[patch net-next 00/19] mlxsw: Introduce TC Flower offload using TCAM

[patch net-next 00/19] mlxsw: Introduce TC Flower offload using TCAM

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

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.

Jiri Pirko (19):
  mlxsw: item: Add 8bit item helpers
  mlxsw: item: Add helpers for getting pointer into payload for char
    buffer item
  mlxsw: reg: Add Policy-Engine ACL Register
  mlxsw: reg: Add Policy-Engine ACL Group Table register
  mlxsw: reg: Add Policy-Engine TCAM Allocation Register
  mlxsw: reg: Add Policy-Engine TCAM Entry Register Version 2
  mlxsw: reg: Add Policy-Engine Port Binding Table
  mlxsw: reg: Add Policy-Engine Rules Copy Register
  mlxsw: reg: Add Policy-Engine Policy Based Switching Register
  mlxsw: reg: Add Policy-Engine Extended Flexible Action Register
  mlxsw: core: Introduce flexible keys support
  mlxsw: core: Introduce flexible actions support
  mlxsw: spectrum: Introduce basic set of flexible key blocks
  mlxsw: resources: Add ACL related resources
  list: introduce list_for_each_entry_from_reverse helper
  lib: Introduce priority array area manager
  mlxsw: spectrum: Introduce ACL core with simple TCAM implementation
  sched: cls_flower: expose priority to offloading netdevice
  mlxsw: spectrum: Implement TC flower offload

 MAINTAINERS                                        |    8 +
 drivers/net/ethernet/mellanox/mlxsw/Kconfig        |    1 +
 drivers/net/ethernet/mellanox/mlxsw/Makefile       |    6 +-
 .../mellanox/mlxsw/core_acl_flex_actions.c         |  685 +++++++++++++
 .../mellanox/mlxsw/core_acl_flex_actions.h         |   66 ++
 .../ethernet/mellanox/mlxsw/core_acl_flex_keys.c   |  475 +++++++++
 .../ethernet/mellanox/mlxsw/core_acl_flex_keys.h   |  238 +++++
 drivers/net/ethernet/mellanox/mlxsw/item.h         |   98 +-
 drivers/net/ethernet/mellanox/mlxsw/reg.h          |  511 ++++++++-
 drivers/net/ethernet/mellanox/mlxsw/resources.h    |   20 +-
 drivers/net/ethernet/mellanox/mlxsw/spectrum.c     |   32 +-
 drivers/net/ethernet/mellanox/mlxsw/spectrum.h     |  106 +-
 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c |  572 +++++++++++
 .../mellanox/mlxsw/spectrum_acl_flex_keys.h        |  109 ++
 .../ethernet/mellanox/mlxsw/spectrum_acl_tcam.c    | 1084 ++++++++++++++++++++
 .../net/ethernet/mellanox/mlxsw/spectrum_flower.c  |  309 ++++++
 include/linux/list.h                               |   13 +
 include/linux/parman.h                             |   76 ++
 include/net/pkt_cls.h                              |    1 +
 lib/Kconfig                                        |    3 +
 lib/Kconfig.debug                                  |   10 +
 lib/Makefile                                       |    3 +
 lib/parman.c                                       |  294 ++++++
 lib/test_parman.c                                  |  395 +++++++
 net/sched/cls_flower.c                             |    3 +
 25 files changed, 5102 insertions(+), 16 deletions(-)
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.c
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.h
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.c
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.h
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_flex_keys.h
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_tcam.c
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_flower.c
 create mode 100644 include/linux/parman.h
 create mode 100644 lib/parman.c
 create mode 100644 lib/test_parman.c

-- 
2.7.4

[patch net-next 01/19] mlxsw: item: Add 8bit item helpers

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

Item heplers for 8bit values are needed, let's add them.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/item.h | 79 +++++++++++++++++++++++++++++-
 1 file changed, 77 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/item.h b/drivers/net/ethernet/mellanox/mlxsw/item.h
index 3c95e3d..09f35de 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/item.h
+++ b/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)
@@ -253,6 +287,47 @@ 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,							\
-- 
2.7.4

[patch net-next 02/19] mlxsw: item: Add helpers for getting pointer into payload for char buffer item

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

Sometimes it is handy to get a pointer to a char buffer item and use it
direcly to write/read data. So add these helpers.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/item.h | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/item.h b/drivers/net/ethernet/mellanox/mlxsw/item.h
index 09f35de..28427f0 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/item.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/item.h
@@ -225,6 +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)
@@ -468,6 +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,	\
@@ -494,6 +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,	\
-- 
2.7.4

[patch net-next 03/19] mlxsw: reg: Add Policy-Engine ACL Register

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

The PACL register is used for configuration of the ACL.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/reg.h | 47 +++++++++++++++++++++++++++++--
 1 file changed, 45 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/reg.h b/drivers/net/ethernet/mellanox/mlxsw/reg.h
index 9fb0316..18b2da4 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/reg.h
+++ b/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 +1757,48 @@ static inline void mlxsw_reg_spvmlr_pack(char *payload, u8 local_port,
 	}
 }
 
+/* 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);
+}
+
 /* QPCR - QoS Policer Configuration Register
  * -----------------------------------------
  * The QPCR register is used to create policers - that limit
@@ -5434,6 +5476,7 @@ static const struct mlxsw_reg_info *mlxsw_reg_infos[] = {
 	MLXSW_REG(svpe),
 	MLXSW_REG(sfmr),
 	MLXSW_REG(spvmlr),
+	MLXSW_REG(pacl),
 	MLXSW_REG(qpcr),
 	MLXSW_REG(qtct),
 	MLXSW_REG(qeec),
-- 
2.7.4

[patch net-next 04/19] mlxsw: reg: Add Policy-Engine ACL Group Table register

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

The PAGT register is used for configuration of the ACL Group Table.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/reg.h | 54 +++++++++++++++++++++++++++++++
 1 file changed, 54 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/reg.h b/drivers/net/ethernet/mellanox/mlxsw/reg.h
index 18b2da4..5f76fbc 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/reg.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/reg.h
@@ -1799,6 +1799,59 @@ static inline void mlxsw_reg_pacl_pack(char *payload, u16 acl_id,
 	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);
+}
+
 /* QPCR - QoS Policer Configuration Register
  * -----------------------------------------
  * The QPCR register is used to create policers - that limit
@@ -5477,6 +5530,7 @@ static const struct mlxsw_reg_info *mlxsw_reg_infos[] = {
 	MLXSW_REG(sfmr),
 	MLXSW_REG(spvmlr),
 	MLXSW_REG(pacl),
+	MLXSW_REG(pagt),
 	MLXSW_REG(qpcr),
 	MLXSW_REG(qtct),
 	MLXSW_REG(qeec),
-- 
2.7.4

[patch net-next 05/19] mlxsw: reg: Add Policy-Engine TCAM Allocation Register

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

The PTAR register is used for allocation of regions in the TCAM.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/reg.h | 106 ++++++++++++++++++++++++++++++
 1 file changed, 106 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/reg.h b/drivers/net/ethernet/mellanox/mlxsw/reg.h
index 5f76fbc..444f0a3 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/reg.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/reg.h
@@ -1852,6 +1852,111 @@ static inline void mlxsw_reg_pagt_acl_id_pack(char *payload, int index,
 	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);
+}
+
 /* QPCR - QoS Policer Configuration Register
  * -----------------------------------------
  * The QPCR register is used to create policers - that limit
@@ -5531,6 +5636,7 @@ static const struct mlxsw_reg_info *mlxsw_reg_infos[] = {
 	MLXSW_REG(spvmlr),
 	MLXSW_REG(pacl),
 	MLXSW_REG(pagt),
+	MLXSW_REG(ptar),
 	MLXSW_REG(qpcr),
 	MLXSW_REG(qtct),
 	MLXSW_REG(qeec),
-- 
2.7.4

