Merge branch 'mlxsw-ERP-sharing-multiple-masks'

Ido Schimmel says:

====================
mlxsw: spectrum: acl: Introduce ERP sharing by multiple masks

Jiri says:

The Spectrum-2 hardware has limitation number of ERPs per-region. In
order to accommodate more masks than number of ERPs, the hardware
supports to insert rules with delta bits. By that, the rules with masks
that differ in up-to 8 consecutive bits can share the same ERP.

Patches 1 and 2 fix couple of issues that would appear in existing
selftests after adding delta support

Patch 3 introduces a generic object aggregation library. Now it is
static, but it will get extended for recalculation of aggregations in
the future in order to reach more optimal aggregation.

Patch 4 just simply converts existing ERP code to use the objagg library
instead of a rhashtable.

Patches 5-9 do more or less small changes to prepare ground for the last
patch.

Patch 10 fills-up delta callbacks of objagg library and utilizes the
delta bits for rule insertion.

The last patch adds selftest to test the mlxsw Spectrum-2 delta flows.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

MAINTAINERS

@@ -10679,6 +10679,14 @@ L:	linux-nfc@lists.01.org (moderated for non-subscribers)
 S:	Supported
 F:	drivers/nfc/nxp-nci
 
+OBJAGG
+M:	Jiri Pirko <jiri@mellanox.com>
+L:	netdev@vger.kernel.org
+S:	Supported
+F:	lib/objagg.c
+F:	lib/test_objagg.c
+F:	include/linux/objagg.h
+
 OBJTOOL
 M:	Josh Poimboeuf <jpoimboe@redhat.com>
 M:	Peter Zijlstra <peterz@infradead.org>

drivers/net/ethernet/mellanox/mlxsw/Kconfig

@@ -80,6 +80,7 @@ config MLXSW_SPECTRUM
 	depends on IPV6_GRE || IPV6_GRE=n
 	select GENERIC_ALLOCATOR
 	select PARMAN
+	select OBJAGG
 	select MLXFW
 	default m
 	---help---

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

@@ -426,15 +426,17 @@ mlxsw_sp_afk_encode_one(const struct mlxsw_afk_element_inst *elinst,
 void mlxsw_afk_encode(struct mlxsw_afk *mlxsw_afk,
 		      struct mlxsw_afk_key_info *key_info,
 		      struct mlxsw_afk_element_values *values,
-		      char *key, char *mask, int block_start, int block_end)
+		      char *key, char *mask)
 {
+	unsigned int blocks_count =
+			mlxsw_afk_key_info_blocks_count_get(key_info);
 	char block_mask[MLXSW_SP_AFK_KEY_BLOCK_MAX_SIZE];
 	char block_key[MLXSW_SP_AFK_KEY_BLOCK_MAX_SIZE];
 	const struct mlxsw_afk_element_inst *elinst;
 	enum mlxsw_afk_element element;
 	int block_index, i;
 
-	for (i = block_start; i <= block_end; i++) {
+	for (i = 0; i < blocks_count; i++) {
 		memset(block_key, 0, MLXSW_SP_AFK_KEY_BLOCK_MAX_SIZE);
 		memset(block_mask, 0, MLXSW_SP_AFK_KEY_BLOCK_MAX_SIZE);
 
@@ -451,10 +453,18 @@ void mlxsw_afk_encode(struct mlxsw_afk *mlxsw_afk,
 						values->storage.mask);
 		}
 
-		if (key)
-			mlxsw_afk->ops->encode_block(block_key, i, key);
-		if (mask)
-			mlxsw_afk->ops->encode_block(block_mask, i, mask);
+		mlxsw_afk->ops->encode_block(key, i, block_key);
+		mlxsw_afk->ops->encode_block(mask, i, block_mask);
 	}
 }
 EXPORT_SYMBOL(mlxsw_afk_encode);
+
+void mlxsw_afk_clear(struct mlxsw_afk *mlxsw_afk, char *key,
+		     int block_start, int block_end)
+{
+	int i;
+
+	for (i = block_start; i <= block_end; i++)
+		mlxsw_afk->ops->clear_block(key, i);
+}
+EXPORT_SYMBOL(mlxsw_afk_clear);

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

@@ -188,7 +188,8 @@ struct mlxsw_afk;
 struct mlxsw_afk_ops {
 	const struct mlxsw_afk_block *blocks;
 	unsigned int blocks_count;
-	void (*encode_block)(char *block, int block_index, char *output);
+	void (*encode_block)(char *output, int block_index, char *block);
+	void (*clear_block)(char *output, int block_index);
 };
 
 struct mlxsw_afk *mlxsw_afk_create(unsigned int max_blocks,
@@ -228,6 +229,8 @@ void mlxsw_afk_values_add_buf(struct mlxsw_afk_element_values *values,
 void mlxsw_afk_encode(struct mlxsw_afk *mlxsw_afk,
 		      struct mlxsw_afk_key_info *key_info,
 		      struct mlxsw_afk_element_values *values,
-		      char *key, char *mask, int block_start, int block_end);
+		      char *key, char *mask);
+void mlxsw_afk_clear(struct mlxsw_afk *mlxsw_afk, char *key,
+		     int block_start, int block_end);
 
 #endif

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

@@ -2834,8 +2834,9 @@ static inline void mlxsw_reg_ptce3_pack(char *payload, bool valid,
 					u32 priority,
 					const char *tcam_region_info,
 					const char *key, u8 erp_id,
-					bool large_exists, u32 lkey_id,
-					u32 action_pointer)
+					u16 delta_start, u8 delta_mask,
+					u8 delta_value, bool large_exists,
+					u32 lkey_id, u32 action_pointer)
 {
 	MLXSW_REG_ZERO(ptce3, payload);
 	mlxsw_reg_ptce3_v_set(payload, valid);
@@ -2844,6 +2845,9 @@ static inline void mlxsw_reg_ptce3_pack(char *payload, bool valid,
 	mlxsw_reg_ptce3_tcam_region_info_memcpy_to(payload, tcam_region_info);
 	mlxsw_reg_ptce3_flex2_key_blocks_memcpy_to(payload, key);
 	mlxsw_reg_ptce3_erp_id_set(payload, erp_id);
+	mlxsw_reg_ptce3_delta_start_set(payload, delta_start);
+	mlxsw_reg_ptce3_delta_mask_set(payload, delta_mask);
+	mlxsw_reg_ptce3_delta_value_set(payload, delta_value);
 	mlxsw_reg_ptce3_large_exists_set(payload, large_exists);
 	mlxsw_reg_ptce3_large_entry_key_id_set(payload, lkey_id);
 	mlxsw_reg_ptce3_action_pointer_set(payload, action_pointer);

drivers/net/ethernet/mellanox/mlxsw/spectrum2_acl_tcam.c

@@ -34,15 +34,15 @@ mlxsw_sp2_acl_ctcam_region_entry_insert(struct mlxsw_sp_acl_ctcam_region *cregio
 {
 	struct mlxsw_sp_acl_atcam_region *aregion;
 	struct mlxsw_sp_acl_atcam_entry *aentry;
-	struct mlxsw_sp_acl_erp *erp;
+	struct mlxsw_sp_acl_erp_mask *erp_mask;
 
 	aregion = mlxsw_sp_acl_tcam_cregion_aregion(cregion);
 	aentry = mlxsw_sp_acl_tcam_centry_aentry(centry);
 
-	erp = mlxsw_sp_acl_erp_get(aregion, mask, true);
-	if (IS_ERR(erp))
-		return PTR_ERR(erp);
-	aentry->erp = erp;
+	erp_mask = mlxsw_sp_acl_erp_mask_get(aregion, mask, true);
+	if (IS_ERR(erp_mask))
+		return PTR_ERR(erp_mask);
+	aentry->erp_mask = erp_mask;
 
 	return 0;
 }
@@ -57,7 +57,7 @@ mlxsw_sp2_acl_ctcam_region_entry_remove(struct mlxsw_sp_acl_ctcam_region *cregio
 	aregion = mlxsw_sp_acl_tcam_cregion_aregion(cregion);
 	aentry = mlxsw_sp_acl_tcam_centry_aentry(centry);
 
-	mlxsw_sp_acl_erp_put(aregion, aentry->erp);
+	mlxsw_sp_acl_erp_mask_put(aregion, aentry->erp_mask);
 }
 
 static const struct mlxsw_sp_acl_ctcam_region_ops

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

@@ -14,8 +14,8 @@
 #include "spectrum_acl_tcam.h"
 #include "core_acl_flex_keys.h"
 
-#define MLXSW_SP_ACL_ATCAM_LKEY_ID_BLOCK_START	6
-#define MLXSW_SP_ACL_ATCAM_LKEY_ID_BLOCK_END	11
+#define MLXSW_SP_ACL_ATCAM_LKEY_ID_BLOCK_CLEAR_START	0
+#define MLXSW_SP_ACL_ATCAM_LKEY_ID_BLOCK_CLEAR_END	5
 
 struct mlxsw_sp_acl_atcam_lkey_id_ht_key {
 	char enc_key[MLXSW_REG_PTCEX_FLEX_KEY_BLOCKS_LEN]; /* MSB blocks */
@@ -34,7 +34,7 @@ struct mlxsw_sp_acl_atcam_region_ops {
 	void (*fini)(struct mlxsw_sp_acl_atcam_region *aregion);
 	struct mlxsw_sp_acl_atcam_lkey_id *
 		(*lkey_id_get)(struct mlxsw_sp_acl_atcam_region *aregion,
-			       struct mlxsw_sp_acl_rule_info *rulei, u8 erp_id);
+			       char *enc_key, u8 erp_id);
 	void (*lkey_id_put)(struct mlxsw_sp_acl_atcam_region *aregion,
 			    struct mlxsw_sp_acl_atcam_lkey_id *lkey_id);
 };
@@ -64,7 +64,7 @@ static const struct rhashtable_params mlxsw_sp_acl_atcam_entries_ht_params = {
 static bool
 mlxsw_sp_acl_atcam_is_centry(const struct mlxsw_sp_acl_atcam_entry *aentry)
 {
-	return mlxsw_sp_acl_erp_is_ctcam_erp(aentry->erp);
+	return mlxsw_sp_acl_erp_mask_is_ctcam(aentry->erp_mask);
 }
 
 static int
@@ -90,8 +90,7 @@ mlxsw_sp_acl_atcam_region_generic_fini(struct mlxsw_sp_acl_atcam_region *aregion
 
 static struct mlxsw_sp_acl_atcam_lkey_id *
 mlxsw_sp_acl_atcam_generic_lkey_id_get(struct mlxsw_sp_acl_atcam_region *aregion,
-				       struct mlxsw_sp_acl_rule_info *rulei,
-				       u8 erp_id)
+				       char *enc_key, u8 erp_id)
 {
 	struct mlxsw_sp_acl_atcam_region_generic *region_generic;
 
