[MLXSW_REG_PXXX_TCAM_REGION_INFO_LEN];
+	struct mlxsw_afk_key_info *key_info;
+	struct {
+		struct parman_prio parman_prio;
+		struct parman_item parman_item;
+		struct mlxsw_sp_acl_rule_info *rulei;
+	} catchall;
+};
+
+struct mlxsw_sp_acl_tcam_chunk {
+	struct list_head list; /* Member of a TCAM region */
+	struct rhash_head ht_node; /* Member of a chunk HT */
+	unsigned int priority; /* Priority within the region and group */
+	struct parman_prio parman_prio;
+	struct mlxsw_sp_acl_tcam_group *group;
+	struct mlxsw_sp_acl_tcam_region *region;
+	unsigned int ref_count;
+};
+
+struct mlxsw_sp_acl_tcam_entry {
+	struct parman_item parman_item;
+	struct mlxsw_sp_acl_tcam_chunk *chunk;
+};
+
+static const struct rhashtable_params mlxsw_sp_acl_tcam_chunk_ht_params = {
+	.key_len = sizeof(unsigned int),
+	.key_offset = offsetof(struct mlxsw_sp_acl_tcam_chunk, priority),
+	.head_offset = offsetof(struct mlxsw_sp_acl_tcam_chunk, ht_node),
+	.automatic_shrinking = true,
+};
+
+static int mlxsw_sp_acl_tcam_group_update(struct mlxsw_sp *mlxsw_sp,
+					  struct mlxsw_sp_acl_tcam_group *group)
+{
+	struct mlxsw_sp_acl_tcam_region *region;
+	char pagt_pl[MLXSW_REG_PAGT_LEN];
+	int acl_index = 0;
+
+	mlxsw_reg_pagt_pack(pagt_pl, group->id);
+	list_for_each_entry(region, &group->region_list, list)
+		mlxsw_reg_pagt_acl_id_pack(pagt_pl, acl_index++, region->id);
+	mlxsw_reg_pagt_size_set(pagt_pl, acl_index);
+	return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(pagt), pagt_pl);
+}
+
+static int
+mlxsw_sp_acl_tcam_group_add(struct mlxsw_sp *mlxsw_sp,
+			    struct mlxsw_sp_acl_tcam *tcam,
+			    struct mlxsw_sp_acl_tcam_group *group,
+			    const struct mlxsw_sp_acl_tcam_pattern *patterns,
+			    unsigned int patterns_count)
+{
+	int err;
+
+	group->tcam = tcam;
+	group->patterns = patterns;
+	group->patterns_count = patterns_count;
+	INIT_LIST_HEAD(&group->region_list);
+	err = mlxsw_sp_acl_tcam_group_id_get(tcam, &group->id);
+	if (err)
+		return err;
+
+	err = mlxsw_sp_acl_tcam_group_update(mlxsw_sp, group);
+	if (err)
+		goto err_group_update;
+
+	err = rhashtable_init(&group->chunk_ht,
+			      &mlxsw_sp_acl_tcam_chunk_ht_params);
+	if (err)
+		goto err_rhashtable_init;
+
+	return 0;
+
+err_rhashtable_init:
+err_group_update:
+	mlxsw_sp_acl_tcam_group_id_put(tcam, group->id);
+	return err;
+}
+
+static void mlxsw_sp_acl_tcam_group_del(struct mlxsw_sp *mlxsw_sp,
+					struct mlxsw_sp_acl_tcam_group *group)
+{
+	struct mlxsw_sp_acl_tcam *tcam = group->tcam;
+
+	rhashtable_destroy(&group->chunk_ht);
+	mlxsw_sp_acl_tcam_group_id_put(tcam, group->id);
+	WARN_ON(!list_empty(&group->region_list));
+}
+
+static int
+mlxsw_sp_acl_tcam_group_bind(struct mlxsw_sp *mlxsw_sp,
+			     struct mlxsw_sp_acl_tcam_group *group,
+			     struct net_device *dev, bool ingress)
+{
+	struct mlxsw_sp_port *mlxsw_sp_port;
+	char ppbt_pl[MLXSW_REG_PPBT_LEN];
+
+	if (!mlxsw_sp_port_dev_check(dev))
+		return -EINVAL;
+
+	mlxsw_sp_port = netdev_priv(dev);
+	group->bound.local_port = mlxsw_sp_port->local_port;
+	group->bound.ingress = ingress;
+	mlxsw_reg_ppbt_pack(ppbt_pl,
+			    group->bound.ingress ? MLXSW_REG_PXBT_E_IACL :
+						   MLXSW_REG_PXBT_E_EACL,
+			    MLXSW_REG_PXBT_OP_BIND, group->bound.local_port,
+			    group->id);
+	return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(ppbt), ppbt_pl);
+}
+
+static void
+mlxsw_sp_acl_tcam_group_unbind(struct mlxsw_sp *mlxsw_sp,
+			       struct mlxsw_sp_acl_tcam_group *group)
+{
+	char ppbt_pl[MLXSW_REG_PPBT_LEN];
+
+	mlxsw_reg_ppbt_pack(ppbt_pl,
+			    group->bound.ingress ? MLXSW_REG_PXBT_E_IACL :
+						   MLXSW_REG_PXBT_E_EACL,
+			    MLXSW_REG_PXBT_OP_UNBIND, group->bound.local_port,
+			    group->id);
+	mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(ppbt), ppbt_pl);
+}
+
+static unsigned int
+mlxsw_sp_acl_tcam_region_prio(struct mlxsw_sp_acl_tcam_region *region)
+{
+	struct mlxsw_sp_acl_tcam_chunk *chunk;
+
+	if (list_empty(&region->chunk_list))
+		return 0;
+	/* As a priority of a region, return priority of the first chunk */
+	chunk = list_first_entry(&region->chunk_list, typeof(*chunk), list);
+	return chunk->priority;
+}
+
+static unsigned int
+mlxsw_sp_acl_tcam_region_max_prio(struct mlxsw_sp_acl_tcam_region *region)
+{
+	struct mlxsw_sp_acl_tcam_chunk *chunk;
+
+	if (list_empty(&region->chunk_list))
+		return 0;
+	chunk = list_last_entry(&region->chunk_list, typeof(*chunk), list);
+	return chunk->priority;
+}
+
+static void
+mlxsw_sp_acl_tcam_group_list_add(struct mlxsw_sp_acl_tcam_group *group,
+				 struct mlxsw_sp_acl_tcam_region *region)
+{
+	struct mlxsw_sp_acl_tcam_region *region2;
+	struct list_head *pos;
+
+	/* Position the region inside the list according to priority */
+	list_for_each(pos, &group->region_list) {
+		region2 = list_entry(pos, typeof(*region2), list);
+		if (mlxsw_sp_acl_tcam_region_prio(region2) >
+		    mlxsw_sp_acl_tcam_region_prio(region))
+			break;
+	}
+	list_add_tail(&region->list, pos);
+	group->region_count++;
+}
+
+static void
+mlxsw_sp_acl_tcam_group_list_del(struct mlxsw_sp_acl_tcam_group *group,
+				 struct mlxsw_sp_acl_tcam_region *region)
+{
+	group->region_count--;
+	list_del(&region->list);
+}
+
+static int
+mlxsw_sp_acl_tcam_group_region_attach(struct mlxsw_sp *mlxsw_sp,
+				      struct mlxsw_sp_acl_tcam_group *group,
+				      struct mlxsw_sp_acl_tcam_region *region)
+{
+	int err;
+
+	if (group->region_count == group->tcam->max_group_size)
+		return -ENOBUFS;
+
+	mlxsw_sp_acl_tcam_group_list_add(group, region);
+
+	err = mlxsw_sp_acl_tcam_group_update(mlxsw_sp, group);
+	if (err)
+		goto err_group_update;
+	region->group = group;
+
+	return 0;
+
+err_group_update:
+	mlxsw_sp_acl_tcam_group_list_del(group, region);
+	mlxsw_sp_acl_tcam_group_update(mlxsw_sp, group);
+	return err;
+}
+
+static void
+mlxsw_sp_acl_tcam_group_region_detach(struct mlxsw_sp *mlxsw_sp,
+				      struct mlxsw_sp_acl_tcam_region *region)
+{
+	struct mlxsw_sp_acl_tcam_group *group = region->group;
+
+	mlxsw_sp_acl_tcam_group_list_del(group, region);
+	mlxsw_sp_acl_tcam_group_update(mlxsw_sp, group);
+}
+
+static struct mlxsw_sp_acl_tcam_region *
+mlxsw_sp_acl_tcam_group_region_find(struct mlxsw_sp_acl_tcam_group *group,
+				    unsigned int priority,
+				    struct mlxsw_afk_element_usage *elusage,
+				    bool *p_need_split)
+{
+	struct mlxsw_sp_acl_tcam_region *region, *region2;
+	struct list_head *pos;
+	bool issubset;
+
+	list_for_each(pos, &group->region_list) {
+		region = list_entry(pos, typeof(*region), list);
+
+		/* First, check if the requested priority does not rather belong
+		 * under some of the next regions.
+		 */
+		if (pos->next != &group->region_list) { /* not last */
+			region2 = list_entry(pos->next, typeof(*region2), list);
+			if (priority >= mlxsw_sp_acl_tcam_region_prio(region2))
+				continue;
+		}
+
+		issubset = mlxsw_afk_key_info_subset(region->key_info, elusage);
+