[patch net-next 06/19] mlxsw: reg: Add Policy-Engine TCAM Entry Register Version 2

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

The PTCE-V2 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.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/reg.h | 100 ++++++++++++++++++++++++++++++
 1 file changed, 100 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/reg.h b/drivers/net/ethernet/mellanox/mlxsw/reg.h
index 444f0a3..1008251 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/reg.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/reg.h
@@ -1957,6 +1957,105 @@ 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);
 }
 
+/* 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);
+
+#define MLXSW_REG_PTCE2_FLEX_ACTION_SET_LEN 0xA8
+
+/* reg_ptce2_flex_action_set
+ * ACL action set.
+ * Access: RW
+ */
+MLXSW_ITEM_BUF(reg, ptce2, flex_action_set, 0xE0,
+	       MLXSW_REG_PTCE2_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
@@ -5637,6 +5736,7 @@ static const struct mlxsw_reg_info *mlxsw_reg_infos[] = {
 	MLXSW_REG(pacl),
 	MLXSW_REG(pagt),
 	MLXSW_REG(ptar),
+	MLXSW_REG(ptce2),
 	MLXSW_REG(qpcr),
 	MLXSW_REG(qtct),
 	MLXSW_REG(qeec),
-- 
2.7.4

[patch net-next 07/19] mlxsw: reg: Add Policy-Engine Port Binding Table

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

The PPBT is used for configuration of the Port Binding Table.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/reg.h | 63 +++++++++++++++++++++++++++++++
 1 file changed, 63 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/reg.h b/drivers/net/ethernet/mellanox/mlxsw/reg.h
index 1008251..ce6d85a 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/reg.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/reg.h
@@ -1757,6 +1757,68 @@ 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.
@@ -5733,6 +5795,7 @@ 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),
-- 
2.7.4

[patch net-next 08/19] mlxsw: reg: Add Policy-Engine Rules Copy Register

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

The PRCR register is used for accessing rules within a TCAM region.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/reg.h | 77 +++++++++++++++++++++++++++++++
 1 file changed, 77 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/reg.h b/drivers/net/ethernet/mellanox/mlxsw/reg.h
index ce6d85a..555cb80 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/reg.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/reg.h
@@ -2019,6 +2019,82 @@ 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);
 }
 
+/* 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);
+}
+
 /* PTCE-V2 - Policy-Engine TCAM Entry Register Version 2
  * -----------------------------------------------------
  * This register is used for accessing rules within a TCAM region.
@@ -5799,6 +5875,7 @@ static const struct mlxsw_reg_info *mlxsw_reg_infos[] = {
 	MLXSW_REG(pacl),
 	MLXSW_REG(pagt),
 	MLXSW_REG(ptar),
+	MLXSW_REG(prcr),
 	MLXSW_REG(ptce2),
 	MLXSW_REG(qpcr),
 	MLXSW_REG(qtct),
-- 
2.7.4

[patch net-next 09/19] mlxsw: reg: Add Policy-Engine Policy Based Switching Register

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

The PPBS register retrieves and sets Policy Based Switching Table entries.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/reg.h | 31 +++++++++++++++++++++++++++++++
 1 file changed, 31 insertions(+)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/reg.h b/drivers/net/ethernet/mellanox/mlxsw/reg.h
index 555cb80..c503363 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/reg.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/reg.h
@@ -2019,6 +2019,36 @@ 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.
@@ -5875,6 +5905,7 @@ static const struct mlxsw_reg_info *mlxsw_reg_infos[] = {
 	MLXSW_REG(pacl),
 	MLXSW_REG(pagt),
 	MLXSW_REG(ptar),
+	MLXSW_REG(ppbs),
 	MLXSW_REG(prcr),
 	MLXSW_REG(ptce2),
 	MLXSW_REG(qpcr),
-- 
2.7.4

[patch net-next 10/19] mlxsw: reg: Add Policy-Engine Extended Flexible Action Register

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

PEFA register is used for accessing an extended flexible action entry
in the central KVD Linear Database.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/reg.h | 39 ++++++++++++++++++++++++++++---
 1 file changed, 36 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/reg.h b/drivers/net/ethernet/mellanox/mlxsw/reg.h
index c503363..b50a312 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/reg.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/reg.h
@@ -2125,6 +2125,40 @@ static inline void mlxsw_reg_prcr_pack(char *payload, enum mlxsw_reg_prcr_op op,
 						       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.
@@ -2203,14 +2237,12 @@ MLXSW_ITEM_BUF(reg, ptce2, flex_key_blocks, 0x20,
 MLXSW_ITEM_BUF(reg, ptce2, mask, 0x80,
 	       MLXSW_REG_PTCE2_FLEX_KEY_BLOCKS_LEN);
 
-#define MLXSW_REG_PTCE2_FLEX_ACTION_SET_LEN 0xA8
-
 /* reg_ptce2_flex_action_set
  * ACL action set.
  * Access: RW
  */
 MLXSW_ITEM_BUF(reg, ptce2, flex_action_set, 0xE0,
-	       MLXSW_REG_PTCE2_FLEX_ACTION_SET_LEN);
+	       MLXSW_REG_PXXX_FLEX_ACTION_SET_LEN);
 
 static inline void mlxsw_reg_ptce2_pack(char *payload, bool valid,
 					enum mlxsw_reg_ptce2_op op,
@@ -5907,6 +5939,7 @@ static const struct mlxsw_reg_info *mlxsw_reg_infos[] = {
 	MLXSW_REG(ptar),
 	MLXSW_REG(ppbs),
 	MLXSW_REG(prcr),
+	MLXSW_REG(pefa),
 	MLXSW_REG(ptce2),
 	MLXSW_REG(qpcr),
 	MLXSW_REG(qtct),
-- 
2.7.4

[patch net-next 11/19] mlxsw: core: Introduce flexible keys support

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

Hardware supports matching on so called "flexible keys". The idea is to
assemble an optimal key to use for matching according to the fields in
packet (elements) requested by user. Certain sets of elements are
combined into pre-defined blocks. There is a picker to find needed blocks.
Keys consist of 1..n blocks.

Alongside with that, an initial portion of elements is introduced in order
to be able to offload basic cls_flower rules.

Picked keys are cached so multiple rules could share them.