@@ -220,8 +219,7 @@ mlxsw_sp_acl_atcam_lkey_id_destroy(struct mlxsw_sp_acl_atcam_region *aregion,
 
 static struct mlxsw_sp_acl_atcam_lkey_id *
 mlxsw_sp_acl_atcam_12kb_lkey_id_get(struct mlxsw_sp_acl_atcam_region *aregion,
-				    struct mlxsw_sp_acl_rule_info *rulei,
-				    u8 erp_id)
+				    char *enc_key, u8 erp_id)
 {
 	struct mlxsw_sp_acl_atcam_region_12kb *region_12kb = aregion->priv;
 	struct mlxsw_sp_acl_tcam_region *region = aregion->region;
@@ -230,9 +228,10 @@ mlxsw_sp_acl_atcam_12kb_lkey_id_get(struct mlxsw_sp_acl_atcam_region *aregion,
 	struct mlxsw_afk *afk = mlxsw_sp_acl_afk(mlxsw_sp->acl);
 	struct mlxsw_sp_acl_atcam_lkey_id *lkey_id;
 
-	mlxsw_afk_encode(afk, region->key_info, &rulei->values, ht_key.enc_key,
-			 NULL, MLXSW_SP_ACL_ATCAM_LKEY_ID_BLOCK_START,
-			 MLXSW_SP_ACL_ATCAM_LKEY_ID_BLOCK_END);
+	memcpy(ht_key.enc_key, enc_key, sizeof(ht_key.enc_key));
+	mlxsw_afk_clear(afk, ht_key.enc_key,
+			MLXSW_SP_ACL_ATCAM_LKEY_ID_BLOCK_CLEAR_START,
+			MLXSW_SP_ACL_ATCAM_LKEY_ID_BLOCK_CLEAR_END);
 	ht_key.erp_id = erp_id;
 	lkey_id = rhashtable_lookup_fast(&region_12kb->lkey_ht, &ht_key,
 					 mlxsw_sp_acl_atcam_lkey_id_ht_params);
@@ -379,7 +378,7 @@ mlxsw_sp_acl_atcam_region_entry_insert(struct mlxsw_sp *mlxsw_sp,
 				       struct mlxsw_sp_acl_rule_info *rulei)
 {
 	struct mlxsw_sp_acl_tcam_region *region = aregion->region;
-	u8 erp_id = mlxsw_sp_acl_erp_id(aentry->erp);
+	u8 erp_id = mlxsw_sp_acl_erp_mask_erp_id(aentry->erp_mask);
 	struct mlxsw_sp_acl_atcam_lkey_id *lkey_id;
 	char ptce3_pl[MLXSW_REG_PTCE3_LEN];
 	u32 kvdl_index, priority;
@@ -389,7 +388,8 @@ mlxsw_sp_acl_atcam_region_entry_insert(struct mlxsw_sp *mlxsw_sp,
 	if (err)
 		return err;
 
-	lkey_id = aregion->ops->lkey_id_get(aregion, rulei, erp_id);
+	lkey_id = aregion->ops->lkey_id_get(aregion, aentry->ht_key.enc_key,
+					    erp_id);
 	if (IS_ERR(lkey_id))
 		return PTR_ERR(lkey_id);
 	aentry->lkey_id = lkey_id;
@@ -398,6 +398,9 @@ mlxsw_sp_acl_atcam_region_entry_insert(struct mlxsw_sp *mlxsw_sp,
 	mlxsw_reg_ptce3_pack(ptce3_pl, true, MLXSW_REG_PTCE3_OP_WRITE_WRITE,
 			     priority, region->tcam_region_info,
 			     aentry->ht_key.enc_key, erp_id,
+			     aentry->delta_info.start,
+			     aentry->delta_info.mask,
+			     aentry->delta_info.value,
 			     refcount_read(&lkey_id->refcnt) != 1, lkey_id->id,
 			     kvdl_index);
 	err = mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(ptce3), ptce3_pl);
@@ -418,12 +421,17 @@ mlxsw_sp_acl_atcam_region_entry_remove(struct mlxsw_sp *mlxsw_sp,
 {
 	struct mlxsw_sp_acl_atcam_lkey_id *lkey_id = aentry->lkey_id;
 	struct mlxsw_sp_acl_tcam_region *region = aregion->region;
-	u8 erp_id = mlxsw_sp_acl_erp_id(aentry->erp);
+	u8 erp_id = mlxsw_sp_acl_erp_mask_erp_id(aentry->erp_mask);
+	char *enc_key = aentry->ht_key.enc_key;
 	char ptce3_pl[MLXSW_REG_PTCE3_LEN];
 
 	mlxsw_reg_ptce3_pack(ptce3_pl, false, MLXSW_REG_PTCE3_OP_WRITE_WRITE, 0,
-			     region->tcam_region_info, aentry->ht_key.enc_key,
-			     erp_id, refcount_read(&lkey_id->refcnt) != 1,
+			     region->tcam_region_info,
+			     enc_key, erp_id,
+			     aentry->delta_info.start,
+			     aentry->delta_info.mask,
+			     aentry->delta_info.value,
+			     refcount_read(&lkey_id->refcnt) != 1,
 			     lkey_id->id, 0);
 	mlxsw_reg_write(mlxsw_sp->core, MLXSW_REG(ptce3), ptce3_pl);
 	aregion->ops->lkey_id_put(aregion, lkey_id);
@@ -438,19 +446,30 @@ __mlxsw_sp_acl_atcam_entry_add(struct mlxsw_sp *mlxsw_sp,
 	struct mlxsw_sp_acl_tcam_region *region = aregion->region;
 	char mask[MLXSW_REG_PTCEX_FLEX_KEY_BLOCKS_LEN] = { 0 };
 	struct mlxsw_afk *afk = mlxsw_sp_acl_afk(mlxsw_sp->acl);
-	struct mlxsw_sp_acl_erp *erp;
-	unsigned int blocks_count;
+	const struct mlxsw_sp_acl_erp_delta *delta;
+	struct mlxsw_sp_acl_erp_mask *erp_mask;
 	int err;
 
-	blocks_count = mlxsw_afk_key_info_blocks_count_get(region->key_info);
 	mlxsw_afk_encode(afk, region->key_info, &rulei->values,
-			 aentry->ht_key.enc_key, mask, 0, blocks_count - 1);
-
-	erp = mlxsw_sp_acl_erp_get(aregion, mask, false);
-	if (IS_ERR(erp))
-		return PTR_ERR(erp);
-	aentry->erp = erp;
-	aentry->ht_key.erp_id = mlxsw_sp_acl_erp_id(erp);
+			 aentry->full_enc_key, mask);
+
+	erp_mask = mlxsw_sp_acl_erp_mask_get(aregion, mask, false);
+	if (IS_ERR(erp_mask))
+		return PTR_ERR(erp_mask);
+	aentry->erp_mask = erp_mask;
+	aentry->ht_key.erp_id = mlxsw_sp_acl_erp_mask_erp_id(erp_mask);
+	memcpy(aentry->ht_key.enc_key, aentry->full_enc_key,
+	       sizeof(aentry->ht_key.enc_key));
+
+	/* Compute all needed delta information and clear the delta bits
+	 * from the encrypted key.
+	 */
+	delta = mlxsw_sp_acl_erp_delta(aentry->erp_mask);
+	aentry->delta_info.start = mlxsw_sp_acl_erp_delta_start(delta);
+	aentry->delta_info.mask = mlxsw_sp_acl_erp_delta_mask(delta);
+	aentry->delta_info.value =
+		mlxsw_sp_acl_erp_delta_value(delta, aentry->full_enc_key);
+	mlxsw_sp_acl_erp_delta_clear(delta, aentry->ht_key.enc_key);
 
 	/* We can't insert identical rules into the A-TCAM, so fail and
 	 * let the rule spill into C-TCAM
@@ -472,7 +491,7 @@ __mlxsw_sp_acl_atcam_entry_add(struct mlxsw_sp *mlxsw_sp,
 	rhashtable_remove_fast(&aregion->entries_ht, &aentry->ht_node,
 			       mlxsw_sp_acl_atcam_entries_ht_params);
 err_rhashtable_insert:
-	mlxsw_sp_acl_erp_put(aregion, erp);
+	mlxsw_sp_acl_erp_mask_put(aregion, erp_mask);
 	return err;
 }
 
@@ -484,7 +503,7 @@ __mlxsw_sp_acl_atcam_entry_del(struct mlxsw_sp *mlxsw_sp,
 	mlxsw_sp_acl_atcam_region_entry_remove(mlxsw_sp, aregion, aentry);
 	rhashtable_remove_fast(&aregion->entries_ht, &aentry->ht_node,
 			       mlxsw_sp_acl_atcam_entries_ht_params);
-	mlxsw_sp_acl_erp_put(aregion, aentry->erp);
+	mlxsw_sp_acl_erp_mask_put(aregion, aentry->erp_mask);
 }
 
 int mlxsw_sp_acl_atcam_entry_add(struct mlxsw_sp *mlxsw_sp,

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

@@ -46,7 +46,6 @@ mlxsw_sp_acl_ctcam_region_entry_insert(struct mlxsw_sp *mlxsw_sp,
 	struct mlxsw_sp_acl_tcam_region *region = cregion->region;
 	struct mlxsw_afk *afk = mlxsw_sp_acl_afk(mlxsw_sp->acl);
 	char ptce2_pl[MLXSW_REG_PTCE2_LEN];
-	unsigned int blocks_count;
 	char *act_set;
 	u32 priority;
 	char *mask;
@@ -63,9 +62,7 @@ mlxsw_sp_acl_ctcam_region_entry_insert(struct mlxsw_sp *mlxsw_sp,
 			     centry->parman_item.index, priority);
 	key = mlxsw_reg_ptce2_flex_key_blocks_data(ptce2_pl);
 	mask = mlxsw_reg_ptce2_mask_data(ptce2_pl);
-	blocks_count = mlxsw_afk_key_info_blocks_count_get(region->key_info);
-	mlxsw_afk_encode(afk, region->key_info, &rulei->values, key, mask, 0,
-			 blocks_count - 1);
+	mlxsw_afk_encode(afk, region->key_info, &rulei->values, key, mask);
 
 	err = cregion->ops->entry_insert(cregion, centry, mask);
 	if (err)