+		/* If requested element usage would not fit and the priority
+		 * is lower than the currently inspected region we cannot
+		 * use this region, so return NULL to indicate new region has
+		 * to be created.
+		 */
+		if (!issubset &&
+		    priority < mlxsw_sp_acl_tcam_region_prio(region))
+			return NULL;
+
+		/* If requested element usage would not fit and the priority
+		 * is higher than the currently inspected region we cannot
+		 * use this region. There is still some hope that the next
+		 * region would be the fit. So let it be processed and
+		 * eventually break at the check right above this.
+		 */
+		if (!issubset &&
+		    priority > mlxsw_sp_acl_tcam_region_max_prio(region))
+			continue;
+
+		/* Indicate if the region needs to be split in order to add
+		 * the requested priority. Split is needed when requested
+		 * element usage won't fit into the found region.
+		 */
+		*p_need_split = !issubset;
+		return region;
+	}
+	return NULL; /* New region has to be created. */
+}
+
+static void
+mlxsw_sp_acl_tcam_group_use_patterns(struct mlxsw_sp_acl_tcam_group *group,
+				     struct mlxsw_afk_element_usage *elusage,
+				     struct mlxsw_afk_element_usage *out)
+{
+	const struct mlxsw_sp_acl_tcam_pattern *pattern;
+	int i;
+
+	for (i = 0; i < group->patterns_count; i++) {
+		pattern = &group->patterns[i];
+		mlxsw_afk_element_usage_fill(out, pattern->elements,
+					     pattern->elements_count);
+		if (mlxsw_afk_element_usage_subset(elusage, out))
+			return;
+	}
+	memcpy(out, elusage, sizeof(*out));
+}
+
+#define MLXSW_SP_ACL_TCAM_REGION_BASE_COUNT 16
+#define MLXSW_SP_ACL_TCAM_REGION_RESIZE_STEP 16
+
+static int
+mlxsw_sp_acl_tcam_region_alloc(struct mlxsw_sp *mlxsw_sp,
+			       struct mlxsw_sp_acl_tcam_region *region)
+{
+	struct mlxsw_afk_key_info *key_info = region->key_info;
+	char ptar_pl[MLXSW_REG_PTAR_LEN];
+	unsigned int encodings_count;
+	int i;
+	int err;
+
+	mlxsw_reg_ptar_pack(ptar_pl, MLXSW_REG_PTAR_OP_ALLOC,
+			    MLXSW_SP_ACL_TCAM_REGION_BASE_COUNT,
+			    region->id, region->tcam_region_info);
+	encodings_count = mlxsw_afk_key_info_blocks_count_get(key_info);
+	for (i = 0; i < encodings_count; i++) {
+		u16 encoding;
+
+		encoding = mlxsw_afk_key_info_block_encoding_get(key_info, i);
+		mlxsw_reg_ptar_key_id_pack(ptar_pl, i, encoding);
+	}
+	err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(ptar), ptar_pl);
+	if (err)
+		return err;
+	mlxsw_reg_ptar_unpack(ptar_pl, region->tcam_region_info);
+	return 0;
+}
+
+static void
+mlxsw_sp_acl_tcam_region_free(struct mlxsw_sp *mlxsw_sp,
+			      struct mlxsw_sp_acl_tcam_region *region)
+{
+	char ptar_pl[MLXSW_REG_PTAR_LEN];
+
+	mlxsw_reg_ptar_pack(ptar_pl, MLXSW_REG_PTAR_OP_FREE, 0, region->id,
+			    region->tcam_region_info);
+	mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(ptar), ptar_pl);
+}
+
+static int
+mlxsw_sp_acl_tcam_region_resize(struct mlxsw_sp *mlxsw_sp,
+				struct mlxsw_sp_acl_tcam_region *region,
+				u16 new_size)
+{
+	char ptar_pl[MLXSW_REG_PTAR_LEN];
+
+	mlxsw_reg_ptar_pack(ptar_pl, MLXSW_REG_PTAR_OP_RESIZE,
+			    new_size, region->id, region->tcam_region_info);
+	return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(ptar), ptar_pl);
+}
+
+static int
+mlxsw_sp_acl_tcam_region_enable(struct mlxsw_sp *mlxsw_sp,
+				struct mlxsw_sp_acl_tcam_region *region)
+{
+	char pacl_pl[MLXSW_REG_PACL_LEN];
+
+	mlxsw_reg_pacl_pack(pacl_pl, region->id, true,
+			    region->tcam_region_info);
+	return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(pacl), pacl_pl);
+}
+
+static void
+mlxsw_sp_acl_tcam_region_disable(struct mlxsw_sp *mlxsw_sp,
+				 struct mlxsw_sp_acl_tcam_region *region)
+{
+	char pacl_pl[MLXSW_REG_PACL_LEN];
+
+	mlxsw_reg_pacl_pack(pacl_pl, region->id, false,
+			    region->tcam_region_info);
+	mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(pacl), pacl_pl);
+}
+
+static int
+mlxsw_sp_acl_tcam_region_entry_insert(struct mlxsw_sp *mlxsw_sp,
+				      struct mlxsw_sp_acl_tcam_region *region,
+				      unsigned int offset,
+				      struct mlxsw_sp_acl_rule_info *rulei)
+{
+	char ptce2_pl[MLXSW_REG_PTCE2_LEN];
+	char *act_set;
+	char *mask;
+	char *key;
+
+	mlxsw_reg_ptce2_pack(ptce2_pl, true, MLXSW_REG_PTCE2_OP_WRITE_WRITE,
+			     region->tcam_region_info, offset);
+	key = mlxsw_reg_ptce2_flex_key_blocks_data(ptce2_pl);
+	mask = mlxsw_reg_ptce2_mask_data(ptce2_pl);
+	mlxsw_afk_encode(region->key_info, &rulei->values, key, mask);
+
+	/* Only the first action set belongs here, the rest is in KVD */
+	act_set = mlxsw_afa_block_first_set(rulei->act_block);
+	mlxsw_reg_ptce2_flex_action_set_memcpy_to(ptce2_pl, act_set);
+
+	return mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(ptce2), ptce2_pl);
+}
+
+static void
+mlxsw_sp_acl_tcam_region_entry_remove(struct mlxsw_sp *mlxsw_sp,
+				      struct mlxsw_sp_acl_tcam_region *region,
+				      unsigned int offset)
+{
+	char ptce2_pl[MLXSW_REG_PTCE2_LEN];
+
+	mlxsw_reg_ptce2_pack(ptce2_pl, false, MLXSW_REG_PTCE2_OP_WRITE_WRITE,
+			     region->tcam_region_info, offset);
+	mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(ptce2), ptce2_pl);
+}
+
+#define MLXSW_SP_ACL_TCAM_CATCHALL_PRIO (-1UL)
+
+static int
+mlxsw_sp_acl_tcam_region_catchall_add(struct mlxsw_sp *mlxsw_sp,
+				      struct mlxsw_sp_acl_tcam_region *region)
+{
+	struct parman_prio *parman_prio = &region->catchall.parman_prio;
+	struct parman_item *parman_item = &region->catchall.parman_item;
+	struct mlxsw_sp_acl_rule_info *rulei;
+	int err;
+
+	parman_prio_init(region->parman, parman_prio,
+			 MLXSW_SP_ACL_TCAM_CATCHALL_PRIO);
+	err = parman_item_add(region->parman, parman_prio, parman_item);
+	if (err)
+		goto err_parman_item_add;
+
+	rulei = mlxsw_sp_acl_rulei_create(mlxsw_sp->acl);
+	if (IS_ERR(rulei)) {
+		err = PTR_ERR(rulei);
+		goto err_rulei_create;
+	}
+
+	mlxsw_sp_acl_rulei_act_continue(rulei);
+	err = mlxsw_sp_acl_rulei_commit(rulei);
+	if (err)
+		goto err_rulei_commit;
+
+	err = mlxsw_sp_acl_tcam_region_entry_insert(mlxsw_sp, region,
+						    parman_item->index, rulei);
+	region->catchall.rulei = rulei;
+	if (err)
+		goto err_rule_insert;
+
+	return 0;
+
+err_rule_insert:
+err_rulei_commit:
+	mlxsw_sp_acl_rulei_destroy(rulei);
+err_rulei_create:
+	parman_item_remove(region->parman, parman_prio, parman_item);
+err_parman_item_add:
+	parman_prio_fini(parman_prio);
+	return err;
+}
+
+static void
+mlxsw_sp_acl_tcam_region_catchall_del(struct mlxsw_sp *mlxsw_sp,
+				      struct mlxsw_sp_acl_tcam_region *region)
+{
+	struct parman_prio *parman_prio = &region->catchall.parman_prio;
+	struct parman_item *parman_item = &region->catchall.parman_item;
+	struct mlxsw_sp_acl_rule_info *rulei = region->catchall.rulei;
+
+	mlxsw_sp_acl_tcam_region_entry_remove(mlxsw_sp, region,
+					      parman_item->index);
+	mlxsw_sp_acl_rulei_destroy(rulei);
+	parman_item_remove(region->parman, parman_prio, parman_item);
+	parman_prio_fini(parman_prio);