There is an encode function provided that takes care of encoding key and
mask values according to given key.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/Makefile       |   2 +-
 .../ethernet/mellanox/mlxsw/core_acl_flex_keys.c   | 475 +++++++++++++++++++++
 .../ethernet/mellanox/mlxsw/core_acl_flex_keys.h   | 238 +++++++++++
 3 files changed, 714 insertions(+), 1 deletion(-)
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.c
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.h

diff --git a/drivers/net/ethernet/mellanox/mlxsw/Makefile b/drivers/net/ethernet/mellanox/mlxsw/Makefile
index fe8dadb..6a83768 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/Makefile
+++ b/drivers/net/ethernet/mellanox/mlxsw/Makefile
@@ -1,5 +1,5 @@
 obj-$(CONFIG_MLXSW_CORE)	+= mlxsw_core.o
-mlxsw_core-objs			:= core.o
+mlxsw_core-objs			:= core.o core_acl_flex_keys.o
 mlxsw_core-$(CONFIG_MLXSW_CORE_HWMON) += core_hwmon.o
 mlxsw_core-$(CONFIG_MLXSW_CORE_THERMAL) += core_thermal.o
 obj-$(CONFIG_MLXSW_PCI)		+= mlxsw_pci.o
diff --git a/drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.c b/drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.c
new file mode 100644
index 0000000..492b1a4
--- /dev/null
+++ b/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);
diff --git a/drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.h b/drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.h
new file mode 100644
index 0000000..809a082
--- /dev/null
+++ b/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
-- 
2.7.4

[patch net-next 12/19] mlxsw: core: Introduce flexible actions support

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

Each entry which is matched during ACL lookup points to an action set.
This action set contains up to three separate actions. If more actions
are needed to be chained, the extended set is created to hold them
in KVD linear area.

This patch implements handling of sets and encoding of actions.
Currectly, only two actions are supported. Drop and forward. Forward
action uses PBS pointer to KVD linear area, so the action code needs to
take care of this as well.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/Makefile       |   3 +-
 .../mellanox/mlxsw/core_acl_flex_actions.c         | 685 +++++++++++++++++++++
 .../mellanox/mlxsw/core_acl_flex_actions.h         |  66 ++
 3 files changed, 753 insertions(+), 1 deletion(-)
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.c
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.h

diff --git a/drivers/net/ethernet/mellanox/mlxsw/Makefile b/drivers/net/ethernet/mellanox/mlxsw/Makefile
index 6a83768..c4c48ba 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/Makefile
+++ b/drivers/net/ethernet/mellanox/mlxsw/Makefile
@@ -1,5 +1,6 @@
 obj-$(CONFIG_MLXSW_CORE)	+= mlxsw_core.o
-mlxsw_core-objs			:= core.o core_acl_flex_keys.o
+mlxsw_core-objs			:= core.o core_acl_flex_keys.o \
+				   core_acl_flex_actions.o
 mlxsw_core-$(CONFIG_MLXSW_CORE_HWMON) += core_hwmon.o
 mlxsw_core-$(CONFIG_MLXSW_CORE_THERMAL) += core_thermal.o
 obj-$(CONFIG_MLXSW_PCI)		+= mlxsw_pci.o
diff --git a/drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.c b/drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.c
new file mode 100644
index 0000000..34e2fef
--- /dev/null
+++ b/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);
diff --git a/drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.h b/drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.h
new file mode 100644
index 0000000..43f78dc
--- /dev/null
+++ b/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
-- 
2.7.4

[patch net-next 13/19] mlxsw: spectrum: Introduce basic set of flexible key blocks

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

Introduce basic set of Spectrum flexible key blocks. It contains blocks
needed to carry all elements defined so far.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 .../mellanox/mlxsw/spectrum_acl_flex_keys.h        | 109 +++++++++++++++++++++
 1 file changed, 109 insertions(+)
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_flex_keys.h

diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_flex_keys.h b/drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_flex_keys.h
new file mode 100644
index 0000000..82b81cf
--- /dev/null
+++ b/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
-- 
2.7.4

[patch net-next 14/19] mlxsw: resources: Add ACL related resources

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

Add couple of resource limits related to ACL.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/resources.h | 20 ++++++++++++++++++--
 1 file changed, 18 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlxsw/resources.h b/drivers/net/ethernet/mellanox/mlxsw/resources.h
index 3c2171d..bce8c2e 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/resources.h
+++ b/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,
-- 
2.7.4

[patch net-next 15/19] list: introduce list_for_each_entry_from_reverse helper

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

Similar to list_for_each_entry_continue and its reverse variant
list_for_each_entry_continue_reverse, introduce reverse helper for
list_for_each_entry_from.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Ido Schimmel <idosch@mellanox.com>
---
 include/linux/list.h | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/include/linux/list.h b/include/linux/list.h
index d1039ec..ae537fa 100644
--- a/include/linux/list.h
+++ b/include/linux/list.h
@@ -527,6 +527,19 @@ static inline void list_splice_tail_init(struct list_head *list,
 	     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.
  * @n:		another type * to use as temporary storage
-- 
2.7.4

[patch net-next 16/19] lib: Introduce priority array area manager

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

This introduces a infrastructure for management of linear priority
areas. Priority order in an array matters, however order of items inside
a priority group does not matter.

As an initial implementation, L-sort algorithm is used. It is quite
trivial. More advanced algorithm called P-sort will be introduced as a
follow-up. The infrastructure is prepared for other algos.

Alongside this, a testing module is introduced as well.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
---
 MAINTAINERS            |   8 +
 include/linux/parman.h |  76 ++++++++++
 lib/Kconfig            |   3 +
 lib/Kconfig.debug      |  10 ++
 lib/Makefile           |   3 +
 lib/parman.c           | 294 ++++++++++++++++++++++++++++++++++++
 lib/test_parman.c      | 395 +++++++++++++++++++++++++++++++++++++++++++++++++
 7 files changed, 789 insertions(+)
 create mode 100644 include/linux/parman.h
 create mode 100644 lib/parman.c
 create mode 100644 lib/test_parman.c

diff --git a/MAINTAINERS b/MAINTAINERS
index 300d2ec..626758b 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -9375,6 +9375,14 @@ 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
diff --git a/include/linux/parman.h b/include/linux/parman.h
new file mode 100644
index 0000000..3c8cccc
--- /dev/null
+++ b/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
diff --git a/lib/Kconfig b/lib/Kconfig
index 260a80e..5d644f1 100644
--- a/lib/Kconfig
+++ b/lib/Kconfig
@@ -550,4 +550,7 @@ config STACKDEPOT
 config SBITMAP
 	bool
 
+config PARMAN
+	tristate "parman"
+
 endmenu
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 15969ab..433a788 100644
--- a/lib/Kconfig.debug
+++ b/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
diff --git a/lib/Makefile b/lib/Makefile
index 7b3008d..1c039a4 100644
--- a/lib/Makefile
+++ b/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
diff --git a/lib/parman.c b/lib/parman.c
new file mode 100644
index 0000000..5b6f4ec
--- /dev/null
+++ b/lib/parman.c
@@ -0,0 +1,294 @@
+/*
+ * 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,
+};
+
+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);
+
+void parman_destroy(struct parman *parman)
+{
+	WARN_ON(!list_empty(&parman->prio_list));
+	kfree(parman);
+}
+EXPORT_SYMBOL(parman_destroy);
+
+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);
+
+void parman_prio_fini(struct parman_prio *prio)
+{
+	WARN_ON(parman_prio_used(prio));
+	list_del(&prio->list);
+}
+EXPORT_SYMBOL(parman_prio_fini);
+
+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);
+
+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");
diff --git a/lib/test_parman.c b/lib/test_parman.c
new file mode 100644
index 0000000..fe9f3a7
--- /dev/null
+++ b/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");
-- 
2.7.4

[patch net-next 17/19] mlxsw: spectrum: Introduce ACL core with simple TCAM implementation

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

Add ACL core infrastructure for Spectrum ASIC. This infra provides an
abstraction layer over specific HW implementations. There are two basic
objects used. One is "rule" and the second is "ruleset" which serves as a
container of multiple rules. In general, within one ruleset the rules are
allowed to have multiple priorities and masks. Each ruleset is bound to
either ingress or egress a of port netdevice.