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

@@ -7,7 +7,7 @@
 #include <linux/gfp.h>
 #include <linux/kernel.h>
 #include <linux/list.h>
-#include <linux/rhashtable.h>
+#include <linux/objagg.h>
 #include <linux/rtnetlink.h>
 #include <linux/slab.h>
 
@@ -29,6 +29,8 @@ struct mlxsw_sp_acl_erp_core {
 
 struct mlxsw_sp_acl_erp_key {
 	char mask[MLXSW_REG_PTCEX_FLEX_KEY_BLOCKS_LEN];
+#define __MASK_LEN 0x38
+#define __MASK_IDX(i) (__MASK_LEN - (i) - 1)
 	bool ctcam;
 };
 
@@ -36,10 +38,8 @@ struct mlxsw_sp_acl_erp {
 	struct mlxsw_sp_acl_erp_key key;
 	u8 id;
 	u8 index;
-	refcount_t refcnt;
 	DECLARE_BITMAP(mask_bitmap, MLXSW_SP_ACL_TCAM_MASK_LEN);
 	struct list_head list;
-	struct rhash_head ht_node;
 	struct mlxsw_sp_acl_erp_table *erp_table;
 };
 
@@ -53,7 +53,6 @@ struct mlxsw_sp_acl_erp_table {
 	DECLARE_BITMAP(erp_id_bitmap, MLXSW_SP_ACL_ERP_MAX_PER_REGION);
 	DECLARE_BITMAP(erp_index_bitmap, MLXSW_SP_ACL_ERP_MAX_PER_REGION);
 	struct list_head atcam_erps_list;
-	struct rhashtable erp_ht;
 	struct mlxsw_sp_acl_erp_core *erp_core;
 	struct mlxsw_sp_acl_atcam_region *aregion;
 	const struct mlxsw_sp_acl_erp_table_ops *ops;
@@ -61,12 +60,8 @@ struct mlxsw_sp_acl_erp_table {
 	unsigned int num_atcam_erps;
 	unsigned int num_max_atcam_erps;
 	unsigned int num_ctcam_erps;
-};
-
-static const struct rhashtable_params mlxsw_sp_acl_erp_ht_params = {
-	.key_len = sizeof(struct mlxsw_sp_acl_erp_key),
-	.key_offset = offsetof(struct mlxsw_sp_acl_erp, key),
-	.head_offset = offsetof(struct mlxsw_sp_acl_erp, ht_node),
+	unsigned int num_deltas;
+	struct objagg *objagg;
 };
 
 struct mlxsw_sp_acl_erp_table_ops {
@@ -119,16 +114,6 @@ static const struct mlxsw_sp_acl_erp_table_ops erp_no_mask_ops = {
 	.erp_destroy = mlxsw_sp_acl_erp_no_mask_destroy,
 };
 
-bool mlxsw_sp_acl_erp_is_ctcam_erp(const struct mlxsw_sp_acl_erp *erp)
-{
-	return erp->key.ctcam;
-}
-
-u8 mlxsw_sp_acl_erp_id(const struct mlxsw_sp_acl_erp *erp)
-{
-	return erp->id;
-}
-
 static unsigned int
 mlxsw_sp_acl_erp_table_entry_size(const struct mlxsw_sp_acl_erp_table *erp_table)
 {
@@ -194,12 +179,15 @@ mlxsw_sp_acl_erp_master_mask_update(struct mlxsw_sp_acl_erp_table *erp_table)
 
 static int
 mlxsw_sp_acl_erp_master_mask_set(struct mlxsw_sp_acl_erp_table *erp_table,
-				 const struct mlxsw_sp_acl_erp *erp)
+				 struct mlxsw_sp_acl_erp_key *key)
 {
+	DECLARE_BITMAP(mask_bitmap, MLXSW_SP_ACL_TCAM_MASK_LEN);
 	unsigned long bit;
 	int err;
 
-	for_each_set_bit(bit, erp->mask_bitmap, MLXSW_SP_ACL_TCAM_MASK_LEN)
+	bitmap_from_arr32(mask_bitmap, (u32 *) key->mask,
+			  MLXSW_SP_ACL_TCAM_MASK_LEN);
+	for_each_set_bit(bit, mask_bitmap, MLXSW_SP_ACL_TCAM_MASK_LEN)
 		mlxsw_sp_acl_erp_master_mask_bit_set(bit,
 						     &erp_table->master_mask);
 
@@ -210,7 +198,7 @@ mlxsw_sp_acl_erp_master_mask_set(struct mlxsw_sp_acl_erp_table *erp_table,
 	return 0;
 
 err_master_mask_update:
-	for_each_set_bit(bit, erp->mask_bitmap, MLXSW_SP_ACL_TCAM_MASK_LEN)
+	for_each_set_bit(bit, mask_bitmap, MLXSW_SP_ACL_TCAM_MASK_LEN)
 		mlxsw_sp_acl_erp_master_mask_bit_clear(bit,
 						       &erp_table->master_mask);
 	return err;
@@ -218,12 +206,15 @@ mlxsw_sp_acl_erp_master_mask_set(struct mlxsw_sp_acl_erp_table *erp_table,
 
 static int
 mlxsw_sp_acl_erp_master_mask_clear(struct mlxsw_sp_acl_erp_table *erp_table,
-				   const struct mlxsw_sp_acl_erp *erp)
+				   struct mlxsw_sp_acl_erp_key *key)
 {
+	DECLARE_BITMAP(mask_bitmap, MLXSW_SP_ACL_TCAM_MASK_LEN);
 	unsigned long bit;
 	int err;
 
-	for_each_set_bit(bit, erp->mask_bitmap, MLXSW_SP_ACL_TCAM_MASK_LEN)
+	bitmap_from_arr32(mask_bitmap, (u32 *) key->mask,
+			  MLXSW_SP_ACL_TCAM_MASK_LEN);
+	for_each_set_bit(bit, mask_bitmap, MLXSW_SP_ACL_TCAM_MASK_LEN)
 		mlxsw_sp_acl_erp_master_mask_bit_clear(bit,
 						       &erp_table->master_mask);
 
@@ -234,7 +225,7 @@ mlxsw_sp_acl_erp_master_mask_clear(struct mlxsw_sp_acl_erp_table *erp_table,
 	return 0;
 
 err_master_mask_update:
-	for_each_set_bit(bit, erp->mask_bitmap, MLXSW_SP_ACL_TCAM_MASK_LEN)
+	for_each_set_bit(bit, mask_bitmap, MLXSW_SP_ACL_TCAM_MASK_LEN)
 		mlxsw_sp_acl_erp_master_mask_bit_set(bit,
 						     &erp_table->master_mask);
 	return err;
@@ -256,26 +247,16 @@ mlxsw_sp_acl_erp_generic_create(struct mlxsw_sp_acl_erp_table *erp_table,
 		goto err_erp_id_get;
 
 	memcpy(&erp->key, key, sizeof(*key));
-	bitmap_from_arr32(erp->mask_bitmap, (u32 *) key->mask,
-			  MLXSW_SP_ACL_TCAM_MASK_LEN);
 	list_add(&erp->list, &erp_table->atcam_erps_list);
-	refcount_set(&erp->refcnt, 1);
 	erp_table->num_atcam_erps++;
 	erp->erp_table = erp_table;
 
-	err = mlxsw_sp_acl_erp_master_mask_set(erp_table, erp);
+	err = mlxsw_sp_acl_erp_master_mask_set(erp_table, &erp->key);
 	if (err)
 		goto err_master_mask_set;
 
-	err = rhashtable_insert_fast(&erp_table->erp_ht, &erp->ht_node,
-				     mlxsw_sp_acl_erp_ht_params);
-	if (err)
-		goto err_rhashtable_insert;
-
 	return erp;
 
-err_rhashtable_insert:
-	mlxsw_sp_acl_erp_master_mask_clear(erp_table, erp);
 err_master_mask_set:
 	erp_table->num_atcam_erps--;
 	list_del(&erp->list);
@@ -290,9 +271,7 @@ mlxsw_sp_acl_erp_generic_destroy(struct mlxsw_sp_acl_erp *erp)
 {
 	struct mlxsw_sp_acl_erp_table *erp_table = erp->erp_table;
 
-	rhashtable_remove_fast(&erp_table->erp_ht, &erp->ht_node,
-			       mlxsw_sp_acl_erp_ht_params);
-	mlxsw_sp_acl_erp_master_mask_clear(erp_table, erp);
+	mlxsw_sp_acl_erp_master_mask_clear(erp_table, &erp->key);
 	erp_table->num_atcam_erps--;
 	list_del(&erp->list);
 	mlxsw_sp_acl_erp_id_put(erp_table, erp->id);
@@ -647,9 +626,56 @@ mlxsw_sp_acl_erp_region_ctcam_disable(struct mlxsw_sp_acl_erp_table *erp_table)
 	mlxsw_sp_acl_erp_table_enable(erp_table, false);
 }
 
+static int
+__mlxsw_sp_acl_erp_table_other_inc(struct mlxsw_sp_acl_erp_table *erp_table,
+				   unsigned int *inc_num)
+{
+	int err;
+
+	/* If there are C-TCAM eRP or deltas in use we need to transition
+	 * the region to use eRP table, if it is not already done
+	 */
+	if (erp_table->ops != &erp_two_masks_ops &&
+	    erp_table->ops != &erp_multiple_masks_ops) {
+		err = mlxsw_sp_acl_erp_region_table_trans(erp_table);
+		if (err)
+			return err;
+	}
+
+	/* When C-TCAM or deltas are used, the eRP table must be used */
+	if (erp_table->ops != &erp_multiple_masks_ops)
+		erp_table->ops = &erp_multiple_masks_ops;
+
+	(*inc_num)++;
+
+	return 0;
+}
+
+static int mlxsw_sp_acl_erp_ctcam_inc(struct mlxsw_sp_acl_erp_table *erp_table)
+{
+	return __mlxsw_sp_acl_erp_table_other_inc(erp_table,
+						  &erp_table->num_ctcam_erps);
+}
+
+static int mlxsw_sp_acl_erp_delta_inc(struct mlxsw_sp_acl_erp_table *erp_table)
+{
+	return __mlxsw_sp_acl_erp_table_other_inc(erp_table,
+						  &erp_table->num_deltas);
+}
+
 static void
-mlxsw_sp_acl_erp_ctcam_table_ops_set(struct mlxsw_sp_acl_erp_table *erp_table)
+__mlxsw_sp_acl_erp_table_other_dec(struct mlxsw_sp_acl_erp_table *erp_table,
+				   unsigned int *dec_num)
 {
+	(*dec_num)--;
+
+	/* If there are no C-TCAM eRP or deltas in use, the state we
+	 * transition to depends on the number of A-TCAM eRPs currently
+	 * in use.
+	 */
+	if (erp_table->num_ctcam_erps > 0 || erp_table->num_deltas > 0)
+		return;
+
 	switch (erp_table->num_atcam_erps) {
 	case 2:
 		/* Keep using the eRP table, but correctly set the
@@ -683,9 +709,21 @@ mlxsw_sp_acl_erp_ctcam_table_ops_set(struct mlxsw_sp_acl_erp_table *erp_table)
 	}
 }
 