+}
+
+static void
+mlxsw_sp_acl_tcam_region_move(struct mlxsw_sp *mlxsw_sp,
+			      struct mlxsw_sp_acl_tcam_region *region,
+			      u16 src_offset, u16 dst_offset, u16 size)
+{
+	char prcr_pl[MLXSW_REG_PRCR_LEN];
+
+	mlxsw_reg_prcr_pack(prcr_pl, MLXSW_REG_PRCR_OP_MOVE,
+			    region->tcam_region_info, src_offset,
+			    region->tcam_region_info, dst_offset, size);
+	mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(prcr), prcr_pl);
+}
+
+static int mlxsw_sp_acl_tcam_region_parman_resize(void *priv,
+						  unsigned long new_count)
+{
+	struct mlxsw_sp_acl_tcam_region *region = priv;
+	struct mlxsw_sp *mlxsw_sp = region->mlxsw_sp;
+	u64 max_tcam_rules;
+
+	max_tcam_rules = MLXSW_CORE_RES_GET(mlxsw_sp->core, ACL_MAX_TCAM_RULES);
+	if (new_count > max_tcam_rules)
+		return -EINVAL;
+	return mlxsw_sp_acl_tcam_region_resize(mlxsw_sp, region, new_count);
+}
+
+static void mlxsw_sp_acl_tcam_region_parman_move(void *priv,
+						 unsigned long from_index,
+						 unsigned long to_index,
+						 unsigned long count)
+{
+	struct mlxsw_sp_acl_tcam_region *region = priv;
+	struct mlxsw_sp *mlxsw_sp = region->mlxsw_sp;
+
+	mlxsw_sp_acl_tcam_region_move(mlxsw_sp, region,
+				      from_index, to_index, count);
+}
+
+static const struct parman_ops mlxsw_sp_acl_tcam_region_parman_ops = {
+	.base_count	= MLXSW_SP_ACL_TCAM_REGION_BASE_COUNT,
+	.resize_step	= MLXSW_SP_ACL_TCAM_REGION_RESIZE_STEP,
+	.resize		= mlxsw_sp_acl_tcam_region_parman_resize,
+	.move		= mlxsw_sp_acl_tcam_region_parman_move,
+	.algo		= PARMAN_ALGO_TYPE_LSORT,
+};
+
+static struct mlxsw_sp_acl_tcam_region *
+mlxsw_sp_acl_tcam_region_create(struct mlxsw_sp *mlxsw_sp,
+				struct mlxsw_sp_acl_tcam *tcam,
+				struct mlxsw_afk_element_usage *elusage)
+{
+	struct mlxsw_afk *afk = mlxsw_sp_acl_afk(mlxsw_sp->acl);
+	struct mlxsw_sp_acl_tcam_region *region;
+	int err;
+
+	region = kzalloc(sizeof(*region), GFP_KERNEL);
+	if (!region)
+		return ERR_PTR(-ENOMEM);
+	INIT_LIST_HEAD(&region->chunk_list);
+	region->mlxsw_sp = mlxsw_sp;
+
+	region->parman = parman_create(&mlxsw_sp_acl_tcam_region_parman_ops,
+				       region);
+	if (!region->parman) {
+		err = -ENOMEM;
+		goto err_parman_create;
+	}
+
+	region->key_info = mlxsw_afk_key_info_get(afk, elusage);
+	if (IS_ERR(region->key_info)) {
+		err = PTR_ERR(region->key_info);
+		goto err_key_info_get;
+	}
+
+	err = mlxsw_sp_acl_tcam_region_id_get(tcam, &region->id);
+	if (err)
+		goto err_region_id_get;
+
+	err = mlxsw_sp_acl_tcam_region_alloc(mlxsw_sp, region);
+	if (err)
+		goto err_tcam_region_alloc;
+
+	err = mlxsw_sp_acl_tcam_region_enable(mlxsw_sp, region);
+	if (err)
+		goto err_tcam_region_enable;
+
+	err = mlxsw_sp_acl_tcam_region_catchall_add(mlxsw_sp, region);
+	if (err)
+		goto err_tcam_region_catchall_add;
+
+	return region;
+
+err_tcam_region_catchall_add:
+	mlxsw_sp_acl_tcam_region_disable(mlxsw_sp, region);
+err_tcam_region_enable:
+	mlxsw_sp_acl_tcam_region_free(mlxsw_sp, region);
+err_tcam_region_alloc:
+	mlxsw_sp_acl_tcam_region_id_put(tcam, region->id);
+err_region_id_get:
+	mlxsw_afk_key_info_put(region->key_info);
+err_key_info_get:
+	parman_destroy(region->parman);
+err_parman_create:
+	kfree(region);
+	return ERR_PTR(err);
+}
+
+static void
+mlxsw_sp_acl_tcam_region_destroy(struct mlxsw_sp *mlxsw_sp,
+				 struct mlxsw_sp_acl_tcam_region *region)
+{
+	mlxsw_sp_acl_tcam_region_catchall_del(mlxsw_sp, region);
+	mlxsw_sp_acl_tcam_region_disable(mlxsw_sp, region);
+	mlxsw_sp_acl_tcam_region_free(mlxsw_sp, region);
+	mlxsw_sp_acl_tcam_region_id_put(region->group->tcam, region->id);
+	mlxsw_afk_key_info_put(region->key_info);
+	parman_destroy(region->parman);
+	kfree(region);
+}
+
+static int
+mlxsw_sp_acl_tcam_chunk_assoc(struct mlxsw_sp *mlxsw_sp,
+			      struct mlxsw_sp_acl_tcam_group *group,
+			      unsigned int priority,
+			      struct mlxsw_afk_element_usage *elusage,
+			      struct mlxsw_sp_acl_tcam_chunk *chunk)
+{
+	struct mlxsw_sp_acl_tcam_region *region;
+	bool region_created = false;
+	bool need_split;
+	int err;
+
+	region = mlxsw_sp_acl_tcam_group_region_find(group, priority, elusage,
+						     &need_split);
+	if (region && need_split) {
+		/* According to priority, the chunk should belong to an
+		 * existing region. However, this chunk needs elements
+		 * that region does not contain. We need to split the existing
+		 * region into two and create a new region for this chunk
+		 * in between. This is not supported now.
+		 */
+		return -EOPNOTSUPP;
+	}
+	if (!region) {
+		struct mlxsw_afk_element_usage region_elusage;
+
+		mlxsw_sp_acl_tcam_group_use_patterns(group, elusage,
+						     &region_elusage);
+		region = mlxsw_sp_acl_tcam_region_create(mlxsw_sp, group->tcam,
+							 &region_elusage);
+		if (IS_ERR(region))
+			return PTR_ERR(region);
+		region_created = true;
+	}
+
+	chunk->region = region;
+	list_add_tail(&chunk->list, &region->chunk_list);
+
+	if (!region_created)
+		return 0;
+
+	err = mlxsw_sp_acl_tcam_group_region_attach(mlxsw_sp, group, region);
+	if (err)
+		goto err_group_region_attach;
+
+	return 0;
+
+err_group_region_attach:
+	mlxsw_sp_acl_tcam_region_destroy(mlxsw_sp, region);
+	return err;
+}
+
+static void
+mlxsw_sp_acl_tcam_chunk_deassoc(struct mlxsw_sp *mlxsw_sp,
+				struct mlxsw_sp_acl_tcam_chunk *chunk)
+{
+	struct mlxsw_sp_acl_tcam_region *region = chunk->region;
+
+	list_del(&chunk->list);
+	if (list_empty(&region->chunk_list)) {
+		mlxsw_sp_acl_tcam_group_region_detach(mlxsw_sp, region);
+		mlxsw_sp_acl_tcam_region_destroy(mlxsw_sp, region);
+	}
+}
+
+static struct mlxsw_sp_acl_tcam_chunk *
+mlxsw_sp_acl_tcam_chunk_create(struct mlxsw_sp *mlxsw_sp,
+			       struct mlxsw_sp_acl_tcam_group *group,
+			       unsigned int priority,
+			       struct mlxsw_afk_element_usage *elusage)
+{
+	struct mlxsw_sp_acl_tcam_chunk *chunk;
+	int err;
+
+	if (priority == MLXSW_SP_ACL_TCAM_CATCHALL_PRIO)
+		return ERR_PTR(-EINVAL);
+
+	chunk = kzalloc(sizeof(*chunk), GFP_KERNEL);
+	if (!chunk)
+		return ERR_PTR(-ENOMEM);
+	chunk->priority = priority;
+	chunk->group = group;
+	chunk->ref_count = 1;
+
+	err = mlxsw_sp_acl_tcam_chunk_assoc(mlxsw_sp, group, priority,
+					    elusage, chunk);
+	if (err)
+		goto err_chunk_assoc;
+
+	parman_prio_init(chunk->region->parman, &chunk->parman_prio, priority);
+
+	err = rhashtable_insert_fast(&group->chunk_ht, &chunk->ht_node,
+				     mlxsw_sp_acl_tcam_chunk_ht_params);
+	if (err)
+		goto err_rhashtable_insert;
+
+	return chunk;
+
+err_rhashtable_insert:
+	parman_prio_fini(&chunk->parman_prio);
+	mlxsw_sp_acl_tcam_chunk_deassoc(mlxsw_sp, chunk);
+err_chunk_assoc:
+	kfree(chunk);
+	return ERR_PTR(err);
+}
+
+static void
+mlxsw_sp_acl_tcam_chunk_destroy(struct mlxsw_sp *mlxsw_sp,
+				struct mlxsw_sp_acl_tcam_chunk *chunk)
+{
+	struct mlxsw_sp_acl_tcam_group *group = chunk->group;
+
+	rhashtable_remove_fast(&group->chunk_ht, &chunk->ht_node,
+			       mlxsw_sp_acl_tcam_chunk_ht_params);