The initial TCAM implementation is very simple and limited. It utilizes
parman lsort manager to take care of TCAM region layout.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/Kconfig        |    1 +
 drivers/net/ethernet/mellanox/mlxsw/Makefile       |    3 +-
 drivers/net/ethernet/mellanox/mlxsw/spectrum.c     |   17 +-
 drivers/net/ethernet/mellanox/mlxsw/spectrum.h     |  100 +-
 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c |  572 +++++++++++
 .../ethernet/mellanox/mlxsw/spectrum_acl_tcam.c    | 1084 ++++++++++++++++++++
 6 files changed, 1769 insertions(+), 8 deletions(-)
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_tcam.c

diff --git a/drivers/net/ethernet/mellanox/mlxsw/Kconfig b/drivers/net/ethernet/mellanox/mlxsw/Kconfig
index 16f44b9..76a7574 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/Kconfig
+++ b/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
diff --git a/drivers/net/ethernet/mellanox/mlxsw/Makefile b/drivers/net/ethernet/mellanox/mlxsw/Makefile
index c4c48ba..1459716 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/Makefile
+++ b/drivers/net/ethernet/mellanox/mlxsw/Makefile
@@ -14,7 +14,8 @@ mlxsw_switchx2-objs		:= switchx2.o
 obj-$(CONFIG_MLXSW_SPECTRUM)	+= mlxsw_spectrum.o
 mlxsw_spectrum-objs		:= spectrum.o spectrum_buffers.o \
 				   spectrum_switchdev.o spectrum_router.o \
-				   spectrum_kvdl.o
+				   spectrum_kvdl.o spectrum_acl.o \
+				   spectrum_acl_tcam.o
 mlxsw_spectrum-$(CONFIG_MLXSW_SPECTRUM_DCB)	+= spectrum_dcb.o
 obj-$(CONFIG_MLXSW_MINIMAL)	+= mlxsw_minimal.o
 mlxsw_minimal-objs		:= minimal.o
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
index 467aa52..b1d77e1 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
+++ b/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)
 {
@@ -3203,6 +3201,12 @@ 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 +3216,8 @@ 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 +3238,7 @@ 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 +3304,7 @@ 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;
 }
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.h b/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
index bc3efe1..cd9b4b2 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
+++ b/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>
  *
@@ -50,6 +50,8 @@
 
 #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 +264,8 @@ struct mlxsw_sp_router {
 	bool aborted;
 };
 
+struct mlxsw_sp_acl;
+
 struct mlxsw_sp {
 	struct {
 		struct list_head list;
@@ -291,6 +295,7 @@ 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 +378,7 @@ 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 +608,94 @@ 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;
+
 #endif
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c
new file mode 100644
index 0000000..8a18b3a
--- /dev/null
+++ b/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);
+}
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_tcam.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_tcam.c
new file mode 100644
index 0000000..a08cf10
--- /dev/null
+++ b/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,
+};
-- 
2.7.4

[patch net-next 18/19] sched: cls_flower: expose priority to offloading netdevice

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

The driver that offloads flower rules needs to know with which priority
user inserted the rules. So add this information into offload struct.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Ido Schimmel <idosch@mellanox.com>
---
 include/net/pkt_cls.h  | 1 +
 net/sched/cls_flower.c | 3 +++
 2 files changed, 4 insertions(+)

diff --git a/include/net/pkt_cls.h b/include/net/pkt_cls.h
index b43077e..dabb00a 100644
--- a/include/net/pkt_cls.h
+++ b/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;
diff --git a/net/sched/cls_flower.c b/net/sched/cls_flower.c
index 9e74b0f..e96ced5 100644
--- a/net/sched/cls_flower.c
+++ b/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;
 
-- 
2.7.4

[patch net-next 19/19] mlxsw: spectrum: Implement TC flower offload

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

Extend the existing setup_tc ndo call and allow to offload cls_flower
rules. Only limited set of dissector keys and actions are supported now.
Use previously introduced ACL infrastructure to offload cls_flower rules
to be processed in the HW.

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlxsw/Makefile       |   4 +-
 drivers/net/ethernet/mellanox/mlxsw/spectrum.c     |  15 +-
 drivers/net/ethernet/mellanox/mlxsw/spectrum.h     |   6 +
 .../net/ethernet/mellanox/mlxsw/spectrum_flower.c  | 309 +++++++++++++++++++++
 4 files changed, 331 insertions(+), 3 deletions(-)
 create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_flower.c

diff --git a/drivers/net/ethernet/mellanox/mlxsw/Makefile b/drivers/net/ethernet/mellanox/mlxsw/Makefile
index 1459716..6b6c30d 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/Makefile
+++ b/drivers/net/ethernet/mellanox/mlxsw/Makefile
@@ -14,8 +14,8 @@ mlxsw_switchx2-objs		:= switchx2.o
 obj-$(CONFIG_MLXSW_SPECTRUM)	+= mlxsw_spectrum.o
 mlxsw_spectrum-objs		:= spectrum.o spectrum_buffers.o \
 				   spectrum_switchdev.o spectrum_router.o \
-				   spectrum_kvdl.o spectrum_acl.o \
-				   spectrum_acl_tcam.o
+				   spectrum_kvdl.o spectrum_acl_tcam.o \
+				   spectrum_acl.o spectrum_flower.o
 mlxsw_spectrum-$(CONFIG_MLXSW_SPECTRUM_DCB)	+= spectrum_dcb.o
 obj-$(CONFIG_MLXSW_MINIMAL)	+= mlxsw_minimal.o
 mlxsw_minimal-objs		:= minimal.o
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
index b1d77e1..8a52c86 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
@@ -1355,7 +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,
@@ -1369,6 +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;
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.h b/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
index cd9b4b2..4d251e0 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
+++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum.h
@@ -47,6 +47,7 @@
 #include <linux/in6.h>
 #include <linux/notifier.h>
 #include <net/psample.h>
+#include <net/pkt_cls.h>
 
 #include "port.h"
 #include "core.h"
@@ -698,4 +699,9 @@ 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
diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_flower.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_flower.c
new file mode 100644
index 0000000..35b147a
--- /dev/null
+++ b/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);
+}
-- 
2.7.4

Re: [patch net-next 19/19] mlxsw: spectrum: Implement TC flower offload

[Florian Fainelli]

On 02/02/2017 07:12 AM, Jiri Pirko wrote:
> From: Jiri Pirko <jiri@mellanox.com>
> 
> Extend the existing setup_tc ndo call and allow to offload cls_flower
> rules. Only limited set of dissector keys and actions are supported now.
> Use previously introduced ACL infrastructure to offload cls_flower rules
> to be processed in the HW.
> 
> Signed-off-by: Jiri Pirko <jiri@mellanox.com>
> Reviewed-by: Ido Schimmel <idosch@mellanox.com>
> ---

> +	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;

You are not checking here that out_dev has the same netdev_ops pointer
(unlike the matchall case), is that expected?
-- 
Florian

Re: [patch net-next 00/19] mlxsw: Introduce TC Flower offload using TCAM

[Florian Fainelli]

On 02/02/2017 07:12 AM, Jiri Pirko wrote:
> From: Jiri Pirko <jiri@mellanox.com>
> 
> 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.