+static void mlxsw_sp_acl_erp_ctcam_dec(struct mlxsw_sp_acl_erp_table *erp_table)
+{
+	__mlxsw_sp_acl_erp_table_other_dec(erp_table,
+					   &erp_table->num_ctcam_erps);
+}
+
+static void mlxsw_sp_acl_erp_delta_dec(struct mlxsw_sp_acl_erp_table *erp_table)
+{
+	__mlxsw_sp_acl_erp_table_other_dec(erp_table,
+					   &erp_table->num_deltas);
+}
+
 static struct mlxsw_sp_acl_erp *
-__mlxsw_sp_acl_erp_ctcam_mask_create(struct mlxsw_sp_acl_erp_table *erp_table,
-				     struct mlxsw_sp_acl_erp_key *key)
+mlxsw_sp_acl_erp_ctcam_mask_create(struct mlxsw_sp_acl_erp_table *erp_table,
+				   struct mlxsw_sp_acl_erp_key *key)
 {
 	struct mlxsw_sp_acl_erp *erp;
 	int err;
@@ -697,89 +735,41 @@ __mlxsw_sp_acl_erp_ctcam_mask_create(struct mlxsw_sp_acl_erp_table *erp_table,
 	memcpy(&erp->key, key, sizeof(*key));
 	bitmap_from_arr32(erp->mask_bitmap, (u32 *) key->mask,
 			  MLXSW_SP_ACL_TCAM_MASK_LEN);
-	refcount_set(&erp->refcnt, 1);
-	erp_table->num_ctcam_erps++;
-	erp->erp_table = erp_table;
 
-	err = mlxsw_sp_acl_erp_master_mask_set(erp_table, erp);
+	err = mlxsw_sp_acl_erp_ctcam_inc(erp_table);
 	if (err)
-		goto err_master_mask_set;
+		goto err_erp_ctcam_inc;
 
-	err = rhashtable_insert_fast(&erp_table->erp_ht, &erp->ht_node,
-				     mlxsw_sp_acl_erp_ht_params);
+	erp->erp_table = erp_table;
+
+	err = mlxsw_sp_acl_erp_master_mask_set(erp_table, &erp->key);
 	if (err)
-		goto err_rhashtable_insert;
+		goto err_master_mask_set;
 
 	err = mlxsw_sp_acl_erp_region_ctcam_enable(erp_table);
 	if (err)
 		goto err_erp_region_ctcam_enable;
 
-	/* When C-TCAM is used, the eRP table must be used */
-	erp_table->ops = &erp_multiple_masks_ops;
-
 	return erp;
 
 err_erp_region_ctcam_enable:
-	rhashtable_remove_fast(&erp_table->erp_ht, &erp->ht_node,
-			       mlxsw_sp_acl_erp_ht_params);
-err_rhashtable_insert:
-	mlxsw_sp_acl_erp_master_mask_clear(erp_table, erp);
+	mlxsw_sp_acl_erp_master_mask_clear(erp_table, &erp->key);
 err_master_mask_set:
-	erp_table->num_ctcam_erps--;
+	mlxsw_sp_acl_erp_ctcam_dec(erp_table);
+err_erp_ctcam_inc:
 	kfree(erp);
 	return ERR_PTR(err);
 }
 
-static struct mlxsw_sp_acl_erp *
-mlxsw_sp_acl_erp_ctcam_mask_create(struct mlxsw_sp_acl_erp_table *erp_table,
-				   struct mlxsw_sp_acl_erp_key *key)
-{
-	struct mlxsw_sp_acl_erp *erp;
-	int err;
-
-	/* There is a special situation where we need to spill rules
-	 * into the C-TCAM, yet the region is still using a master
-	 * mask and thus not performing a lookup in the C-TCAM. This
-	 * can happen when two rules that only differ in priority - and
-	 * thus sharing the same key - are programmed. In this case
-	 * we transition the region to use an eRP table
-	 */
-	err = mlxsw_sp_acl_erp_region_table_trans(erp_table);
-	if (err)
-		return ERR_PTR(err);
-
-	erp = __mlxsw_sp_acl_erp_ctcam_mask_create(erp_table, key);
-	if (IS_ERR(erp)) {
-		err = PTR_ERR(erp);
-		goto err_erp_create;
-	}
-
-	return erp;
-
-err_erp_create:
-	mlxsw_sp_acl_erp_region_master_mask_trans(erp_table);
-	return ERR_PTR(err);
-}
-
 static void
 mlxsw_sp_acl_erp_ctcam_mask_destroy(struct mlxsw_sp_acl_erp *erp)
 {
 	struct mlxsw_sp_acl_erp_table *erp_table = erp->erp_table;
 
 	mlxsw_sp_acl_erp_region_ctcam_disable(erp_table);
-	rhashtable_remove_fast(&erp_table->erp_ht, &erp->ht_node,
-			       mlxsw_sp_acl_erp_ht_params);
-	mlxsw_sp_acl_erp_master_mask_clear(erp_table, erp);
-	erp_table->num_ctcam_erps--;
+	mlxsw_sp_acl_erp_master_mask_clear(erp_table, &erp->key);
+	mlxsw_sp_acl_erp_ctcam_dec(erp_table);
 	kfree(erp);
-
-	/* Once the last C-TCAM eRP was destroyed, the state we
-	 * transition to depends on the number of A-TCAM eRPs currently
-	 * in use
-	 */
-	if (erp_table->num_ctcam_erps > 0)
-		return;
-	mlxsw_sp_acl_erp_ctcam_table_ops_set(erp_table);
 }
 
 static struct mlxsw_sp_acl_erp *
@@ -790,7 +780,7 @@ mlxsw_sp_acl_erp_mask_create(struct mlxsw_sp_acl_erp_table *erp_table,
 	int err;
 
 	if (key->ctcam)
-		return __mlxsw_sp_acl_erp_ctcam_mask_create(erp_table, key);
+		return mlxsw_sp_acl_erp_ctcam_mask_create(erp_table, key);
 
 	/* Expand the eRP table for the new eRP, if needed */
 	err = mlxsw_sp_acl_erp_table_expand(erp_table);
@@ -838,7 +828,8 @@ mlxsw_sp_acl_erp_mask_destroy(struct mlxsw_sp_acl_erp_table *erp_table,
 	mlxsw_sp_acl_erp_index_put(erp_table, erp->index);
 	mlxsw_sp_acl_erp_generic_destroy(erp);
 
-	if (erp_table->num_atcam_erps == 2 && erp_table->num_ctcam_erps == 0)
+	if (erp_table->num_atcam_erps == 2 && erp_table->num_ctcam_erps == 0 &&
+	    erp_table->num_deltas == 0)
 		erp_table->ops = &erp_two_masks_ops;
 }
 
@@ -940,13 +931,12 @@ mlxsw_sp_acl_erp_no_mask_destroy(struct mlxsw_sp_acl_erp_table *erp_table,
 	WARN_ON(1);
 }
 
-struct mlxsw_sp_acl_erp *
-mlxsw_sp_acl_erp_get(struct mlxsw_sp_acl_atcam_region *aregion,
-		     const char *mask, bool ctcam)
+struct mlxsw_sp_acl_erp_mask *
+mlxsw_sp_acl_erp_mask_get(struct mlxsw_sp_acl_atcam_region *aregion,
+			  const char *mask, bool ctcam)
 {
-	struct mlxsw_sp_acl_erp_table *erp_table = aregion->erp_table;
 	struct mlxsw_sp_acl_erp_key key;
-	struct mlxsw_sp_acl_erp *erp;
+	struct objagg_obj *objagg_obj;
 
 	/* eRPs are allocated from a shared resource, but currently all
 	 * allocations are done under RTNL.
@@ -955,29 +945,238 @@ mlxsw_sp_acl_erp_get(struct mlxsw_sp_acl_atcam_region *aregion,
 
 	memcpy(key.mask, mask, MLXSW_REG_PTCEX_FLEX_KEY_BLOCKS_LEN);
 	key.ctcam = ctcam;
-	erp = rhashtable_lookup_fast(&erp_table->erp_ht, &key,
-				     mlxsw_sp_acl_erp_ht_params);
-	if (erp) {
-		refcount_inc(&erp->refcnt);
-		return erp;
+	objagg_obj = objagg_obj_get(aregion->erp_table->objagg, &key);
+	if (IS_ERR(objagg_obj))
+		return ERR_CAST(objagg_obj);
+	return (struct mlxsw_sp_acl_erp_mask *) objagg_obj;
+}
+
+void mlxsw_sp_acl_erp_mask_put(struct mlxsw_sp_acl_atcam_region *aregion,
+			       struct mlxsw_sp_acl_erp_mask *erp_mask)
+{
+	struct objagg_obj *objagg_obj = (struct objagg_obj *) erp_mask;
+
+	ASSERT_RTNL();
+	objagg_obj_put(aregion->erp_table->objagg, objagg_obj);
+}
+
+bool
+mlxsw_sp_acl_erp_mask_is_ctcam(const struct mlxsw_sp_acl_erp_mask *erp_mask)
+{
+	struct objagg_obj *objagg_obj = (struct objagg_obj *) erp_mask;
+	const struct mlxsw_sp_acl_erp_key *key = objagg_obj_raw(objagg_obj);
+
+	return key->ctcam;
+}
+
+u8 mlxsw_sp_acl_erp_mask_erp_id(const struct mlxsw_sp_acl_erp_mask *erp_mask)
+{
+	struct objagg_obj *objagg_obj = (struct objagg_obj *) erp_mask;
+	const struct mlxsw_sp_acl_erp *erp = objagg_obj_root_priv(objagg_obj);
+
+	return erp->id;
+}
+
+struct mlxsw_sp_acl_erp_delta {
+	struct mlxsw_sp_acl_erp_key key;
+	u16 start;
+	u8 mask;
+};
+
+u16 mlxsw_sp_acl_erp_delta_start(const struct mlxsw_sp_acl_erp_delta *delta)
+{
+	return delta->start;
+}
+
+u8 mlxsw_sp_acl_erp_delta_mask(const struct mlxsw_sp_acl_erp_delta *delta)
+{
+	return delta->mask;
+}
+
+u8 mlxsw_sp_acl_erp_delta_value(const struct mlxsw_sp_acl_erp_delta *delta,
+				const char *enc_key)
+{
+	u16 start = delta->start;
+	u8 mask = delta->mask;
+	u16 tmp;
+
+	if (!mask)
+		return 0;
+
+	tmp = (unsigned char) enc_key[__MASK_IDX(start / 8)];
+	if (start / 8 + 1 < __MASK_LEN)
+		tmp |= (unsigned char) enc_key[__MASK_IDX(start / 8 + 1)] << 8;
+	tmp >>= start % 8;
+	tmp &= mask;
+	return tmp;
+}
+
+void mlxsw_sp_acl_erp_delta_clear(const struct mlxsw_sp_acl_erp_delta *delta,
+				  const char *enc_key)
+{
+	u16 start = delta->start;
+	u8 mask = delta->mask;
+	unsigned char *byte;
+	u16 tmp;
+
+	tmp = mask;
+	tmp <<= start % 8;
+	tmp = ~tmp;
+
+	byte = (unsigned char *) &enc_key[__MASK_IDX(start / 8)];
+	*byte &= tmp & 0xff;
+	if (start / 8 + 1 < __MASK_LEN) {
+		byte = (unsigned char *) &enc_key[__MASK_IDX(start / 8 + 1)];
+		*byte &= (tmp >> 8) & 0xff;
 	}
+}
+
+static const struct mlxsw_sp_acl_erp_delta
+mlxsw_sp_acl_erp_delta_default = {};
 