+	parman_prio_fini(&chunk->parman_prio);
+	mlxsw_sp_acl_tcam_chunk_deassoc(mlxsw_sp, chunk);
+	kfree(chunk);
+}
+
+static struct mlxsw_sp_acl_tcam_chunk *
+mlxsw_sp_acl_tcam_chunk_get(struct mlxsw_sp *mlxsw_sp,
+			    struct mlxsw_sp_acl_tcam_group *group,
+			    unsigned int priority,
+			    struct mlxsw_afk_element_usage *elusage)
+{
+	struct mlxsw_sp_acl_tcam_chunk *chunk;
+
+	chunk = rhashtable_lookup_fast(&group->chunk_ht, &priority,
+				       mlxsw_sp_acl_tcam_chunk_ht_params);
+	if (chunk) {
+		if (WARN_ON(!mlxsw_afk_key_info_subset(chunk->region->key_info,
+						       elusage)))
+			return ERR_PTR(-EINVAL);
+		chunk->ref_count++;
+		return chunk;
+	}
+	return mlxsw_sp_acl_tcam_chunk_create(mlxsw_sp, group,
+					      priority, elusage);
+}
+
+static void mlxsw_sp_acl_tcam_chunk_put(struct mlxsw_sp *mlxsw_sp,
+					struct mlxsw_sp_acl_tcam_chunk *chunk)
+{
+	if (--chunk->ref_count)
+		return;
+	mlxsw_sp_acl_tcam_chunk_destroy(mlxsw_sp, chunk);
+}
+
+static int mlxsw_sp_acl_tcam_entry_add(struct mlxsw_sp *mlxsw_sp,
+				       struct mlxsw_sp_acl_tcam_group *group,
+				       struct mlxsw_sp_acl_tcam_entry *entry,
+				       struct mlxsw_sp_acl_rule_info *rulei)
+{
+	struct mlxsw_sp_acl_tcam_chunk *chunk;
+	struct mlxsw_sp_acl_tcam_region *region;
+	int err;
+
+	chunk = mlxsw_sp_acl_tcam_chunk_get(mlxsw_sp, group, rulei->priority,
+					    &rulei->values.elusage);
+	if (IS_ERR(chunk))
+		return PTR_ERR(chunk);
+
+	region = chunk->region;
+	err = parman_item_add(region->parman, &chunk->parman_prio,
+			      &entry->parman_item);
+	if (err)
+		goto err_parman_item_add;
+
+	err = mlxsw_sp_acl_tcam_region_entry_insert(mlxsw_sp, region,
+						    entry->parman_item.index,
+						    rulei);
+	if (err)
+		goto err_rule_insert;
+	entry->chunk = chunk;
+
+	return 0;
+
+err_rule_insert:
+	parman_item_remove(region->parman, &chunk->parman_prio,
+			   &entry->parman_item);
+err_parman_item_add:
+	mlxsw_sp_acl_tcam_chunk_put(mlxsw_sp, chunk);
+	return err;
+}
+
+static void mlxsw_sp_acl_tcam_entry_del(struct mlxsw_sp *mlxsw_sp,
+					struct mlxsw_sp_acl_tcam_entry *entry)
+{
+	struct mlxsw_sp_acl_tcam_chunk *chunk = entry->chunk;
+	struct mlxsw_sp_acl_tcam_region *region = chunk->region;
+
+	mlxsw_sp_acl_tcam_region_entry_remove(mlxsw_sp, region,
+					      entry->parman_item.index);
+	parman_item_remove(region->parman, &chunk->parman_prio,
+			   &entry->parman_item);
+	mlxsw_sp_acl_tcam_chunk_put(mlxsw_sp, chunk);
+}
+
+static const enum mlxsw_afk_element mlxsw_sp_acl_tcam_pattern_ipv4[] = {
+	MLXSW_AFK_ELEMENT_SRC_SYS_PORT,
+	MLXSW_AFK_ELEMENT_DMAC,
+	MLXSW_AFK_ELEMENT_SMAC,
+	MLXSW_AFK_ELEMENT_ETHERTYPE,
+	MLXSW_AFK_ELEMENT_IP_PROTO,
+	MLXSW_AFK_ELEMENT_SRC_IP4,
+	MLXSW_AFK_ELEMENT_DST_IP4,
+	MLXSW_AFK_ELEMENT_DST_L4_PORT,
+	MLXSW_AFK_ELEMENT_SRC_L4_PORT,
+};
+
+static const enum mlxsw_afk_element mlxsw_sp_acl_tcam_pattern_ipv6[] = {
+	MLXSW_AFK_ELEMENT_ETHERTYPE,
+	MLXSW_AFK_ELEMENT_IP_PROTO,
+	MLXSW_AFK_ELEMENT_SRC_IP6_HI,
+	MLXSW_AFK_ELEMENT_SRC_IP6_LO,
+	MLXSW_AFK_ELEMENT_DST_IP6_HI,
+	MLXSW_AFK_ELEMENT_DST_IP6_LO,
+	MLXSW_AFK_ELEMENT_DST_L4_PORT,
+	MLXSW_AFK_ELEMENT_SRC_L4_PORT,
+};
+
+static const struct mlxsw_sp_acl_tcam_pattern mlxsw_sp_acl_tcam_patterns[] = {
+	{
+		.elements = mlxsw_sp_acl_tcam_pattern_ipv4,
+		.elements_count = ARRAY_SIZE(mlxsw_sp_acl_tcam_pattern_ipv4),
+	},
+	{
+		.elements = mlxsw_sp_acl_tcam_pattern_ipv6,
+		.elements_count = ARRAY_SIZE(mlxsw_sp_acl_tcam_pattern_ipv6),
+	},
+};
+
+#define MLXSW_SP_ACL_TCAM_PATTERNS_COUNT \
+	ARRAY_SIZE(mlxsw_sp_acl_tcam_patterns)
+
+struct mlxsw_sp_acl_tcam_flower_ruleset {
+	struct mlxsw_sp_acl_tcam_group group;
+};
+
+struct mlxsw_sp_acl_tcam_flower_rule {
+	struct mlxsw_sp_acl_tcam_entry entry;
+};
+
+static int
+mlxsw_sp_acl_tcam_flower_ruleset_add(struct mlxsw_sp *mlxsw_sp,
+				     void *priv, void *ruleset_priv)
+{
+	struct mlxsw_sp_acl_tcam_flower_ruleset *ruleset = ruleset_priv;
+	struct mlxsw_sp_acl_tcam *tcam = priv;
+
+	return mlxsw_sp_acl_tcam_group_add(mlxsw_sp, tcam, &ruleset->group,
+					   mlxsw_sp_acl_tcam_patterns,
+					   MLXSW_SP_ACL_TCAM_PATTERNS_COUNT);
+}
+
+static void
+mlxsw_sp_acl_tcam_flower_ruleset_del(struct mlxsw_sp *mlxsw_sp,
+				     void *ruleset_priv)
+{
+	struct mlxsw_sp_acl_tcam_flower_ruleset *ruleset = ruleset_priv;
+
+	mlxsw_sp_acl_tcam_group_del(mlxsw_sp, &ruleset->group);
+}
+
+static int
+mlxsw_sp_acl_tcam_flower_ruleset_bind(struct mlxsw_sp *mlxsw_sp,
+				      void *ruleset_priv,
+				      struct net_device *dev, bool ingress)
+{
+	struct mlxsw_sp_acl_tcam_flower_ruleset *ruleset = ruleset_priv;
+
+	return mlxsw_sp_acl_tcam_group_bind(mlxsw_sp, &ruleset->group,
+					    dev, ingress);
+}
+
+static void
+mlxsw_sp_acl_tcam_flower_ruleset_unbind(struct mlxsw_sp *mlxsw_sp,
+					void *ruleset_priv)
+{
+	struct mlxsw_sp_acl_tcam_flower_ruleset *ruleset = ruleset_priv;
+
+	mlxsw_sp_acl_tcam_group_unbind(mlxsw_sp, &ruleset->group);
+}
+
+static int
+mlxsw_sp_acl_tcam_flower_rule_add(struct mlxsw_sp *mlxsw_sp,
+				  void *ruleset_priv, void *rule_priv,
+				  struct mlxsw_sp_acl_rule_info *rulei)
+{
+	struct mlxsw_sp_acl_tcam_flower_ruleset *ruleset = ruleset_priv;
+	struct mlxsw_sp_acl_tcam_flower_rule *rule = rule_priv;
+
+	return mlxsw_sp_acl_tcam_entry_add(mlxsw_sp, &ruleset->group,
+					   &rule->entry, rulei);
+}
+
+static void
+mlxsw_sp_acl_tcam_flower_rule_del(struct mlxsw_sp *mlxsw_sp, void *rule_priv)
+{
+	struct mlxsw_sp_acl_tcam_flower_rule *rule = rule_priv;
+
+	mlxsw_sp_acl_tcam_entry_del(mlxsw_sp, &rule->entry);
+}
+
+static const struct mlxsw_sp_acl_profile_ops mlxsw_sp_acl_tcam_flower_ops = {
+	.ruleset_priv_size	= sizeof(struct mlxsw_sp_acl_tcam_flower_ruleset),
+	.ruleset_add		= mlxsw_sp_acl_tcam_flower_ruleset_add,
+	.ruleset_del		= mlxsw_sp_acl_tcam_flower_ruleset_del,
+	.ruleset_bind		= mlxsw_sp_acl_tcam_flower_ruleset_bind,
+	.ruleset_unbind		= mlxsw_sp_acl_tcam_flower_ruleset_unbind,
+	.rule_priv_size		= sizeof(struct mlxsw_sp_acl_tcam_flower_rule),
+	.rule_add		= mlxsw_sp_acl_tcam_flower_rule_add,
+	.rule_del		= mlxsw_sp_acl_tcam_flower_rule_del,
+};
+
+static const struct mlxsw_sp_acl_profile_ops *
+mlxsw_sp_acl_tcam_profile_ops_arr[] = {
+	[MLXSW_SP_ACL_PROFILE_FLOWER] = &mlxsw_sp_acl_tcam_flower_ops,
+};
+
+static const struct mlxsw_sp_acl_profile_ops *
+mlxsw_sp_acl_tcam_profile_ops(struct mlxsw_sp *mlxsw_sp,
+			      enum mlxsw_sp_acl_profile profile)
+{
+	const struct mlxsw_sp_acl_profile_ops *ops;
+
+	if (WARN_ON(profile >= ARRAY_SIZE(mlxsw_sp_acl_tcam_profile_ops_arr)))
+		return NULL;
+	ops = mlxsw_sp_acl_tcam_profile_ops_arr[profile];
+	if (WARN_ON(!ops))
+		return NULL;
+	return ops;
+}
+
+const struct mlxsw_sp_acl_ops mlxsw_sp_acl_tcam_ops = {
+	.priv_size		= sizeof(struct mlxsw_sp_acl_tcam),
+	.init			= mlxsw_sp_acl_tcam_init,
+	.fini			= mlxsw_sp_acl_tcam_fini,
+	.profile_ops		= mlxsw_sp_acl_tcam_profile_ops,
+};