This looks really great (except all the input parameters validation
using flow_dissector keys, but there is already a thread about that, and
you are working with what you have)!

One thing I found missing with cls_flower is the ability to specify the
location of a rule, or if not specified have the switch driver return to
user which rule index was selected. Should we consider adding that so we
could finally move out of ethtool::rxfnc for NICs and other drivers?

Thanks

> 
> Jiri Pirko (19):
>   mlxsw: item: Add 8bit item helpers
>   mlxsw: item: Add helpers for getting pointer into payload for char
>     buffer item
>   mlxsw: reg: Add Policy-Engine ACL Register
>   mlxsw: reg: Add Policy-Engine ACL Group Table register
>   mlxsw: reg: Add Policy-Engine TCAM Allocation Register
>   mlxsw: reg: Add Policy-Engine TCAM Entry Register Version 2
>   mlxsw: reg: Add Policy-Engine Port Binding Table
>   mlxsw: reg: Add Policy-Engine Rules Copy Register
>   mlxsw: reg: Add Policy-Engine Policy Based Switching Register
>   mlxsw: reg: Add Policy-Engine Extended Flexible Action Register
>   mlxsw: core: Introduce flexible keys support
>   mlxsw: core: Introduce flexible actions support
>   mlxsw: spectrum: Introduce basic set of flexible key blocks
>   mlxsw: resources: Add ACL related resources
>   list: introduce list_for_each_entry_from_reverse helper
>   lib: Introduce priority array area manager
>   mlxsw: spectrum: Introduce ACL core with simple TCAM implementation
>   sched: cls_flower: expose priority to offloading netdevice
>   mlxsw: spectrum: Implement TC flower offload
> 
>  MAINTAINERS                                        |    8 +
>  drivers/net/ethernet/mellanox/mlxsw/Kconfig        |    1 +
>  drivers/net/ethernet/mellanox/mlxsw/Makefile       |    6 +-
>  .../mellanox/mlxsw/core_acl_flex_actions.c         |  685 +++++++++++++
>  .../mellanox/mlxsw/core_acl_flex_actions.h         |   66 ++
>  .../ethernet/mellanox/mlxsw/core_acl_flex_keys.c   |  475 +++++++++
>  .../ethernet/mellanox/mlxsw/core_acl_flex_keys.h   |  238 +++++
>  drivers/net/ethernet/mellanox/mlxsw/item.h         |   98 +-
>  drivers/net/ethernet/mellanox/mlxsw/reg.h          |  511 ++++++++-
>  drivers/net/ethernet/mellanox/mlxsw/resources.h    |   20 +-
>  drivers/net/ethernet/mellanox/mlxsw/spectrum.c     |   32 +-
>  drivers/net/ethernet/mellanox/mlxsw/spectrum.h     |  106 +-
>  drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c |  572 +++++++++++
>  .../mellanox/mlxsw/spectrum_acl_flex_keys.h        |  109 ++
>  .../ethernet/mellanox/mlxsw/spectrum_acl_tcam.c    | 1084 ++++++++++++++++++++
>  .../net/ethernet/mellanox/mlxsw/spectrum_flower.c  |  309 ++++++
>  include/linux/list.h                               |   13 +
>  include/linux/parman.h                             |   76 ++
>  include/net/pkt_cls.h                              |    1 +
>  lib/Kconfig                                        |    3 +
>  lib/Kconfig.debug                                  |   10 +
>  lib/Makefile                                       |    3 +
>  lib/parman.c                                       |  294 ++++++
>  lib/test_parman.c                                  |  395 +++++++
>  net/sched/cls_flower.c                             |    3 +
>  25 files changed, 5102 insertions(+), 16 deletions(-)
>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.c
>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.h
>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.c
>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.h
>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c
>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_flex_keys.h
>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_tcam.c
>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_flower.c
>  create mode 100644 include/linux/parman.h
>  create mode 100644 lib/parman.c
>  create mode 100644 lib/test_parman.c
> 


-- 
Florian

Re: [patch net-next 16/19] lib: Introduce priority array area manager

[Tom Herbert]

I have no idea what a "priority array area manager" is. Googling it
comes up with nothing and there are no comments in this whole patch
that either describe what it is or how any of the functions should be
used. Am I missing something that is supposed to be obvious?

Thanks,
Tom

On Thu, Feb 2, 2017 at 7:12 AM, Jiri Pirko <jiri@resnulli.us> wrote:
> From: Jiri Pirko <jiri@mellanox.com>
>
> This introduces a infrastructure for management of linear priority
> areas. Priority order in an array matters, however order of items inside
> a priority group does not matter.
>
> As an initial implementation, L-sort algorithm is used. It is quite
> trivial. More advanced algorithm called P-sort will be introduced as a
> follow-up. The infrastructure is prepared for other algos.
>
> Alongside this, a testing module is introduced as well.
>
> Signed-off-by: Jiri Pirko <jiri@mellanox.com>
> ---
>  MAINTAINERS            |   8 +
>  include/linux/parman.h |  76 ++++++++++
>  lib/Kconfig            |   3 +
>  lib/Kconfig.debug      |  10 ++
>  lib/Makefile           |   3 +
>  lib/parman.c           | 294 ++++++++++++++++++++++++++++++++++++
>  lib/test_parman.c      | 395 +++++++++++++++++++++++++++++++++++++++++++++++++
>  7 files changed, 789 insertions(+)
>  create mode 100644 include/linux/parman.h
>  create mode 100644 lib/parman.c
>  create mode 100644 lib/test_parman.c
>
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 300d2ec..626758b 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -9375,6 +9375,14 @@ 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
> diff --git a/include/linux/parman.h b/include/linux/parman.h
> new file mode 100644
> index 0000000..3c8cccc
> --- /dev/null
> +++ b/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
> diff --git a/lib/Kconfig b/lib/Kconfig
> index 260a80e..5d644f1 100644
> --- a/lib/Kconfig
> +++ b/lib/Kconfig
> @@ -550,4 +550,7 @@ config STACKDEPOT
>  config SBITMAP
>         bool
>
> +config PARMAN
> +       tristate "parman"
> +
>  endmenu
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index 15969ab..433a788 100644
> --- a/lib/Kconfig.debug
> +++ b/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
> diff --git a/lib/Makefile b/lib/Makefile
> index 7b3008d..1c039a4 100644
> --- a/lib/Makefile
> +++ b/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
> diff --git a/lib/parman.c b/lib/parman.c
> new file mode 100644
> index 0000000..5b6f4ec
> --- /dev/null
> +++ b/lib/parman.c
> @@ -0,0 +1,294 @@
> +/*
> + * 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,
> +};
> +
> +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);
> +
> +void parman_destroy(struct parman *parman)
> +{
> +       WARN_ON(!list_empty(&parman->prio_list));
> +       kfree(parman);
> +}
> +EXPORT_SYMBOL(parman_destroy);
> +
> +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);
> +
> +void parman_prio_fini(struct parman_prio *prio)
> +{
> +       WARN_ON(parman_prio_used(prio));
> +       list_del(&prio->list);
> +}
> +EXPORT_SYMBOL(parman_prio_fini);
> +
> +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);
> +
> +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");
> diff --git a/lib/test_parman.c b/lib/test_parman.c
> new file mode 100644
> index 0000000..fe9f3a7
> --- /dev/null
> +++ b/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");
> --
> 2.7.4
>

Re: [patch net-next 16/19] lib: Introduce priority array area manager

[Eric Dumazet]

On Thu, 2017-02-02 at 13:58 -0800, Tom Herbert wrote:
> I have no idea what a "priority array area manager" is. Googling it
> comes up with nothing and there are no comments in this whole patch
> that either describe what it is or how any of the functions should be
> used. Am I missing something that is supposed to be obvious?