-	return erp_table->ops->erp_create(erp_table, &key);
+const struct mlxsw_sp_acl_erp_delta *
+mlxsw_sp_acl_erp_delta(const struct mlxsw_sp_acl_erp_mask *erp_mask)
+{
+	struct objagg_obj *objagg_obj = (struct objagg_obj *) erp_mask;
+	const struct mlxsw_sp_acl_erp_delta *delta;
+
+	delta = objagg_obj_delta_priv(objagg_obj);
+	if (!delta)
+		delta = &mlxsw_sp_acl_erp_delta_default;
+	return delta;
 }
 
-void mlxsw_sp_acl_erp_put(struct mlxsw_sp_acl_atcam_region *aregion,
-			  struct mlxsw_sp_acl_erp *erp)
+static int
+mlxsw_sp_acl_erp_delta_fill(const struct mlxsw_sp_acl_erp_key *parent_key,
+			    const struct mlxsw_sp_acl_erp_key *key,
+			    u16 *delta_start, u8 *delta_mask)
 {
+	int offset = 0;
+	int si = -1;
+	u16 pmask;
+	u16 mask;
+	int i;
+
+	/* The difference between 2 masks can be up to 8 consecutive bits. */
+	for (i = 0; i < __MASK_LEN; i++) {
+		if (parent_key->mask[__MASK_IDX(i)] == key->mask[__MASK_IDX(i)])
+			continue;
+		if (si == -1)
+			si = i;
+		else if (si != i - 1)
+			return -EINVAL;
+	}
+	if (si == -1) {
+		/* The masks are the same, this cannot happen.
+		 * That means the caller is broken.
+		 */
+		WARN_ON(1);
+		*delta_start = 0;
+		*delta_mask = 0;
+		return 0;
+	}
+	pmask = (unsigned char) parent_key->mask[__MASK_IDX(si)];
+	mask = (unsigned char) key->mask[__MASK_IDX(si)];
+	if (si + 1 < __MASK_LEN) {
+		pmask |= (unsigned char) parent_key->mask[__MASK_IDX(si + 1)] << 8;
+		mask |= (unsigned char) key->mask[__MASK_IDX(si + 1)] << 8;
+	}
+
+	if ((pmask ^ mask) & pmask)
+		return -EINVAL;
+	mask &= ~pmask;
+	while (!(mask & (1 << offset)))
+		offset++;
+	while (!(mask & 1))
+		mask >>= 1;
+	if (mask & 0xff00)
+		return -EINVAL;
+
+	*delta_start = si * 8 + offset;
+	*delta_mask = mask;
+
+	return 0;
+}
+
+static void *mlxsw_sp_acl_erp_delta_create(void *priv, void *parent_obj,
+					   void *obj)
+{
+	struct mlxsw_sp_acl_erp_key *parent_key = parent_obj;
+	struct mlxsw_sp_acl_atcam_region *aregion = priv;
 	struct mlxsw_sp_acl_erp_table *erp_table = aregion->erp_table;
+	struct mlxsw_sp_acl_erp_key *key = obj;
+	struct mlxsw_sp_acl_erp_delta *delta;
+	u16 delta_start;
+	u8 delta_mask;
+	int err;
 
-	ASSERT_RTNL();
+	if (parent_key->ctcam || key->ctcam)
+		return ERR_PTR(-EINVAL);
+	err = mlxsw_sp_acl_erp_delta_fill(parent_key, key,
+					  &delta_start, &delta_mask);
+	if (err)
+		return ERR_PTR(-EINVAL);
 
-	if (!refcount_dec_and_test(&erp->refcnt))
-		return;
+	delta = kzalloc(sizeof(*delta), GFP_KERNEL);
+	if (!delta)
+		return ERR_PTR(-ENOMEM);
+	delta->start = delta_start;
+	delta->mask = delta_mask;
+
+	err = mlxsw_sp_acl_erp_delta_inc(erp_table);
+	if (err)
+		goto err_erp_delta_inc;
+
+	memcpy(&delta->key, key, sizeof(*key));
+	err = mlxsw_sp_acl_erp_master_mask_set(erp_table, &delta->key);
+	if (err)
+		goto err_master_mask_set;
+
+	return delta;
 
-	erp_table->ops->erp_destroy(erp_table, erp);
+err_master_mask_set:
+	mlxsw_sp_acl_erp_delta_dec(erp_table);
+err_erp_delta_inc:
+	kfree(delta);
+	return ERR_PTR(err);
 }
 
+static void mlxsw_sp_acl_erp_delta_destroy(void *priv, void *delta_priv)
+{
+	struct mlxsw_sp_acl_erp_delta *delta = delta_priv;
+	struct mlxsw_sp_acl_atcam_region *aregion = priv;
+	struct mlxsw_sp_acl_erp_table *erp_table = aregion->erp_table;
+
+	mlxsw_sp_acl_erp_master_mask_clear(erp_table, &delta->key);
+	mlxsw_sp_acl_erp_delta_dec(erp_table);
+	kfree(delta);
+}
+
+static void *mlxsw_sp_acl_erp_root_create(void *priv, void *obj)
+{
+	struct mlxsw_sp_acl_atcam_region *aregion = priv;
+	struct mlxsw_sp_acl_erp_table *erp_table = aregion->erp_table;
+	struct mlxsw_sp_acl_erp_key *key = obj;
+
+	return erp_table->ops->erp_create(erp_table, key);
+}
+
+static void mlxsw_sp_acl_erp_root_destroy(void *priv, void *root_priv)
+{
+	struct mlxsw_sp_acl_atcam_region *aregion = priv;
+	struct mlxsw_sp_acl_erp_table *erp_table = aregion->erp_table;
+
+	erp_table->ops->erp_destroy(erp_table, root_priv);
+}
+
+static const struct objagg_ops mlxsw_sp_acl_erp_objagg_ops = {
+	.obj_size = sizeof(struct mlxsw_sp_acl_erp_key),
+	.delta_create = mlxsw_sp_acl_erp_delta_create,
+	.delta_destroy = mlxsw_sp_acl_erp_delta_destroy,
+	.root_create = mlxsw_sp_acl_erp_root_create,
+	.root_destroy = mlxsw_sp_acl_erp_root_destroy,
+};
+
 static struct mlxsw_sp_acl_erp_table *
 mlxsw_sp_acl_erp_table_create(struct mlxsw_sp_acl_atcam_region *aregion)
 {
@@ -988,9 +1187,12 @@ mlxsw_sp_acl_erp_table_create(struct mlxsw_sp_acl_atcam_region *aregion)
 	if (!erp_table)
 		return ERR_PTR(-ENOMEM);
 
-	err = rhashtable_init(&erp_table->erp_ht, &mlxsw_sp_acl_erp_ht_params);
-	if (err)
-		goto err_rhashtable_init;
+	erp_table->objagg = objagg_create(&mlxsw_sp_acl_erp_objagg_ops,
+					  aregion);
+	if (IS_ERR(erp_table->objagg)) {
+		err = PTR_ERR(erp_table->objagg);
+		goto err_objagg_create;
+	}
 
 	erp_table->erp_core = aregion->atcam->erp_core;
 	erp_table->ops = &erp_no_mask_ops;
@@ -999,7 +1201,7 @@ mlxsw_sp_acl_erp_table_create(struct mlxsw_sp_acl_atcam_region *aregion)
 
 	return erp_table;
 
-err_rhashtable_init:
+err_objagg_create:
 	kfree(erp_table);
 	return ERR_PTR(err);
 }
@@ -1008,7 +1210,7 @@ static void
 mlxsw_sp_acl_erp_table_destroy(struct mlxsw_sp_acl_erp_table *erp_table)
 {
 	WARN_ON(!list_empty(&erp_table->atcam_erps_list));
-	rhashtable_destroy(&erp_table->erp_ht);
+	objagg_destroy(erp_table->objagg);
 	kfree(erp_table);
 }
 

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

@@ -98,8 +98,8 @@ static const struct mlxsw_afk_block mlxsw_sp1_afk_blocks[] = {
 
 #define MLXSW_SP1_AFK_KEY_BLOCK_SIZE 16
 
-static void mlxsw_sp1_afk_encode_block(char *block, int block_index,
-				       char *output)
+static void mlxsw_sp1_afk_encode_block(char *output, int block_index,
+				       char *block)
 {
 	unsigned int offset = block_index * MLXSW_SP1_AFK_KEY_BLOCK_SIZE;
 	char *output_indexed = output + offset;
@@ -107,10 +107,19 @@ static void mlxsw_sp1_afk_encode_block(char *block, int block_index,
 	memcpy(output_indexed, block, MLXSW_SP1_AFK_KEY_BLOCK_SIZE);
 }
 