drivers/net/ethernet/mellanox/mlxsw/spectrum_flower.c

@@ -0,0 +1,309 @@
+/*
+ * drivers/net/ethernet/mellanox/mlxsw/spectrum_flower.c
+ * Copyright (c) 2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2017 Jiri Pirko <jiri@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/netdevice.h>
+#include <net/flow_dissector.h>
+#include <net/pkt_cls.h>
+#include <net/tc_act/tc_gact.h>
+#include <net/tc_act/tc_mirred.h>
+
+#include "spectrum.h"
+#include "core_acl_flex_keys.h"
+
+static int mlxsw_sp_flower_parse_actions(struct mlxsw_sp *mlxsw_sp,
+					 struct net_device *dev,
+					 struct mlxsw_sp_acl_rule_info *rulei,
+					 struct tcf_exts *exts)
+{
+	const struct tc_action *a;
+	LIST_HEAD(actions);
+	int err;
+
+	if (tc_no_actions(exts))
+		return 0;
+
+	tcf_exts_to_list(exts, &actions);
+	list_for_each_entry(a, &actions, list) {
+		if (is_tcf_gact_shot(a)) {
+			err = mlxsw_sp_acl_rulei_act_drop(rulei);
+			if (err)
+				return err;
+		} else if (is_tcf_mirred_egress_redirect(a)) {
+			int ifindex = tcf_mirred_ifindex(a);
+			struct net_device *out_dev;
+
+			out_dev = __dev_get_by_index(dev_net(dev), ifindex);
+			if (out_dev == dev)
+				out_dev = NULL;
+
+			err = mlxsw_sp_acl_rulei_act_fwd(mlxsw_sp, rulei,
+							 out_dev);
+			if (err)
+				return err;
+		} else {
+			dev_err(mlxsw_sp->bus_info->dev, "Unsupported action\n");
+			return -EOPNOTSUPP;
+		}
+	}
+	return 0;
+}
+
+static void mlxsw_sp_flower_parse_ipv4(struct mlxsw_sp_acl_rule_info *rulei,
+				       struct tc_cls_flower_offload *f)
+{
+	struct flow_dissector_key_ipv4_addrs *key =
+		skb_flow_dissector_target(f->dissector,
+					  FLOW_DISSECTOR_KEY_IPV4_ADDRS,
+					  f->key);
+	struct flow_dissector_key_ipv4_addrs *mask =
+		skb_flow_dissector_target(f->dissector,
+					  FLOW_DISSECTOR_KEY_IPV4_ADDRS,
+					  f->mask);
+
+	mlxsw_sp_acl_rulei_keymask_u32(rulei, MLXSW_AFK_ELEMENT_SRC_IP4,
+				       ntohl(key->src), ntohl(mask->src));
+	mlxsw_sp_acl_rulei_keymask_u32(rulei, MLXSW_AFK_ELEMENT_DST_IP4,
+				       ntohl(key->dst), ntohl(mask->dst));
+}
+
+static void mlxsw_sp_flower_parse_ipv6(struct mlxsw_sp_acl_rule_info *rulei,
+				       struct tc_cls_flower_offload *f)
+{
+	struct flow_dissector_key_ipv6_addrs *key =
+		skb_flow_dissector_target(f->dissector,
+					  FLOW_DISSECTOR_KEY_IPV6_ADDRS,
+					  f->key);
+	struct flow_dissector_key_ipv6_addrs *mask =
+		skb_flow_dissector_target(f->dissector,
+					  FLOW_DISSECTOR_KEY_IPV6_ADDRS,
+					  f->mask);
+	size_t addr_half_size = sizeof(key->src) / 2;
+
+	mlxsw_sp_acl_rulei_keymask_buf(rulei, MLXSW_AFK_ELEMENT_SRC_IP6_HI,
+				       &key->src.s6_addr[0],
+				       &mask->src.s6_addr[0],
+				       addr_half_size);
+	mlxsw_sp_acl_rulei_keymask_buf(rulei, MLXSW_AFK_ELEMENT_SRC_IP6_LO,
+				       &key->src.s6_addr[addr_half_size],
+				       &mask->src.s6_addr[addr_half_size],
+				       addr_half_size);
+	mlxsw_sp_acl_rulei_keymask_buf(rulei, MLXSW_AFK_ELEMENT_DST_IP6_HI,
+				       &key->dst.s6_addr[0],
+				       &mask->dst.s6_addr[0],
+				       addr_half_size);
+	mlxsw_sp_acl_rulei_keymask_buf(rulei, MLXSW_AFK_ELEMENT_DST_IP6_LO,
+				       &key->dst.s6_addr[addr_half_size],
+				       &mask->dst.s6_addr[addr_half_size],
+				       addr_half_size);
+}
+
+static int mlxsw_sp_flower_parse_ports(struct mlxsw_sp *mlxsw_sp,
+				       struct mlxsw_sp_acl_rule_info *rulei,
+				       struct tc_cls_flower_offload *f,
+				       u8 ip_proto)
+{
+	struct flow_dissector_key_ports *key, *mask;
+
+	if (!dissector_uses_key(f->dissector, FLOW_DISSECTOR_KEY_PORTS))
+		return 0;
+
+	if (ip_proto != IPPROTO_TCP && ip_proto != IPPROTO_UDP) {
+		dev_err(mlxsw_sp->bus_info->dev, "Only UDP and TCP keys are supported\n");
+		return -EINVAL;
+	}
+
+	key = skb_flow_dissector_target(f->dissector,
+					FLOW_DISSECTOR_KEY_PORTS,
+					f->key);
+	mask = skb_flow_dissector_target(f->dissector,
+					 FLOW_DISSECTOR_KEY_PORTS,
+					 f->mask);
+	mlxsw_sp_acl_rulei_keymask_u32(rulei, MLXSW_AFK_ELEMENT_DST_L4_PORT,
+				       ntohs(key->dst), ntohs(mask->dst));
+	mlxsw_sp_acl_rulei_keymask_u32(rulei, MLXSW_AFK_ELEMENT_SRC_L4_PORT,
+				       ntohs(key->src), ntohs(mask->src));
+	return 0;
+}
+
+static int mlxsw_sp_flower_parse(struct mlxsw_sp *mlxsw_sp,
+				 struct net_device *dev,
+				 struct mlxsw_sp_acl_rule_info *rulei,
+				 struct tc_cls_flower_offload *f)
+{
+	u16 addr_type = 0;
+	u8 ip_proto = 0;
+	int err;
+
+	if (f->dissector->used_keys &
+	    ~(BIT(FLOW_DISSECTOR_KEY_CONTROL) |
+	      BIT(FLOW_DISSECTOR_KEY_BASIC) |
+	      BIT(FLOW_DISSECTOR_KEY_ETH_ADDRS) |
+	      BIT(FLOW_DISSECTOR_KEY_IPV4_ADDRS) |
+	      BIT(FLOW_DISSECTOR_KEY_IPV6_ADDRS) |
+	      BIT(FLOW_DISSECTOR_KEY_PORTS))) {
+		dev_err(mlxsw_sp->bus_info->dev, "Unsupported key\n");
+		return -EOPNOTSUPP;
+	}
+
+	mlxsw_sp_acl_rulei_priority(rulei, f->prio);
+
+	if (dissector_uses_key(f->dissector, FLOW_DISSECTOR_KEY_CONTROL)) {
+		struct flow_dissector_key_control *key =
+			skb_flow_dissector_target(f->dissector,
+						  FLOW_DISSECTOR_KEY_CONTROL,
+						  f->key);
+		addr_type = key->addr_type;
+	}
+
+	if (dissector_uses_key(f->dissector, FLOW_DISSECTOR_KEY_BASIC)) {
+		struct flow_dissector_key_basic *key =
+			skb_flow_dissector_target(f->dissector,
+						  FLOW_DISSECTOR_KEY_BASIC,
+						  f->key);
+		struct flow_dissector_key_basic *mask =
+			skb_flow_dissector_target(f->dissector,
+						  FLOW_DISSECTOR_KEY_BASIC,
+						  f->mask);
+		ip_proto = key->ip_proto;
+		mlxsw_sp_acl_rulei_keymask_u32(rulei,
+					       MLXSW_AFK_ELEMENT_ETHERTYPE,
+					       ntohs(key->n_proto),
+					       ntohs(mask->n_proto));
+		mlxsw_sp_acl_rulei_keymask_u32(rulei,
+					       MLXSW_AFK_ELEMENT_IP_PROTO,
+					       key->ip_proto, mask->ip_proto);
+	}
+
+	if (dissector_uses_key(f->dissector, FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
+		struct flow_dissector_key_eth_addrs *key =
+			skb_flow_dissector_target(f->dissector,
+						  FLOW_DISSECTOR_KEY_ETH_ADDRS,
+						  f->key);
+		struct flow_dissector_key_eth_addrs *mask =
+			skb_flow_dissector_target(f->dissector,
+						  FLOW_DISSECTOR_KEY_ETH_ADDRS,
+						  f->mask);
+
+		mlxsw_sp_acl_rulei_keymask_buf(rulei,
+					       MLXSW_AFK_ELEMENT_DMAC,
+					       key->dst, mask->dst,
+					       sizeof(key->dst));
+		mlxsw_sp_acl_rulei_keymask_buf(rulei,
+					       MLXSW_AFK_ELEMENT_SMAC,
+					       key->src, mask->src,
+					       sizeof(key->src));
+	}
+
+	if (addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS)
+		mlxsw_sp_flower_parse_ipv4(rulei, f);
+
+	if (addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS)
+		mlxsw_sp_flower_parse_ipv6(rulei, f);
+
+	err = mlxsw_sp_flower_parse_ports(mlxsw_sp, rulei, f, ip_proto);
+	if (err)
+		return err;
+
+	return mlxsw_sp_flower_parse_actions(mlxsw_sp, dev, rulei, f->exts);
+}
+
+int mlxsw_sp_flower_replace(struct mlxsw_sp_port *mlxsw_sp_port, bool ingress,
+			    __be16 protocol, struct tc_cls_flower_offload *f)
+{
+	struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+	struct net_device *dev = mlxsw_sp_port->dev;
+	struct mlxsw_sp_acl_rule_info *rulei;
+	struct mlxsw_sp_acl_ruleset *ruleset;
+	struct mlxsw_sp_acl_rule *rule;
+	int err;
+
+	ruleset = mlxsw_sp_acl_ruleset_get(mlxsw_sp, dev, ingress,
+					   MLXSW_SP_ACL_PROFILE_FLOWER);
+	if (IS_ERR(ruleset))
+		return PTR_ERR(ruleset);
+
+	rule = mlxsw_sp_acl_rule_create(mlxsw_sp, ruleset, f->cookie);
+	if (IS_ERR(rule)) {
+		err = PTR_ERR(rule);
+		goto err_rule_create;
+	}
+
+	rulei = mlxsw_sp_acl_rule_rulei(rule);
+	err = mlxsw_sp_flower_parse(mlxsw_sp, dev, rulei, f);
+	if (err)
+		goto err_flower_parse;
+
+	err = mlxsw_sp_acl_rulei_commit(rulei);
+	if (err)
+		goto err_rulei_commit;
+
+	err = mlxsw_sp_acl_rule_add(mlxsw_sp, rule);
+	if (err)
+		goto err_rule_add;
+
+	mlxsw_sp_acl_ruleset_put(mlxsw_sp, ruleset);
+	return 0;
+
+err_rule_add:
+err_rulei_commit:
+err_flower_parse:
+	mlxsw_sp_acl_rule_destroy(mlxsw_sp, rule);
+err_rule_create:
+	mlxsw_sp_acl_ruleset_put(mlxsw_sp, ruleset);
+	return err;
+}
+
+void mlxsw_sp_flower_destroy(struct mlxsw_sp_port *mlxsw_sp_port, bool ingress,
+			     struct tc_cls_flower_offload *f)
+{
+	struct mlxsw_sp *mlxsw_sp = mlxsw_sp_port->mlxsw_sp;
+	struct mlxsw_sp_acl_ruleset *ruleset;
+	struct mlxsw_sp_acl_rule *rule;
+
+	ruleset = mlxsw_sp_acl_ruleset_get(mlxsw_sp, mlxsw_sp_port->dev,
+					   ingress,
+					   MLXSW_SP_ACL_PROFILE_FLOWER);
+	if (WARN_ON(IS_ERR(ruleset)))
+		return;
+
+	rule = mlxsw_sp_acl_rule_lookup(mlxsw_sp, ruleset, f->cookie);
+	if (!WARN_ON(!rule)) {
+		mlxsw_sp_acl_rule_del(mlxsw_sp, rule);
+		mlxsw_sp_acl_rule_destroy(mlxsw_sp, rule);
+	}
+
+	mlxsw_sp_acl_ruleset_put(mlxsw_sp, ruleset);
+}