Okay, but please trim your replies on netdev.

Thanks.

Re: [patch net-next 17/19] mlxsw: spectrum: Introduce ACL core with simple TCAM implementation

[David Miller]

From: Jiri Pirko <jiri@resnulli.us>
Date: Thu,  2 Feb 2017 16:12:57 +0100

> +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);
 ...
> +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);
 ...
> +static void mlxsw_sp_acl_tcam_group_id_put(struct mlxsw_sp_acl_tcam *tcam,
> +					   u16 id)
> +{
> +	clear_bit(id, tcam->used_groups);
> +}

Please use __set_bit() and __clear_bit() here since it seems very clear that
you do not require atomic operations.

Re: [patch net-next 16/19] lib: Introduce priority array area manager

[Joe Perches]

On Thu, 2017-02-02 at 16:12 +0100, Jiri Pirko wrote:
> From: Jiri Pirko <jiri@mellanox.com>
> 
> This introduces a infrastructure for management of linear priority
> areas. Priority order in an array matters, however order of items inside
> a priority group does not matter.
> 
> As an initial implementation, L-sort algorithm is used. It is quite
> trivial. More advanced algorithm called P-sort will be introduced as a
> follow-up. The infrastructure is prepared for other algos.
> 
> Alongside this, a testing module is introduced as well.
> 
> Signed-off-by: Jiri Pirko <jiri@mellanox.com>
> ---
>  MAINTAINERS            |   8 +
>  include/linux/parman.h |  76 ++++++++++
[]
> diff --git a/MAINTAINERS b/MAINTAINERS
[]
> @@ -9375,6 +9375,14 @@ F:	drivers/video/fbdev/sti*
>  F:	drivers/video/console/sti*
>  F:	drivers/video/logo/logo_parisc*
>  
> +PARMAN

This is not particularly descriptive.
Perhaps expand the section header a little like:

PARMAN - Linear priority array areas manager

> diff --git a/include/linux/parman.h b/include/linux/parman.h
[]
> @@ -0,0 +1,76 @@
> +/*
> + * include/linux/parman.h - Manager for linear priority array areas
[]
> diff --git a/lib/Kconfig b/lib/Kconfig
[]
> @@ -550,4 +550,7 @@ config STACKDEPOT
>  config SBITMAP
>  	bool
>  
> +config PARMAN
> +	tristate "parman"

help section?  Why should this be selected?

Re: [patch net-next 17/19] mlxsw: spectrum: Introduce ACL core with simple TCAM implementation

[Jiri Pirko]

Thu, Feb 02, 2017 at 11:34:48PM CET, davem@davemloft.net wrote:
>From: Jiri Pirko <jiri@resnulli.us>
>Date: Thu,  2 Feb 2017 16:12:57 +0100
>
>> +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);
> ...
>> +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);
> ...
>> +static void mlxsw_sp_acl_tcam_group_id_put(struct mlxsw_sp_acl_tcam *tcam,
>> +					   u16 id)
>> +{
>> +	clear_bit(id, tcam->used_groups);
>> +}
>
>Please use __set_bit() and __clear_bit() here since it seems very clear that
>you do not require atomic operations.

Will do.

Re: [patch net-next 00/19] mlxsw: Introduce TC Flower offload using TCAM

[Jiri Pirko]

Thu, Feb 02, 2017 at 10:40:42PM CET, f.fainelli@gmail.com wrote:
>On 02/02/2017 07:12 AM, Jiri Pirko wrote:
>> From: Jiri Pirko <jiri@mellanox.com>
>> 
>> 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.
>
>This looks really great (except all the input parameters validation
>using flow_dissector keys, but there is already a thread about that, and
>you are working with what you have)!
>
>One thing I found missing with cls_flower is the ability to specify the
>location of a rule, or if not specified have the switch driver return to
>user which rule index was selected. Should we consider adding that so we
>could finally move out of ethtool::rxfnc for NICs and other drivers?

What exactly do you mean by "location"? When you add the rule, you
specify "pref/prio". Rules with same prio have always indentical mask
and go into one hashtable. For mlxsw offload, the rule with prio goes
down to driver and the chunks of rules with same priority are arranged
in TCAM accordingly (using parman).



>
>Thanks
>
>> 
>> Jiri Pirko (19):
>>   mlxsw: item: Add 8bit item helpers
>>   mlxsw: item: Add helpers for getting pointer into payload for char
>>     buffer item
>>   mlxsw: reg: Add Policy-Engine ACL Register
>>   mlxsw: reg: Add Policy-Engine ACL Group Table register
>>   mlxsw: reg: Add Policy-Engine TCAM Allocation Register
>>   mlxsw: reg: Add Policy-Engine TCAM Entry Register Version 2
>>   mlxsw: reg: Add Policy-Engine Port Binding Table
>>   mlxsw: reg: Add Policy-Engine Rules Copy Register
>>   mlxsw: reg: Add Policy-Engine Policy Based Switching Register
>>   mlxsw: reg: Add Policy-Engine Extended Flexible Action Register
>>   mlxsw: core: Introduce flexible keys support
>>   mlxsw: core: Introduce flexible actions support
>>   mlxsw: spectrum: Introduce basic set of flexible key blocks
>>   mlxsw: resources: Add ACL related resources
>>   list: introduce list_for_each_entry_from_reverse helper
>>   lib: Introduce priority array area manager
>>   mlxsw: spectrum: Introduce ACL core with simple TCAM implementation
>>   sched: cls_flower: expose priority to offloading netdevice
>>   mlxsw: spectrum: Implement TC flower offload
>> 
>>  MAINTAINERS                                        |    8 +
>>  drivers/net/ethernet/mellanox/mlxsw/Kconfig        |    1 +
>>  drivers/net/ethernet/mellanox/mlxsw/Makefile       |    6 +-
>>  .../mellanox/mlxsw/core_acl_flex_actions.c         |  685 +++++++++++++
>>  .../mellanox/mlxsw/core_acl_flex_actions.h         |   66 ++
>>  .../ethernet/mellanox/mlxsw/core_acl_flex_keys.c   |  475 +++++++++
>>  .../ethernet/mellanox/mlxsw/core_acl_flex_keys.h   |  238 +++++
>>  drivers/net/ethernet/mellanox/mlxsw/item.h         |   98 +-
>>  drivers/net/ethernet/mellanox/mlxsw/reg.h          |  511 ++++++++-
>>  drivers/net/ethernet/mellanox/mlxsw/resources.h    |   20 +-
>>  drivers/net/ethernet/mellanox/mlxsw/spectrum.c     |   32 +-
>>  drivers/net/ethernet/mellanox/mlxsw/spectrum.h     |  106 +-
>>  drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c |  572 +++++++++++
>>  .../mellanox/mlxsw/spectrum_acl_flex_keys.h        |  109 ++
>>  .../ethernet/mellanox/mlxsw/spectrum_acl_tcam.c    | 1084 ++++++++++++++++++++
>>  .../net/ethernet/mellanox/mlxsw/spectrum_flower.c  |  309 ++++++
>>  include/linux/list.h                               |   13 +
>>  include/linux/parman.h                             |   76 ++
>>  include/net/pkt_cls.h                              |    1 +
>>  lib/Kconfig                                        |    3 +
>>  lib/Kconfig.debug                                  |   10 +
>>  lib/Makefile                                       |    3 +
>>  lib/parman.c                                       |  294 ++++++
>>  lib/test_parman.c                                  |  395 +++++++
>>  net/sched/cls_flower.c                             |    3 +
>>  25 files changed, 5102 insertions(+), 16 deletions(-)
>>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.c
>>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_actions.h
>>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.c
>>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/core_acl_flex_keys.h
>>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl.c
>>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_flex_keys.h
>>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_acl_tcam.c
>>  create mode 100644 drivers/net/ethernet/mellanox/mlxsw/spectrum_flower.c
>>  create mode 100644 include/linux/parman.h
>>  create mode 100644 lib/parman.c
>>  create mode 100644 lib/test_parman.c
>> 
>
>
>-- 
>Florian