+static void mlxsw_sp1_afk_clear_block(char *output, int block_index)
+{
+	unsigned int offset = block_index * MLXSW_SP1_AFK_KEY_BLOCK_SIZE;
+	char *output_indexed = output + offset;
+
+	memset(output_indexed, 0, MLXSW_SP1_AFK_KEY_BLOCK_SIZE);
+}
+
 const struct mlxsw_afk_ops mlxsw_sp1_afk_ops = {
 	.blocks		= mlxsw_sp1_afk_blocks,
 	.blocks_count	= ARRAY_SIZE(mlxsw_sp1_afk_blocks),
 	.encode_block	= mlxsw_sp1_afk_encode_block,
+	.clear_block	= mlxsw_sp1_afk_clear_block,
 };
 
 static struct mlxsw_afk_element_inst mlxsw_sp_afk_element_info_mac_0[] = {
@@ -263,10 +272,9 @@ static const struct mlxsw_sp2_afk_block_layout mlxsw_sp2_afk_blocks_layout[] = {
 	MLXSW_SP2_AFK_BLOCK_LAYOUT(block11, 0x00, 12),
 };
 
-static void mlxsw_sp2_afk_encode_block(char *block, int block_index,
-				       char *output)
+static void __mlxsw_sp2_afk_block_value_set(char *output, int block_index,
+					    u64 block_value)
 {
-	u64 block_value = mlxsw_sp2_afk_block_value_get(block);
 	const struct mlxsw_sp2_afk_block_layout *block_layout;
 
 	if (WARN_ON(block_index < 0 ||
@@ -278,8 +286,22 @@ static void mlxsw_sp2_afk_encode_block(char *block, int block_index,
 			   &block_layout->item, 0, block_value);
 }
 
+static void mlxsw_sp2_afk_encode_block(char *output, int block_index,
+				       char *block)
+{
+	u64 block_value = mlxsw_sp2_afk_block_value_get(block);
+
+	__mlxsw_sp2_afk_block_value_set(output, block_index, block_value);
+}
+
+static void mlxsw_sp2_afk_clear_block(char *output, int block_index)
+{
+	__mlxsw_sp2_afk_block_value_set(output, block_index, 0);
+}
+
 const struct mlxsw_afk_ops mlxsw_sp2_afk_ops = {
 	.blocks		= mlxsw_sp2_afk_blocks,
 	.blocks_count	= ARRAY_SIZE(mlxsw_sp2_afk_blocks),
 	.encode_block	= mlxsw_sp2_afk_encode_block,
+	.clear_block	= mlxsw_sp2_afk_clear_block,
 };

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

@@ -154,7 +154,9 @@ struct mlxsw_sp_acl_atcam_region {
 };
 
 struct mlxsw_sp_acl_atcam_entry_ht_key {
-	char enc_key[MLXSW_REG_PTCEX_FLEX_KEY_BLOCKS_LEN]; /* Encoded key */
+	char enc_key[MLXSW_REG_PTCEX_FLEX_KEY_BLOCKS_LEN]; /* Encoded key,
+							    * minus delta bits.
+							    */
 	u8 erp_id;
 };
 
@@ -165,9 +167,15 @@ struct mlxsw_sp_acl_atcam_chunk {
 struct mlxsw_sp_acl_atcam_entry {
 	struct rhash_head ht_node;
 	struct mlxsw_sp_acl_atcam_entry_ht_key ht_key;
+	char full_enc_key[MLXSW_REG_PTCEX_FLEX_KEY_BLOCKS_LEN]; /* Encoded key */
+	struct {
+		u16 start;
+		u8 mask;
+		u8 value;
+	} delta_info;
 	struct mlxsw_sp_acl_ctcam_entry centry;
 	struct mlxsw_sp_acl_atcam_lkey_id *lkey_id;
-	struct mlxsw_sp_acl_erp *erp;
+	struct mlxsw_sp_acl_erp_mask *erp_mask;
 };
 
 static inline struct mlxsw_sp_acl_atcam_region *
@@ -209,15 +217,27 @@ int mlxsw_sp_acl_atcam_init(struct mlxsw_sp *mlxsw_sp,
 void mlxsw_sp_acl_atcam_fini(struct mlxsw_sp *mlxsw_sp,
 			     struct mlxsw_sp_acl_atcam *atcam);
 
-struct mlxsw_sp_acl_erp;
-
-bool mlxsw_sp_acl_erp_is_ctcam_erp(const struct mlxsw_sp_acl_erp *erp);
-u8 mlxsw_sp_acl_erp_id(const struct mlxsw_sp_acl_erp *erp);
-struct mlxsw_sp_acl_erp *
-mlxsw_sp_acl_erp_get(struct mlxsw_sp_acl_atcam_region *aregion,
-		     const char *mask, bool ctcam);
-void mlxsw_sp_acl_erp_put(struct mlxsw_sp_acl_atcam_region *aregion,
-			  struct mlxsw_sp_acl_erp *erp);
+struct mlxsw_sp_acl_erp_delta;
+
+u16 mlxsw_sp_acl_erp_delta_start(const struct mlxsw_sp_acl_erp_delta *delta);
+u8 mlxsw_sp_acl_erp_delta_mask(const struct mlxsw_sp_acl_erp_delta *delta);
+u8 mlxsw_sp_acl_erp_delta_value(const struct mlxsw_sp_acl_erp_delta *delta,
+				const char *enc_key);
+void mlxsw_sp_acl_erp_delta_clear(const struct mlxsw_sp_acl_erp_delta *delta,
+				  const char *enc_key);
+
+struct mlxsw_sp_acl_erp_mask;
+
+bool
+mlxsw_sp_acl_erp_mask_is_ctcam(const struct mlxsw_sp_acl_erp_mask *erp_mask);
+u8 mlxsw_sp_acl_erp_mask_erp_id(const struct mlxsw_sp_acl_erp_mask *erp_mask);
+const struct mlxsw_sp_acl_erp_delta *
+mlxsw_sp_acl_erp_delta(const struct mlxsw_sp_acl_erp_mask *erp_mask);
+struct mlxsw_sp_acl_erp_mask *
+mlxsw_sp_acl_erp_mask_get(struct mlxsw_sp_acl_atcam_region *aregion,
+			  const char *mask, bool ctcam);
+void mlxsw_sp_acl_erp_mask_put(struct mlxsw_sp_acl_atcam_region *aregion,
+			       struct mlxsw_sp_acl_erp_mask *erp_mask);
 int mlxsw_sp_acl_erp_region_init(struct mlxsw_sp_acl_atcam_region *aregion);
 void mlxsw_sp_acl_erp_region_fini(struct mlxsw_sp_acl_atcam_region *aregion);
 int mlxsw_sp_acl_erps_init(struct mlxsw_sp *mlxsw_sp,

include/linux/objagg.h

+/* SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0 */
+/* Copyright (c) 2018 Mellanox Technologies. All rights reserved */
+
+#ifndef _OBJAGG_H
+#define _OBJAGG_H
+
+struct objagg_ops {
+	size_t obj_size;
+	void * (*delta_create)(void *priv, void *parent_obj, void *obj);
+	void (*delta_destroy)(void *priv, void *delta_priv);
+	void * (*root_create)(void *priv, void *obj);
+	void (*root_destroy)(void *priv, void *root_priv);
+};
+
+struct objagg;
+struct objagg_obj;
+
+const void *objagg_obj_root_priv(const struct objagg_obj *objagg_obj);
+const void *objagg_obj_delta_priv(const struct objagg_obj *objagg_obj);
+const void *objagg_obj_raw(const struct objagg_obj *objagg_obj);
+
+struct objagg_obj *objagg_obj_get(struct objagg *objagg, void *obj);
+void objagg_obj_put(struct objagg *objagg, struct objagg_obj *objagg_obj);
+struct objagg *objagg_create(const struct objagg_ops *ops, void *priv);
+void objagg_destroy(struct objagg *objagg);
+
+struct objagg_obj_stats {
+	unsigned int user_count;
+	unsigned int delta_user_count; /* includes delta object users */
+};
+
+struct objagg_obj_stats_info {
+	struct objagg_obj_stats stats;
+	struct objagg_obj *objagg_obj; /* associated object */
+	bool is_root;
+};
+
+struct objagg_stats {
+	unsigned int stats_info_count;
+	struct objagg_obj_stats_info stats_info[];
+};
+
+const struct objagg_stats *objagg_stats_get(struct objagg *objagg);
+void objagg_stats_put(const struct objagg_stats *objagg_stats);
+
+#endif