include/linux/list.h

@@ -526,6 +526,19 @@ static inline void list_splice_tail_init(struct list_head *list,
 	for (; &pos->member != (head);					\
 	     pos = list_next_entry(pos, member))
 
+/**
+ * list_for_each_entry_from_reverse - iterate backwards over list of given type
+ *                                    from the current point
+ * @pos:	the type * to use as a loop cursor.
+ * @head:	the head for your list.
+ * @member:	the name of the list_head within the struct.
+ *
+ * Iterate backwards over list of given type, continuing from current position.
+ */
+#define list_for_each_entry_from_reverse(pos, head, member)		\
+	for (; &pos->member != (head);					\
+	     pos = list_prev_entry(pos, member))
+
 /**
  * list_for_each_entry_safe - iterate over list of given type safe against removal of list entry
  * @pos:	the type * to use as a loop cursor.

include/linux/parman.h

@@ -0,0 +1,76 @@
+/*
+ * include/linux/parman.h - Manager for linear priority array areas
+ * Copyright (c) 2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2017 Jiri Pirko <jiri@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _PARMAN_H
+#define _PARMAN_H
+
+#include <linux/list.h>
+
+enum parman_algo_type {
+	PARMAN_ALGO_TYPE_LSORT,
+};
+
+struct parman_item {
+	struct list_head list;
+	unsigned long index;
+};
+
+struct parman_prio {
+	struct list_head list;
+	struct list_head item_list;
+	unsigned long priority;
+};
+
+struct parman_ops {
+	unsigned long base_count;
+	unsigned long resize_step;
+	int (*resize)(void *priv, unsigned long new_count);
+	void (*move)(void *priv, unsigned long from_index,
+		     unsigned long to_index, unsigned long count);
+	enum parman_algo_type algo;
+};
+
+struct parman;
+
+struct parman *parman_create(const struct parman_ops *ops, void *priv);
+void parman_destroy(struct parman *parman);
+void parman_prio_init(struct parman *parman, struct parman_prio *prio,
+		      unsigned long priority);
+void parman_prio_fini(struct parman_prio *prio);
+int parman_item_add(struct parman *parman, struct parman_prio *prio,
+		    struct parman_item *item);
+void parman_item_remove(struct parman *parman, struct parman_prio *prio,
+			struct parman_item *item);
+
+#endif

include/net/pkt_cls.h

@@ -481,6 +481,7 @@ enum tc_fl_command {
 
 struct tc_cls_flower_offload {
 	enum tc_fl_command command;
+	u32 prio;
 	unsigned long cookie;
 	struct flow_dissector *dissector;
 	struct fl_flow_key *mask;

lib/Kconfig

@@ -550,4 +550,7 @@ config STACKDEPOT
 config SBITMAP
 	bool
 
+config PARMAN
+	tristate "parman"
+
 endmenu

lib/Kconfig.debug

@@ -1826,6 +1826,16 @@ config TEST_HASH
 	  This is intended to help people writing architecture-specific
 	  optimized versions.  If unsure, say N.
 
+config TEST_PARMAN
+	tristate "Perform selftest on priority array manager"
+	default n
+	depends on PARMAN
+	help
+	  Enable this option to test priority array manager on boot
+	  (or module load).
+
+	  If unsure, say N.
+
 endmenu # runtime tests
 
 config PROVIDE_OHCI1394_DMA_INIT

lib/Makefile

@@ -56,6 +56,7 @@ obj-$(CONFIG_TEST_STATIC_KEYS) += test_static_key_base.o
 obj-$(CONFIG_TEST_PRINTF) += test_printf.o
 obj-$(CONFIG_TEST_BITMAP) += test_bitmap.o
 obj-$(CONFIG_TEST_UUID) += test_uuid.o
+obj-$(CONFIG_TEST_PARMAN) += test_parman.o
 
 ifeq ($(CONFIG_DEBUG_KOBJECT),y)
 CFLAGS_kobject.o += -DDEBUG
@@ -230,3 +231,5 @@ obj-$(CONFIG_UBSAN) += ubsan.o
 UBSAN_SANITIZE_ubsan.o := n
 