Re: [patch net-next 16/19] lib: Introduce priority array area manager

[Jiri Pirko]

Thu, Feb 02, 2017 at 11:51:10PM CET, joe@perches.com wrote:
>On Thu, 2017-02-02 at 16:12 +0100, Jiri Pirko wrote:
>> From: Jiri Pirko <jiri@mellanox.com>
>> 
>> This introduces a infrastructure for management of linear priority
>> areas. Priority order in an array matters, however order of items inside
>> a priority group does not matter.
>> 
>> As an initial implementation, L-sort algorithm is used. It is quite
>> trivial. More advanced algorithm called P-sort will be introduced as a
>> follow-up. The infrastructure is prepared for other algos.
>> 
>> Alongside this, a testing module is introduced as well.
>> 
>> Signed-off-by: Jiri Pirko <jiri@mellanox.com>
>> ---
>>  MAINTAINERS            |   8 +
>>  include/linux/parman.h |  76 ++++++++++
>[]
>> diff --git a/MAINTAINERS b/MAINTAINERS
>[]
>> @@ -9375,6 +9375,14 @@ F:	drivers/video/fbdev/sti*
>>  F:	drivers/video/console/sti*
>>  F:	drivers/video/logo/logo_parisc*
>>  
>> +PARMAN
>
>This is not particularly descriptive.
>Perhaps expand the section header a little like:
>
>PARMAN - Linear priority array areas manager

Okay


>
>> diff --git a/include/linux/parman.h b/include/linux/parman.h
>[]
>> @@ -0,0 +1,76 @@
>> +/*
>> + * include/linux/parman.h - Manager for linear priority array areas
>[]
>> diff --git a/lib/Kconfig b/lib/Kconfig
>[]
>> @@ -550,4 +550,7 @@ config STACKDEPOT
>>  config SBITMAP
>>  	bool
>>  
>> +config PARMAN
>> +	tristate "parman"
>
>help section?  Why should this be selected?

I did not put help on purpose, as for other libs. User should not never
need to select this directly because as standalone, it has no sense to
have it enabled. This is always selected by users - only mlxsw for now.


>

Re: [patch net-next 16/19] lib: Introduce priority array area manager

[Jiri Pirko]

Thu, Feb 02, 2017 at 10:58:15PM CET, tom@herbertland.com wrote:
>I have no idea what a "priority array area manager" is. Googling it
>comes up with nothing and there are no comments in this whole patch
>that either describe what it is or how any of the functions should be
>used. Am I missing something that is supposed to be obvious?

This lib manages array areas with chunks of rules with same priority.
Consider folowing example:

entry 1 with prio 10
entry 2 with prio 10
entry 3 with prio 10
entry 4 with prio 20
entry 5 with prio 20
entry 6 with prio 20
entry 7 with prio 30
entry 8 with prio 30
entry 9 with prio 30

In this example there are 3 priority chunks. The order of the prio
matters, however the order within a single priority chunk does not
matter. So the same array would be ordered as follows:

entry 2 with prio 10
entry 3 with prio 10
entry 1 with prio 10
entry 5 with prio 20
entry 4 with prio 20
entry 6 with prio 20
entry 9 with prio 30
entry 8 with prio 30
entry 7 with prio 30

Our usecase of this is ordering of entries within TCAM regions.
I could put it directly into mlxsw driver, yet I thought that this is
solving a generic problem and could be re-used by other drivers.
Therefore I decided to put it to lib, with test module.

I will add some description to the code.


>
>Thanks,
>Tom
>
>On Thu, Feb 2, 2017 at 7:12 AM, Jiri Pirko <jiri@resnulli.us> wrote:
>> From: Jiri Pirko <jiri@mellanox.com>
>>
>> This introduces a infrastructure for management of linear priority
>> areas. Priority order in an array matters, however order of items inside
>> a priority group does not matter.
>>
>> As an initial implementation, L-sort algorithm is used. It is quite
>> trivial. More advanced algorithm called P-sort will be introduced as a
>> follow-up. The infrastructure is prepared for other algos.
>>
>> Alongside this, a testing module is introduced as well.
>>
>> Signed-off-by: Jiri Pirko <jiri@mellanox.com>
>> ---
>>  MAINTAINERS            |   8 +
>>  include/linux/parman.h |  76 ++++++++++
>>  lib/Kconfig            |   3 +
>>  lib/Kconfig.debug      |  10 ++
>>  lib/Makefile           |   3 +
>>  lib/parman.c           | 294 ++++++++++++++++++++++++++++++++++++
>>  lib/test_parman.c      | 395 +++++++++++++++++++++++++++++++++++++++++++++++++
>>  7 files changed, 789 insertions(+)
>>  create mode 100644 include/linux/parman.h
>>  create mode 100644 lib/parman.c
>>  create mode 100644 lib/test_parman.c
>>
>> diff --git a/MAINTAINERS b/MAINTAINERS
>> index 300d2ec..626758b 100644
>> --- a/MAINTAINERS
>> +++ b/MAINTAINERS
>> @@ -9375,6 +9375,14 @@ 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
>> diff --git a/include/linux/parman.h b/include/linux/parman.h
>> new file mode 100644
>> index 0000000..3c8cccc
>> --- /dev/null
>> +++ b/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
>> diff --git a/lib/Kconfig b/lib/Kconfig
>> index 260a80e..5d644f1 100644
>> --- a/lib/Kconfig
>> +++ b/lib/Kconfig
>> @@ -550,4 +550,7 @@ config STACKDEPOT
>>  config SBITMAP
>>         bool
>>
>> +config PARMAN
>> +       tristate "parman"
>> +
>>  endmenu
>> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
>> index 15969ab..433a788 100644
>> --- a/lib/Kconfig.debug
>> +++ b/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
>> diff --git a/lib/Makefile b/lib/Makefile
>> index 7b3008d..1c039a4 100644
>> --- a/lib/Makefile
>> +++ b/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
>> diff --git a/lib/parman.c b/lib/parman.c
>> new file mode 100644
>> index 0000000..5b6f4ec
>> --- /dev/null
>> +++ b/lib/parman.c
>> @@ -0,0 +1,294 @@
>> +/*
>> + * 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,
>> +};
>> +
>> +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);
>> +
>> +void parman_destroy(struct parman *parman)
>> +{
>> +       WARN_ON(!list_empty(&parman->prio_list));
>> +       kfree(parman);
>> +}
>> +EXPORT_SYMBOL(parman_destroy);
>> +
>> +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);
>> +
>> +void parman_prio_fini(struct parman_prio *prio)
>> +{
>> +       WARN_ON(parman_prio_used(prio));
>> +       list_del(&prio->list);
>> +}
>> +EXPORT_SYMBOL(parman_prio_fini);
>> +
>> +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);
>> +
>> +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");
>> diff --git a/lib/test_parman.c b/lib/test_parman.c
>> new file mode 100644
>> index 0000000..fe9f3a7
>> --- /dev/null
>> +++ b/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");
>> --
>> 2.7.4
>>