include/trace/events/objagg.h

+/* SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0 */
+/* Copyright (c) 2018 Mellanox Technologies. All rights reserved */
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM objagg
+
+#if !defined(__TRACE_OBJAGG_H) || defined(TRACE_HEADER_MULTI_READ)
+#define __TRACE_OBJAGG_H
+
+#include <linux/tracepoint.h>
+
+struct objagg;
+struct objagg_obj;
+
+TRACE_EVENT(objagg_create,
+	TP_PROTO(const struct objagg *objagg),
+
+	TP_ARGS(objagg),
+
+	TP_STRUCT__entry(
+		__field(const void *, objagg)
+	),
+
+	TP_fast_assign(
+		__entry->objagg = objagg;
+	),
+
+	TP_printk("objagg %p", __entry->objagg)
+);
+
+TRACE_EVENT(objagg_destroy,
+	TP_PROTO(const struct objagg *objagg),
+
+	TP_ARGS(objagg),
+
+	TP_STRUCT__entry(
+		__field(const void *, objagg)
+	),
+
+	TP_fast_assign(
+		__entry->objagg = objagg;
+	),
+
+	TP_printk("objagg %p", __entry->objagg)
+);
+
+TRACE_EVENT(objagg_obj_create,
+	TP_PROTO(const struct objagg *objagg,
+		 const struct objagg_obj *obj),
+
+	TP_ARGS(objagg, obj),
+
+	TP_STRUCT__entry(
+		__field(const void *, objagg)
+		__field(const void *, obj)
+	),
+
+	TP_fast_assign(
+		__entry->objagg = objagg;
+		__entry->obj = obj;
+	),
+
+	TP_printk("objagg %p, obj %p", __entry->objagg, __entry->obj)
+);
+
+TRACE_EVENT(objagg_obj_destroy,
+	TP_PROTO(const struct objagg *objagg,
+		 const struct objagg_obj *obj),
+
+	TP_ARGS(objagg, obj),
+
+	TP_STRUCT__entry(
+		__field(const void *, objagg)
+		__field(const void *, obj)
+	),
+
+	TP_fast_assign(
+		__entry->objagg = objagg;
+		__entry->obj = obj;
+	),
+
+	TP_printk("objagg %p, obj %p", __entry->objagg, __entry->obj)
+);
+
+TRACE_EVENT(objagg_obj_get,
+	TP_PROTO(const struct objagg *objagg,
+		 const struct objagg_obj *obj,
+		 unsigned int refcount),
+
+	TP_ARGS(objagg, obj, refcount),
+
+	TP_STRUCT__entry(
+		__field(const void *, objagg)
+		__field(const void *, obj)
+		__field(unsigned int, refcount)
+	),
+
+	TP_fast_assign(
+		__entry->objagg = objagg;
+		__entry->obj = obj;
+		__entry->refcount = refcount;
+	),
+
+	TP_printk("objagg %p, obj %p, refcount %u",
+		  __entry->objagg, __entry->obj, __entry->refcount)
+);
+
+TRACE_EVENT(objagg_obj_put,
+	TP_PROTO(const struct objagg *objagg,
+		 const struct objagg_obj *obj,
+		 unsigned int refcount),
+
+	TP_ARGS(objagg, obj, refcount),
+
+	TP_STRUCT__entry(
+		__field(const void *, objagg)
+		__field(const void *, obj)
+		__field(unsigned int, refcount)
+	),
+
+	TP_fast_assign(
+		__entry->objagg = objagg;
+		__entry->obj = obj;
+		__entry->refcount = refcount;
+	),
+
+	TP_printk("objagg %p, obj %p, refcount %u",
+		  __entry->objagg, __entry->obj, __entry->refcount)
+);
+
+TRACE_EVENT(objagg_obj_parent_assign,
+	TP_PROTO(const struct objagg *objagg,
+		 const struct objagg_obj *obj,
+		 const struct objagg_obj *parent,
+		 unsigned int parent_refcount),
+
+	TP_ARGS(objagg, obj, parent, parent_refcount),
+
+	TP_STRUCT__entry(
+		__field(const void *, objagg)
+		__field(const void *, obj)
+		__field(const void *, parent)
+		__field(unsigned int, parent_refcount)
+	),
+
+	TP_fast_assign(
+		__entry->objagg = objagg;
+		__entry->obj = obj;
+		__entry->parent = parent;
+		__entry->parent_refcount = parent_refcount;
+	),
+
+	TP_printk("objagg %p, obj %p, parent %p, parent_refcount %u",
+		  __entry->objagg, __entry->obj,
+		  __entry->parent, __entry->parent_refcount)
+);
+
+TRACE_EVENT(objagg_obj_parent_unassign,
+	TP_PROTO(const struct objagg *objagg,
+		 const struct objagg_obj *obj,
+		 const struct objagg_obj *parent,
+		 unsigned int parent_refcount),
+
+	TP_ARGS(objagg, obj, parent, parent_refcount),
+
+	TP_STRUCT__entry(
+		__field(const void *, objagg)
+		__field(const void *, obj)
+		__field(const void *, parent)
+		__field(unsigned int, parent_refcount)
+	),
+
+	TP_fast_assign(
+		__entry->objagg = objagg;
+		__entry->obj = obj;
+		__entry->parent = parent;
+		__entry->parent_refcount = parent_refcount;
+	),
+
+	TP_printk("objagg %p, obj %p, parent %p, parent_refcount %u",
+		  __entry->objagg, __entry->obj,
+		  __entry->parent, __entry->parent_refcount)
+);
+
+TRACE_EVENT(objagg_obj_root_create,
+	TP_PROTO(const struct objagg *objagg,
+		 const struct objagg_obj *obj),
+
+	TP_ARGS(objagg, obj),
+
+	TP_STRUCT__entry(
+		__field(const void *, objagg)
+		__field(const void *, obj)
+	),
+
+	TP_fast_assign(
+		__entry->objagg = objagg;
+		__entry->obj = obj;
+	),
+
+	TP_printk("objagg %p, obj %p",
+		  __entry->objagg, __entry->obj)
+);
+
+TRACE_EVENT(objagg_obj_root_destroy,
+	TP_PROTO(const struct objagg *objagg,
+		 const struct objagg_obj *obj),
+
+	TP_ARGS(objagg, obj),
+
+	TP_STRUCT__entry(
+		__field(const void *, objagg)
+		__field(const void *, obj)
+	),
+
+	TP_fast_assign(
+		__entry->objagg = objagg;
+		__entry->obj = obj;
+	),
+
+	TP_printk("objagg %p, obj %p",
+		  __entry->objagg, __entry->obj)
+);
+
+#endif /* __TRACE_OBJAGG_H */
+
+/* This part must be outside protection */
+#include <trace/define_trace.h>

lib/Kconfig

@@ -624,3 +624,6 @@ config GENERIC_LIB_CMPDI2
 
 config GENERIC_LIB_UCMPDI2
 	bool
+
+config OBJAGG
+	tristate "objagg" if COMPILE_TEST

lib/Kconfig.debug

@@ -1976,6 +1976,16 @@ config TEST_MEMCAT_P
 
 	  If unsure, say N.
 
+config TEST_OBJAGG
+	tristate "Perform selftest on object aggreration manager"
+	default n
+	depends on OBJAGG
+	help
+	  Enable this option to test object aggregation manager on boot
+	  (or module load).
+
+	  If unsure, say N.
+
 endif # RUNTIME_TESTING_MENU
 
 config MEMTEST

lib/Makefile

@@ -75,6 +75,7 @@ obj-$(CONFIG_TEST_PARMAN) += test_parman.o
 obj-$(CONFIG_TEST_KMOD) += test_kmod.o
 obj-$(CONFIG_TEST_DEBUG_VIRTUAL) += test_debug_virtual.o
 obj-$(CONFIG_TEST_MEMCAT_P) += test_memcat_p.o
+obj-$(CONFIG_TEST_OBJAGG) += test_objagg.o
 
 ifeq ($(CONFIG_DEBUG_KOBJECT),y)
 CFLAGS_kobject.o += -DDEBUG
@@ -274,3 +275,4 @@ obj-$(CONFIG_GENERIC_LIB_LSHRDI3) += lshrdi3.o
 obj-$(CONFIG_GENERIC_LIB_MULDI3) += muldi3.o
 obj-$(CONFIG_GENERIC_LIB_CMPDI2) += cmpdi2.o
 obj-$(CONFIG_GENERIC_LIB_UCMPDI2) += ucmpdi2.o
+obj-$(CONFIG_OBJAGG) += objagg.o