 obj-$(CONFIG_SBITMAP) += sbitmap.o
+
+obj-$(CONFIG_PARMAN) += parman.o

lib/parman.c

@@ -0,0 +1,376 @@
+/*
+ * lib/parman.c - Manager for linear priority array areas
+ * Copyright (c) 2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2017 Jiri Pirko <jiri@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/export.h>
+#include <linux/list.h>
+#include <linux/err.h>
+#include <linux/parman.h>
+
+struct parman_algo {
+	int (*item_add)(struct parman *parman, struct parman_prio *prio,
+			struct parman_item *item);
+	void (*item_remove)(struct parman *parman, struct parman_prio *prio,
+			    struct parman_item *item);
+};
+
+struct parman {
+	const struct parman_ops *ops;
+	void *priv;
+	const struct parman_algo *algo;
+	unsigned long count;
+	unsigned long limit_count;
+	struct list_head prio_list;
+};
+
+static int parman_enlarge(struct parman *parman)
+{
+	unsigned long new_count = parman->limit_count +
+				  parman->ops->resize_step;
+	int err;
+
+	err = parman->ops->resize(parman->priv, new_count);
+	if (err)
+		return err;
+	parman->limit_count = new_count;
+	return 0;
+}
+
+static int parman_shrink(struct parman *parman)
+{
+	unsigned long new_count = parman->limit_count -
+				  parman->ops->resize_step;
+	int err;
+
+	if (new_count < parman->ops->base_count)
+		return 0;
+	err = parman->ops->resize(parman->priv, new_count);
+	if (err)
+		return err;
+	parman->limit_count = new_count;
+	return 0;
+}
+
+static bool parman_prio_used(struct parman_prio *prio)
+
+{
+	return !list_empty(&prio->item_list);
+}
+
+static struct parman_item *parman_prio_first_item(struct parman_prio *prio)
+{
+	return list_first_entry(&prio->item_list,
+				typeof(struct parman_item), list);
+}
+
+static unsigned long parman_prio_first_index(struct parman_prio *prio)
+{
+	return parman_prio_first_item(prio)->index;
+}
+
+static struct parman_item *parman_prio_last_item(struct parman_prio *prio)
+{
+	return list_last_entry(&prio->item_list,
+			       typeof(struct parman_item), list);
+}
+
+static unsigned long parman_prio_last_index(struct parman_prio *prio)
+{
+	return parman_prio_last_item(prio)->index;
+}
+
+static unsigned long parman_lsort_new_index_find(struct parman *parman,
+						 struct parman_prio *prio)
+{
+	list_for_each_entry_from_reverse(prio, &parman->prio_list, list) {
+		if (!parman_prio_used(prio))
+			continue;
+		return parman_prio_last_index(prio) + 1;
+	}
+	return 0;
+}
+
+static void __parman_prio_move(struct parman *parman, struct parman_prio *prio,
+			       struct parman_item *item, unsigned long to_index,
+			       unsigned long count)
+{
+	parman->ops->move(parman->priv, item->index, to_index, count);
+}
+
+static void parman_prio_shift_down(struct parman *parman,
+				   struct parman_prio *prio)
+{
+	struct parman_item *item;
+	unsigned long to_index;
+
+	if (!parman_prio_used(prio))
+		return;
+	item = parman_prio_first_item(prio);
+	to_index = parman_prio_last_index(prio) + 1;
+	__parman_prio_move(parman, prio, item, to_index, 1);
+	list_move_tail(&item->list, &prio->item_list);
+	item->index = to_index;
+}
+
+static void parman_prio_shift_up(struct parman *parman,
+				 struct parman_prio *prio)
+{
+	struct parman_item *item;
+	unsigned long to_index;
+
+	if (!parman_prio_used(prio))
+		return;
+	item = parman_prio_last_item(prio);
+	to_index = parman_prio_first_index(prio) - 1;
+	__parman_prio_move(parman, prio, item, to_index, 1);
+	list_move(&item->list, &prio->item_list);
+	item->index = to_index;
+}
+
+static void parman_prio_item_remove(struct parman *parman,
+				    struct parman_prio *prio,
+				    struct parman_item *item)
+{
+	struct parman_item *last_item;
+	unsigned long to_index;
+
+	last_item = parman_prio_last_item(prio);
+	if (last_item == item) {
+		list_del(&item->list);
+		return;
+	}
+	to_index = item->index;
+	__parman_prio_move(parman, prio, last_item, to_index, 1);
+	list_del(&last_item->list);
+	list_replace(&item->list, &last_item->list);
+	last_item->index = to_index;
+}
+
+static int parman_lsort_item_add(struct parman *parman,
+				 struct parman_prio *prio,
+				 struct parman_item *item)
+{
+	struct parman_prio *prio2;
+	unsigned long new_index;
+	int err;
+
+	if (parman->count + 1 > parman->limit_count) {
+		err = parman_enlarge(parman);
+		if (err)
+			return err;
+	}
+
+	new_index = parman_lsort_new_index_find(parman, prio);
+	list_for_each_entry_reverse(prio2, &parman->prio_list, list) {
+		if (prio2 == prio)
+			break;
+		parman_prio_shift_down(parman, prio2);
+	}
+	item->index = new_index;
+	list_add_tail(&item->list, &prio->item_list);
+	parman->count++;
+	return 0;
+}
+
+static void parman_lsort_item_remove(struct parman *parman,
+				     struct parman_prio *prio,
+				     struct parman_item *item)
+{
+	parman_prio_item_remove(parman, prio, item);
+	list_for_each_entry_continue(prio, &parman->prio_list, list)
+		parman_prio_shift_up(parman, prio);
+	parman->count--;
+	if (parman->limit_count - parman->count >= parman->ops->resize_step)
+		parman_shrink(parman);
+}
+
+static const struct parman_algo parman_lsort = {
+	.item_add	= parman_lsort_item_add,
+	.item_remove	= parman_lsort_item_remove,
+};
+
+static const struct parman_algo *parman_algos[] = {
+	&parman_lsort,
+};
+
+/**
+ * parman_create - creates a new parman instance
+ * @ops:	caller-specific callbacks
+ * @priv:	pointer to a private data passed to the ops
+ *
+ * Note: all locking must be provided by the caller.
+ *
+ * Each parman instance manages an array area with chunks of entries
+ * with the same priority. Consider following example:
+ *
+ * item 1 with prio 10
+ * item 2 with prio 10
+ * item 3 with prio 10
+ * item 4 with prio 20
+ * item 5 with prio 20
+ * item 6 with prio 30
+ * item 7 with prio 30
+ * item 8 with prio 30
+ *
+ * In this example, there are 3 priority chunks. The order of the priorities
+ * matters, however the order of items within a single priority chunk does not
+ * matter. So the same array could be ordered as follows:
+ *
+ * item 2 with prio 10
+ * item 3 with prio 10
+ * item 1 with prio 10
+ * item 5 with prio 20
+ * item 4 with prio 20
+ * item 7 with prio 30
+ * item 8 with prio 30
+ * item 6 with prio 30
+ *
+ * The goal of parman is to maintain the priority ordering. The caller
+ * provides @ops with callbacks parman uses to move the items
+ * and resize the array area.
+ *
+ * Returns a pointer to newly created parman instance in case of success,
+ * otherwise it returns NULL.
+ */
+struct parman *parman_create(const struct parman_ops *ops, void *priv)
+{
+	struct parman *parman;
+
+	parman = kzalloc(sizeof(*parman), GFP_KERNEL);
+	if (!parman)
+		return NULL;
+	INIT_LIST_HEAD(&parman->prio_list);
+	parman->ops = ops;
+	parman->priv = priv;
+	parman->limit_count = ops->base_count;
+	parman->algo = parman_algos[ops->algo];
+	return parman;
+}
+EXPORT_SYMBOL(parman_create);
+
+/**
+ * parman_destroy - destroys existing parman instance
+ * @parman:	parman instance
+ *
+ * Note: all locking must be provided by the caller.
+ */
+void parman_destroy(struct parman *parman)
+{
+	WARN_ON(!list_empty(&parman->prio_list));
+	kfree(parman);
+}
+EXPORT_SYMBOL(parman_destroy);
+
+/**
+ * parman_prio_init - initializes a parman priority chunk
+ * @parman:	parman instance
+ * @prio:	parman prio structure to be initialized
+ * @prority:	desired priority of the chunk
+ *
+ * Note: all locking must be provided by the caller.
+ *
+ * Before caller could add an item with certain priority, he has to
+ * initialize a priority chunk for it using this function.
+ */
+void parman_prio_init(struct parman *parman, struct parman_prio *prio,
+		      unsigned long priority)
+{
+	struct parman_prio *prio2;
+	struct list_head *pos;
+
+	INIT_LIST_HEAD(&prio->item_list);
+	prio->priority = priority;
+
+	/* Position inside the list according to priority */
+	list_for_each(pos, &parman->prio_list) {
+		prio2 = list_entry(pos, typeof(*prio2), list);
+		if (prio2->priority > prio->priority)
+			break;
+	}
+	list_add_tail(&prio->list, pos);
+}
+EXPORT_SYMBOL(parman_prio_init);
+
+/**
+ * parman_prio_fini - finalizes use of parman priority chunk
+ * @prio:	parman prio structure
+ *
+ * Note: all locking must be provided by the caller.
+ */
+void parman_prio_fini(struct parman_prio *prio)
+{
+	WARN_ON(parman_prio_used(prio));
+	list_del(&prio->list);
+}
+EXPORT_SYMBOL(parman_prio_fini);
+
+/**
+ * parman_item_add - adds a parman item under defined priority
+ * @parman:	parman instance
+ * @prio:	parman prio instance to add the item to
+ * @item:	parman item instance
+ *
+ * Note: all locking must be provided by the caller.
+ *
+ * Adds item to a array managed by parman instance under the specified priority.
+ *
+ * Returns 0 in case of success, negative number to indicate an error.
+ */
+int parman_item_add(struct parman *parman, struct parman_prio *prio,
+		    struct parman_item *item)
+{
+	return parman->algo->item_add(parman, prio, item);
+}
+EXPORT_SYMBOL(parman_item_add);
+
+/**
+ * parman_item_del - deletes parman item
+ * @parman:	parman instance
+ * @prio:	parman prio instance to delete the item from
+ * @item:	parman item instance
+ *
+ * Note: all locking must be provided by the caller.
+ */
+void parman_item_remove(struct parman *parman, struct parman_prio *prio,
+			struct parman_item *item)
+{
+	parman->algo->item_remove(parman, prio, item);
+}
+EXPORT_SYMBOL(parman_item_remove);
+
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_AUTHOR("Jiri Pirko <jiri@mellanox.com>");
+MODULE_DESCRIPTION("Priority-based array manager");