Re: [patch net-next 19/19] mlxsw: spectrum: Implement TC flower offload

[Jiri Pirko]

Thu, Feb 02, 2017 at 10:37:23PM CET, f.fainelli@gmail.com wrote:
>On 02/02/2017 07:12 AM, Jiri Pirko wrote:
>> From: Jiri Pirko <jiri@mellanox.com>
>> 
>> Extend the existing setup_tc ndo call and allow to offload cls_flower
>> rules. Only limited set of dissector keys and actions are supported now.
>> Use previously introduced ACL infrastructure to offload cls_flower rules
>> to be processed in the HW.
>> 
>> Signed-off-by: Jiri Pirko <jiri@mellanox.com>
>> Reviewed-by: Ido Schimmel <idosch@mellanox.com>
>> ---
>
>> +	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;
>
>You are not checking here that out_dev has the same netdev_ops pointer
>(unlike the matchall case), is that expected?

This is done later on in mlxsw_sp_acl_rulei_act_fwd.

Re: [patch net-next 00/19] mlxsw: Introduce TC Flower offload using TCAM

[Florian Fainelli]

On 02/02/2017 10:58 PM, Jiri Pirko wrote:
> Thu, Feb 02, 2017 at 10:40:42PM CET, f.fainelli@gmail.com wrote:
>> On 02/02/2017 07:12 AM, Jiri Pirko wrote:
>>> From: Jiri Pirko <jiri@mellanox.com>
>>>
>>> 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.
>>
>> This looks really great (except all the input parameters validation
>> using flow_dissector keys, but there is already a thread about that, and
>> you are working with what you have)!
>>
>> One thing I found missing with cls_flower is the ability to specify the
>> location of a rule, or if not specified have the switch driver return to
>> user which rule index was selected. Should we consider adding that so we
>> could finally move out of ethtool::rxfnc for NICs and other drivers?
> 
> What exactly do you mean by "location"? When you add the rule, you
> specify "pref/prio". Rules with same prio have always indentical mask
> and go into one hashtable. For mlxsw offload, the rule with prio goes
> down to driver and the chunks of rules with same priority are arranged
> in TCAM accordingly (using parman).

location parameter in ethtool_rx_flow_spec allows you to either let the
driver decide where to place the rule in its TCAM (RX_CLS_LOC_ANY), or
you can influence this decision by setting it to 0 ... MAX.

In the Broadcom SF2 switch case, this rule location is used during
packet matching, for instance:

- you program a rule to match a given 5-tuple on Port 0, queue 0 that
redirects the packet to Port 7 queue 8, with a rule ID of X (that is
both position in TCAM, and unique global identifier)

- a packet is matched, switch classifies packets, find a matching rule
and the packet ingresses Port 7 with Broadcom tag (per-packet metadata)
and the classification ID is set to X indicating that the packet you
receive has been matched by a rule

The driver knows how many rules are available (HW limitation), and can
either do a first unused rule location, or let the user specify where it
wants it (look at bcm_sf2_cfp.c).

The part that seems to be possibly missing here is that if the driver
did decide where to place this rule in its TCAM, it can be useful to
return that to user-space, so it can be communicated to other parts of
the system.

Of course, this only works if the programming paradigm is that rules are
identified by some kind of location...
-- 
Florian

Re: [patch net-next 00/19] mlxsw: Introduce TC Flower offload using TCAM

[Jiri Pirko]

Fri, Feb 03, 2017 at 06:57:25PM CET, f.fainelli@gmail.com wrote:
>On 02/02/2017 10:58 PM, Jiri Pirko wrote:
>> Thu, Feb 02, 2017 at 10:40:42PM CET, f.fainelli@gmail.com wrote:
>>> On 02/02/2017 07:12 AM, Jiri Pirko wrote:
>>>> From: Jiri Pirko <jiri@mellanox.com>
>>>>
>>>> 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.
>>>
>>> This looks really great (except all the input parameters validation
>>> using flow_dissector keys, but there is already a thread about that, and
>>> you are working with what you have)!
>>>
>>> One thing I found missing with cls_flower is the ability to specify the
>>> location of a rule, or if not specified have the switch driver return to
>>> user which rule index was selected. Should we consider adding that so we
>>> could finally move out of ethtool::rxfnc for NICs and other drivers?
>> 
>> What exactly do you mean by "location"? When you add the rule, you
>> specify "pref/prio". Rules with same prio have always indentical mask
>> and go into one hashtable. For mlxsw offload, the rule with prio goes
>> down to driver and the chunks of rules with same priority are arranged
>> in TCAM accordingly (using parman).
>
>location parameter in ethtool_rx_flow_spec allows you to either let the
>driver decide where to place the rule in its TCAM (RX_CLS_LOC_ANY), or
>you can influence this decision by setting it to 0 ... MAX.

Got it. There is no concept of this in tc/cls_flower.
	
Also, our TCAM is not realy one long array of rules, but a hierarchy
(see spectrum_acl_tcam.c) So it really makes no sense for user to allow
to specify anything as he has no clue how the driver will store the rule
inside the TCAM.


>
>In the Broadcom SF2 switch case, this rule location is used during
>packet matching, for instance:
>
>- you program a rule to match a given 5-tuple on Port 0, queue 0 that
>redirects the packet to Port 7 queue 8, with a rule ID of X (that is
>both position in TCAM, and unique global identifier)
>
>- a packet is matched, switch classifies packets, find a matching rule
>and the packet ingresses Port 7 with Broadcom tag (per-packet metadata)
>and the classification ID is set to X indicating that the packet you
>receive has been matched by a rule
>
>The driver knows how many rules are available (HW limitation), and can
>either do a first unused rule location, or let the user specify where it
>wants it (look at bcm_sf2_cfp.c).
>
>The part that seems to be possibly missing here is that if the driver
>did decide where to place this rule in its TCAM, it can be useful to
>return that to user-space, so it can be communicated to other parts of
>the system.

We are now working on a generic debugging (RO!) extension to devlink
which allows user to see the inner tables of the ASIC and how they are
configured and filled up. The TCAM area will be exposed byt this
interface as well.


>
>Of course, this only works if the programming paradigm is that rules are
>identified by some kind of location...
>-- 
>Florian