lib/test_parman.c

@@ -0,0 +1,395 @@
+/*
+ * lib/test_parman.c - Test module for parman
+ * Copyright (c) 2017 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2017 Jiri Pirko <jiri@mellanox.com>
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ * 3. Neither the names of the copyright holders nor the names of its
+ *    contributors may be used to endorse or promote products derived from
+ *    this software without specific prior written permission.
+ *
+ * Alternatively, this software may be distributed under the terms of the
+ * GNU General Public License ("GPL") version 2 as published by the Free
+ * Software Foundation.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/bitops.h>
+#include <linux/err.h>
+#include <linux/random.h>
+#include <linux/parman.h>
+
+#define TEST_PARMAN_PRIO_SHIFT 7 /* defines number of prios for testing */
+#define TEST_PARMAN_PRIO_COUNT BIT(TEST_PARMAN_PRIO_SHIFT)
+#define TEST_PARMAN_PRIO_MASK (TEST_PARMAN_PRIO_COUNT - 1)
+
+#define TEST_PARMAN_ITEM_SHIFT 13 /* defines a total number
+				   * of items for testing
+				   */
+#define TEST_PARMAN_ITEM_COUNT BIT(TEST_PARMAN_ITEM_SHIFT)
+#define TEST_PARMAN_ITEM_MASK (TEST_PARMAN_ITEM_COUNT - 1)
+
+#define TEST_PARMAN_BASE_SHIFT 8
+#define TEST_PARMAN_BASE_COUNT BIT(TEST_PARMAN_BASE_SHIFT)
+#define TEST_PARMAN_RESIZE_STEP_SHIFT 7
+#define TEST_PARMAN_RESIZE_STEP_COUNT BIT(TEST_PARMAN_RESIZE_STEP_SHIFT)
+
+#define TEST_PARMAN_BULK_MAX_SHIFT (2 + TEST_PARMAN_RESIZE_STEP_SHIFT)
+#define TEST_PARMAN_BULK_MAX_COUNT BIT(TEST_PARMAN_BULK_MAX_SHIFT)
+#define TEST_PARMAN_BULK_MAX_MASK (TEST_PARMAN_BULK_MAX_COUNT - 1)
+
+#define TEST_PARMAN_RUN_BUDGET (TEST_PARMAN_ITEM_COUNT * 256)
+
+struct test_parman_prio {
+	struct parman_prio parman_prio;
+	unsigned long priority;
+};
+
+struct test_parman_item {
+	struct parman_item parman_item;
+	struct test_parman_prio *prio;
+	bool used;
+};
+
+struct test_parman {
+	struct parman *parman;
+	struct test_parman_item **prio_array;
+	unsigned long prio_array_limit;
+	struct test_parman_prio prios[TEST_PARMAN_PRIO_COUNT];
+	struct test_parman_item items[TEST_PARMAN_ITEM_COUNT];
+	struct rnd_state rnd;
+	unsigned long run_budget;
+	unsigned long bulk_budget;
+	bool bulk_noop;
+	unsigned int used_items;
+};
+
+#define ITEM_PTRS_SIZE(count) (sizeof(struct test_parman_item *) * (count))
+
+static int test_parman_resize(void *priv, unsigned long new_count)
+{
+	struct test_parman *test_parman = priv;
+	struct test_parman_item **prio_array;
+	unsigned long old_count;
+
+	prio_array = krealloc(test_parman->prio_array,
+			      ITEM_PTRS_SIZE(new_count), GFP_KERNEL);
+	if (new_count == 0)
+		return 0;
+	if (!prio_array)
+		return -ENOMEM;
+	old_count = test_parman->prio_array_limit;
+	if (new_count > old_count)
+		memset(&prio_array[old_count], 0,
+		       ITEM_PTRS_SIZE(new_count - old_count));
+	test_parman->prio_array = prio_array;
+	test_parman->prio_array_limit = new_count;
+	return 0;
+}
+
+static void test_parman_move(void *priv, unsigned long from_index,
+			     unsigned long to_index, unsigned long count)
+{
+	struct test_parman *test_parman = priv;
+	struct test_parman_item **prio_array = test_parman->prio_array;
+
+	memmove(&prio_array[to_index], &prio_array[from_index],
+		ITEM_PTRS_SIZE(count));
+	memset(&prio_array[from_index], 0, ITEM_PTRS_SIZE(count));
+}
+
+static const struct parman_ops test_parman_lsort_ops = {
+	.base_count	= TEST_PARMAN_BASE_COUNT,
+	.resize_step	= TEST_PARMAN_RESIZE_STEP_COUNT,
+	.resize		= test_parman_resize,
+	.move		= test_parman_move,
+	.algo		= PARMAN_ALGO_TYPE_LSORT,
+};
+
+static void test_parman_rnd_init(struct test_parman *test_parman)
+{
+	prandom_seed_state(&test_parman->rnd, 3141592653589793238ULL);
+}
+
+static u32 test_parman_rnd_get(struct test_parman *test_parman)
+{
+	return prandom_u32_state(&test_parman->rnd);
+}
+
+static unsigned long test_parman_priority_gen(struct test_parman *test_parman)
+{
+	unsigned long priority;
+	int i;
+
+again:
+	priority = test_parman_rnd_get(test_parman);
+	if (priority == 0)
+		goto again;
+
+	for (i = 0; i < TEST_PARMAN_PRIO_COUNT; i++) {
+		struct test_parman_prio *prio = &test_parman->prios[i];
+
+		if (prio->priority == 0)
+			break;
+		if (prio->priority == priority)
+			goto again;
+	}
+	return priority;
+}
+
+static void test_parman_prios_init(struct test_parman *test_parman)
+{
+	int i;
+
+	for (i = 0; i < TEST_PARMAN_PRIO_COUNT; i++) {
+		struct test_parman_prio *prio = &test_parman->prios[i];
+
+		/* Assign random uniqueue priority to each prio structure */
+		prio->priority = test_parman_priority_gen(test_parman);
+		parman_prio_init(test_parman->parman, &prio->parman_prio,
+				 prio->priority);
+	}
+}
+
+static void test_parman_prios_fini(struct test_parman *test_parman)
+{
+	int i;
+
+	for (i = 0; i < TEST_PARMAN_PRIO_COUNT; i++) {
+		struct test_parman_prio *prio = &test_parman->prios[i];
+
+		parman_prio_fini(&prio->parman_prio);
+	}
+}
+
+static void test_parman_items_init(struct test_parman *test_parman)
+{
+	int i;
+
+	for (i = 0; i < TEST_PARMAN_ITEM_COUNT; i++) {
+		struct test_parman_item *item = &test_parman->items[i];
+		unsigned int prio_index = test_parman_rnd_get(test_parman) &
+					  TEST_PARMAN_PRIO_MASK;
+
+		/* Assign random prio to each item structure */
+		item->prio = &test_parman->prios[prio_index];
+	}
+}
+
+static void test_parman_items_fini(struct test_parman *test_parman)
+{
+	int i;
+
+	for (i = 0; i < TEST_PARMAN_ITEM_COUNT; i++) {
+		struct test_parman_item *item = &test_parman->items[i];
+
+		if (!item->used)
+			continue;
+		parman_item_remove(test_parman->parman,
+				   &item->prio->parman_prio,
+				   &item->parman_item);
+	}
+}
+
+static struct test_parman *test_parman_create(const struct parman_ops *ops)
+{
+	struct test_parman *test_parman;
+	int err;
+
+	test_parman = kzalloc(sizeof(*test_parman), GFP_KERNEL);
+	if (!test_parman)
+		return ERR_PTR(-ENOMEM);
+	err = test_parman_resize(test_parman, TEST_PARMAN_BASE_COUNT);
+	if (err)
+		goto err_resize;
+	test_parman->parman = parman_create(ops, test_parman);
+	if (!test_parman->parman) {
+		err = -ENOMEM;
+		goto err_parman_create;
+	}
+	test_parman_rnd_init(test_parman);
+	test_parman_prios_init(test_parman);
+	test_parman_items_init(test_parman);
+	test_parman->run_budget = TEST_PARMAN_RUN_BUDGET;
+	return test_parman;
+
+err_parman_create:
+	test_parman_resize(test_parman, 0);
+err_resize:
+	kfree(test_parman);
+	return ERR_PTR(err);
+}
+
+static void test_parman_destroy(struct test_parman *test_parman)
+{
+	test_parman_items_fini(test_parman);
+	test_parman_prios_fini(test_parman);
+	parman_destroy(test_parman->parman);
+	test_parman_resize(test_parman, 0);
+	kfree(test_parman);
+}
+
+static bool test_parman_run_check_budgets(struct test_parman *test_parman)
+{
+	if (test_parman->run_budget-- == 0)
+		return false;
+	if (test_parman->bulk_budget-- != 0)
+		return true;
+
+	test_parman->bulk_budget = test_parman_rnd_get(test_parman) &
+				   TEST_PARMAN_BULK_MAX_MASK;
+	test_parman->bulk_noop = test_parman_rnd_get(test_parman) & 1;
+	return true;
+}
+
+static int test_parman_run(struct test_parman *test_parman)
+{
+	unsigned int i = test_parman_rnd_get(test_parman);
+	int err;
+
+	while (test_parman_run_check_budgets(test_parman)) {
+		unsigned int item_index = i++ & TEST_PARMAN_ITEM_MASK;
+		struct test_parman_item *item = &test_parman->items[item_index];
+
+		if (test_parman->bulk_noop)
+			continue;
+
+		if (!item->used) {
+			err = parman_item_add(test_parman->parman,
+					      &item->prio->parman_prio,
+					      &item->parman_item);
+			if (err)
+				return err;
+			test_parman->prio_array[item->parman_item.index] = item;
+			test_parman->used_items++;
+		} else {
+			test_parman->prio_array[item->parman_item.index] = NULL;
+			parman_item_remove(test_parman->parman,
+					   &item->prio->parman_prio,
+					   &item->parman_item);
+			test_parman->used_items--;
+		}
+		item->used = !item->used;
+	}
+	return 0;
+}
+
+static int test_parman_check_array(struct test_parman *test_parman,
+				   bool gaps_allowed)
+{
+	unsigned int last_unused_items = 0;
+	unsigned long last_priority = 0;
+	unsigned int used_items = 0;
+	int i;
+
+	if (test_parman->prio_array_limit < TEST_PARMAN_BASE_COUNT) {
+		pr_err("Array limit is lower than the base count (%lu < %lu)\n",
+		       test_parman->prio_array_limit, TEST_PARMAN_BASE_COUNT);
+		return -EINVAL;
+	}
+
+	for (i = 0; i < test_parman->prio_array_limit; i++) {
+		struct test_parman_item *item = test_parman->prio_array[i];
+
+		if (!item) {
+			last_unused_items++;
+			continue;
+		}
+		if (last_unused_items && !gaps_allowed) {
+			pr_err("Gap found in array even though they are forbidden\n");
+			return -EINVAL;
+		}
+
+		last_unused_items = 0;
+		used_items++;
+
+		if (item->prio->priority < last_priority) {
+			pr_err("Item belongs under higher priority then the last one (current: %lu, previous: %lu)\n",
+			       item->prio->priority, last_priority);
+			return -EINVAL;
+		}
+		last_priority = item->prio->priority;
+
+		if (item->parman_item.index != i) {
+			pr_err("Item has different index in compare to where it actualy is (%lu != %d)\n",
+			       item->parman_item.index, i);
+			return -EINVAL;
+		}
+	}
+
+	if (used_items != test_parman->used_items) {
+		pr_err("Number of used items in array does not match (%u != %u)\n",
+		       used_items, test_parman->used_items);
+		return -EINVAL;
+	}
+
+	if (last_unused_items >= TEST_PARMAN_RESIZE_STEP_COUNT) {
+		pr_err("Number of unused item at the end of array is bigger than resize step (%u >= %lu)\n",
+		       last_unused_items, TEST_PARMAN_RESIZE_STEP_COUNT);
+		return -EINVAL;
+	}
+
+	pr_info("Priority array check successful\n");
+
+	return 0;
+}
+
+static int test_parman_lsort(void)
+{
+	struct test_parman *test_parman;
+	int err;
+
+	test_parman = test_parman_create(&test_parman_lsort_ops);
+	if (IS_ERR(test_parman))
+		return PTR_ERR(test_parman);
+
+	err = test_parman_run(test_parman);
+	if (err)
+		goto out;
+
+	err = test_parman_check_array(test_parman, false);
+	if (err)
+		goto out;
+out:
+	test_parman_destroy(test_parman);
+	return err;
+}
+
+static int __init test_parman_init(void)
+{
+	return test_parman_lsort();
+}
+
+static void __exit test_parman_exit(void)
+{
+}
+
+module_init(test_parman_init);
+module_exit(test_parman_exit);
+
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_AUTHOR("Jiri Pirko <jiri@mellanox.com>");
+MODULE_DESCRIPTION("Test module for parman");

net/sched/cls_flower.c

@@ -229,6 +229,7 @@ static void fl_hw_destroy_filter(struct tcf_proto *tp, struct cls_fl_filter *f)
 		return;
 
 	offload.command = TC_CLSFLOWER_DESTROY;
+	offload.prio = tp->prio;
 	offload.cookie = (unsigned long)f;
 
 	tc->type = TC_SETUP_CLSFLOWER;
@@ -260,6 +261,7 @@ static int fl_hw_replace_filter(struct tcf_proto *tp,
 	}
 
 	offload.command = TC_CLSFLOWER_REPLACE;
+	offload.prio = tp->prio;
 	offload.cookie = (unsigned long)f;
 	offload.dissector = dissector;
 	offload.mask = mask;
@@ -287,6 +289,7 @@ static void fl_hw_update_stats(struct tcf_proto *tp, struct cls_fl_filter *f)
 		return;
 
 	offload.command = TC_CLSFLOWER_STATS;
+	offload.prio = tp->prio;
 	offload.cookie = (unsigned long)f;
 	offload.exts = &f->exts;