lib/objagg.c

+// SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0
+/* Copyright (c) 2018 Mellanox Technologies. All rights reserved */
+
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/rhashtable.h>
+#include <linux/list.h>
+#include <linux/sort.h>
+#include <linux/objagg.h>
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/objagg.h>
+
+struct objagg {
+	const struct objagg_ops *ops;
+	void *priv;
+	struct rhashtable obj_ht;
+	struct rhashtable_params ht_params;
+	struct list_head obj_list;
+	unsigned int obj_count;
+};
+
+struct objagg_obj {
+	struct rhash_head ht_node; /* member of objagg->obj_ht */
+	struct list_head list; /* member of objagg->obj_list */
+	struct objagg_obj *parent; /* if the object is nested, this
+				    * holds pointer to parent, otherwise NULL
+				    */
+	union {
+		void *delta_priv; /* user delta private */
+		void *root_priv; /* user root private */
+	};
+	unsigned int refcount; /* counts number of users of this object
+				* including nested objects
+				*/
+	struct objagg_obj_stats stats;
+	unsigned long obj[0];
+};
+
+static unsigned int objagg_obj_ref_inc(struct objagg_obj *objagg_obj)
+{
+	return ++objagg_obj->refcount;
+}
+
+static unsigned int objagg_obj_ref_dec(struct objagg_obj *objagg_obj)
+{
+	return --objagg_obj->refcount;
+}
+
+static void objagg_obj_stats_inc(struct objagg_obj *objagg_obj)
+{
+	objagg_obj->stats.user_count++;
+	objagg_obj->stats.delta_user_count++;
+	if (objagg_obj->parent)
+		objagg_obj->parent->stats.delta_user_count++;
+}
+
+static void objagg_obj_stats_dec(struct objagg_obj *objagg_obj)
+{
+	objagg_obj->stats.user_count--;
+	objagg_obj->stats.delta_user_count--;
+	if (objagg_obj->parent)
+		objagg_obj->parent->stats.delta_user_count--;
+}
+
+static bool objagg_obj_is_root(const struct objagg_obj *objagg_obj)
+{
+	/* Nesting is not supported, so we can use ->parent
+	 * to figure out if the object is root.
+	 */
+	return !objagg_obj->parent;
+}
+
+/**
+ * objagg_obj_root_priv - obtains root private for an object
+ * @objagg_obj:	objagg object instance
+ *
+ * Note: all locking must be provided by the caller.
+ *
+ * Either the object is root itself when the private is returned
+ * directly, or the parent is root and its private is returned
+ * instead.
+ *
+ * Returns a user private root pointer.
+ */
+const void *objagg_obj_root_priv(const struct objagg_obj *objagg_obj)
+{
+	if (objagg_obj_is_root(objagg_obj))
+		return objagg_obj->root_priv;
+	WARN_ON(!objagg_obj_is_root(objagg_obj->parent));
+	return objagg_obj->parent->root_priv;
+}
+EXPORT_SYMBOL(objagg_obj_root_priv);
+
+/**
+ * objagg_obj_delta_priv - obtains delta private for an object
+ * @objagg_obj:	objagg object instance
+ *
+ * Note: all locking must be provided by the caller.
+ *
+ * Returns user private delta pointer or NULL in case the passed
+ * object is root.
+ */
+const void *objagg_obj_delta_priv(const struct objagg_obj *objagg_obj)
+{
+	if (objagg_obj_is_root(objagg_obj))
+		return NULL;
+	return objagg_obj->delta_priv;
+}
+EXPORT_SYMBOL(objagg_obj_delta_priv);
+
+/**
+ * objagg_obj_raw - obtains object user private pointer
+ * @objagg_obj:	objagg object instance
+ *
+ * Note: all locking must be provided by the caller.
+ *
+ * Returns user private pointer as was passed to objagg_obj_get() by "obj" arg.
+ */
+const void *objagg_obj_raw(const struct objagg_obj *objagg_obj)
+{
+	return objagg_obj->obj;
+}
+EXPORT_SYMBOL(objagg_obj_raw);
+
+static struct objagg_obj *objagg_obj_lookup(struct objagg *objagg, void *obj)
+{
+	return rhashtable_lookup_fast(&objagg->obj_ht, obj, objagg->ht_params);
+}
+
+static int objagg_obj_parent_assign(struct objagg *objagg,
+				    struct objagg_obj *objagg_obj,
+				    struct objagg_obj *parent)
+{
+	void *delta_priv;
+
+	delta_priv = objagg->ops->delta_create(objagg->priv, parent->obj,
+					       objagg_obj->obj);
+	if (IS_ERR(delta_priv))
+		return PTR_ERR(delta_priv);
+
+	/* User returned a delta private, that means that
+	 * our object can be aggregated into the parent.
+	 */
+	objagg_obj->parent = parent;
+	objagg_obj->delta_priv = delta_priv;
+	objagg_obj_ref_inc(objagg_obj->parent);
+	trace_objagg_obj_parent_assign(objagg, objagg_obj,
+				       parent,
+				       parent->refcount);
+	return 0;
+}
+
+static int objagg_obj_parent_lookup_assign(struct objagg *objagg,
+					   struct objagg_obj *objagg_obj)
+{
+	struct objagg_obj *objagg_obj_cur;
+	int err;
+
+	list_for_each_entry(objagg_obj_cur, &objagg->obj_list, list) {
+		/* Nesting is not supported. In case the object
+		 * is not root, it cannot be assigned as parent.
+		 */
+		if (!objagg_obj_is_root(objagg_obj_cur))
+			continue;
+		err = objagg_obj_parent_assign(objagg, objagg_obj,
+					       objagg_obj_cur);
+		if (!err)
+			return 0;
+	}
+	return -ENOENT;
+}
+
+static void __objagg_obj_put(struct objagg *objagg,
+			     struct objagg_obj *objagg_obj);
+
+static void objagg_obj_parent_unassign(struct objagg *objagg,
+				       struct objagg_obj *objagg_obj)
+{
+	trace_objagg_obj_parent_unassign(objagg, objagg_obj,
+					 objagg_obj->parent,
+					 objagg_obj->parent->refcount);
+	objagg->ops->delta_destroy(objagg->priv, objagg_obj->delta_priv);
+	__objagg_obj_put(objagg, objagg_obj->parent);
+}
+
+static int objagg_obj_root_create(struct objagg *objagg,
+				  struct objagg_obj *objagg_obj)
+{
+	objagg_obj->root_priv = objagg->ops->root_create(objagg->priv,
+							 objagg_obj->obj);
+	if (IS_ERR(objagg_obj->root_priv))
+		return PTR_ERR(objagg_obj->root_priv);
+
+	trace_objagg_obj_root_create(objagg, objagg_obj);
+	return 0;
+}
+
+static void objagg_obj_root_destroy(struct objagg *objagg,
+				    struct objagg_obj *objagg_obj)
+{
+	trace_objagg_obj_root_destroy(objagg, objagg_obj);
+	objagg->ops->root_destroy(objagg->priv, objagg_obj->root_priv);
+}
+
+static int objagg_obj_init(struct objagg *objagg,
+			   struct objagg_obj *objagg_obj)
+{
+	int err;
+
+	/* Try to find if the object can be aggregated under an existing one. */
+	err = objagg_obj_parent_lookup_assign(objagg, objagg_obj);
+	if (!err)
+		return 0;
+	/* If aggregation is not possible, make the object a root. */
+	return objagg_obj_root_create(objagg, objagg_obj);
+}
+
+static void objagg_obj_fini(struct objagg *objagg,
+			    struct objagg_obj *objagg_obj)
+{
+	if (!objagg_obj_is_root(objagg_obj))
+		objagg_obj_parent_unassign(objagg, objagg_obj);
+	else
+		objagg_obj_root_destroy(objagg, objagg_obj);
+}
+
+static struct objagg_obj *objagg_obj_create(struct objagg *objagg, void *obj)
+{
+	struct objagg_obj *objagg_obj;
+	int err;
+
+	objagg_obj = kzalloc(sizeof(*objagg_obj) + objagg->ops->obj_size,
+			     GFP_KERNEL);
+	if (!objagg_obj)
+		return ERR_PTR(-ENOMEM);
+	objagg_obj_ref_inc(objagg_obj);
+	memcpy(objagg_obj->obj, obj, objagg->ops->obj_size);
+
+	err = objagg_obj_init(objagg, objagg_obj);
+	if (err)
+		goto err_obj_init;
+
+	err = rhashtable_insert_fast(&objagg->obj_ht, &objagg_obj->ht_node,
+				     objagg->ht_params);
+	if (err)
+		goto err_ht_insert;
+	list_add(&objagg_obj->list, &objagg->obj_list);
+	objagg->obj_count++;
+	trace_objagg_obj_create(objagg, objagg_obj);
+
+	return objagg_obj;
+
+err_ht_insert:
+	objagg_obj_fini(objagg, objagg_obj);
+err_obj_init:
+	kfree(objagg_obj);
+	return ERR_PTR(err);
+}
+
+static struct objagg_obj *__objagg_obj_get(struct objagg *objagg, void *obj)
+{
+	struct objagg_obj *objagg_obj;
+
+	/* First, try to find the object exactly as user passed it,
+	 * perhaps it is already in use.
+	 */
+	objagg_obj = objagg_obj_lookup(objagg, obj);
+	if (objagg_obj) {
+		objagg_obj_ref_inc(objagg_obj);
+		return objagg_obj;
+	}
+
+	return objagg_obj_create(objagg, obj);
+}
+
+/**
+ * objagg_obj_get - gets an object within objagg instance
+ * @objagg:	objagg instance
+ * @obj:	user-specific private object pointer
+ *
+ * Note: all locking must be provided by the caller.
+ *
+ * Size of the "obj" memory is specified in "objagg->ops".
+ *
+ * There are 3 main options this function wraps:
+ * 1) The object according to "obj" already exist. In that case
+ *    the reference counter is incrementes and the object is returned.
+ * 2) The object does not exist, but it can be aggregated within
+ *    another object. In that case, user ops->delta_create() is called
+ *    to obtain delta data and a new object is created with returned
+ *    user-delta private pointer.
+ * 3) The object does not exist and cannot be aggregated into
+ *    any of the existing objects. In that case, user ops->root_create()
+ *    is called to create the root and a new object is created with
+ *    returned user-root private pointer.
+ *
+ * Returns a pointer to objagg object instance in case of success,
+ * otherwise it returns pointer error using ERR_PTR macro.
+ */
+struct objagg_obj *objagg_obj_get(struct objagg *objagg, void *obj)
+{
+	struct objagg_obj *objagg_obj;
+
+	objagg_obj = __objagg_obj_get(objagg, obj);
+	if (IS_ERR(objagg_obj))
+		return objagg_obj;
+	objagg_obj_stats_inc(objagg_obj);
+	trace_objagg_obj_get(objagg, objagg_obj, objagg_obj->refcount);
+	return objagg_obj;
+}
+EXPORT_SYMBOL(objagg_obj_get);
+
+static void objagg_obj_destroy(struct objagg *objagg,
+			       struct objagg_obj *objagg_obj)
+{
+	trace_objagg_obj_destroy(objagg, objagg_obj);
+	--objagg->obj_count;
+	list_del(&objagg_obj->list);
+	rhashtable_remove_fast(&objagg->obj_ht, &objagg_obj->ht_node,
+			       objagg->ht_params);
+	objagg_obj_fini(objagg, objagg_obj);
+	kfree(objagg_obj);
+}
+
+static void __objagg_obj_put(struct objagg *objagg,
+			     struct objagg_obj *objagg_obj)
+{
+	if (!objagg_obj_ref_dec(objagg_obj))
+		objagg_obj_destroy(objagg, objagg_obj);
+}
+
+/**
+ * objagg_obj_put - puts an object within objagg instance
+ * @objagg:	objagg instance
+ * @objagg_obj:	objagg object instance
+ *
+ * Note: all locking must be provided by the caller.
+ *
+ * Symmetric to objagg_obj_get().
+ */
+void objagg_obj_put(struct objagg *objagg, struct objagg_obj *objagg_obj)
+{
+	trace_objagg_obj_put(objagg, objagg_obj, objagg_obj->refcount);
+	objagg_obj_stats_dec(objagg_obj);
+	__objagg_obj_put(objagg, objagg_obj);
+}
+EXPORT_SYMBOL(objagg_obj_put);
+
+/**
+ * objagg_create - creates a new objagg instance
+ * @ops:	user-specific callbacks
+ * @priv:	pointer to a private data passed to the ops
+ *
+ * Note: all locking must be provided by the caller.
+ *
+ * The purpose of the library is to provide an infrastructure to
+ * aggregate user-specified objects. Library does not care about the type
+ * of the object. User fills-up ops which take care of the specific
+ * user object manipulation.
+ *
+ * As a very stupid example, consider integer numbers. For example
+ * number 8 as a root object. That can aggregate number 9 with delta 1,
+ * number 10 with delta 2, etc. This example is implemented as
+ * a part of a testing module in test_objagg.c file.
+ *
+ * Each objagg instance contains multiple trees. Each tree node is
+ * represented by "an object". In the current implementation there can be
+ * only roots and leafs nodes. Leaf nodes are called deltas.
+ * But in general, this can be easily extended for intermediate nodes.
+ * In that extension, a delta would be associated with all non-root
+ * nodes.
+ *
+ * Returns a pointer to newly created objagg instance in case of success,
+ * otherwise it returns pointer error using ERR_PTR macro.
+ */
+struct objagg *objagg_create(const struct objagg_ops *ops, void *priv)
+{
+	struct objagg *objagg;
+	int err;
+
+	if (WARN_ON(!ops || !ops->root_create || !ops->root_destroy ||
+		    !ops->delta_create || !ops->delta_destroy))
+		return ERR_PTR(-EINVAL);
+	objagg = kzalloc(sizeof(*objagg), GFP_KERNEL);
+	if (!objagg)
+		return ERR_PTR(-ENOMEM);
+	objagg->ops = ops;
+	objagg->priv = priv;
+	INIT_LIST_HEAD(&objagg->obj_list);
+
+	objagg->ht_params.key_len = ops->obj_size;
+	objagg->ht_params.key_offset = offsetof(struct objagg_obj, obj);
+	objagg->ht_params.head_offset = offsetof(struct objagg_obj, ht_node);
+
+	err = rhashtable_init(&objagg->obj_ht, &objagg->ht_params);
+	if (err)
+		goto err_rhashtable_init;
+
+	trace_objagg_create(objagg);
+	return objagg;
+
+err_rhashtable_init:
+	kfree(objagg);
+	return ERR_PTR(err);
+}
+EXPORT_SYMBOL(objagg_create);
+
+/**
+ * objagg_destroy - destroys a new objagg instance
+ * @objagg:	objagg instance
+ *
+ * Note: all locking must be provided by the caller.
+ */
+void objagg_destroy(struct objagg *objagg)
+{
+	trace_objagg_destroy(objagg);
+	WARN_ON(!list_empty(&objagg->obj_list));
+	rhashtable_destroy(&objagg->obj_ht);
+	kfree(objagg);
+}
+EXPORT_SYMBOL(objagg_destroy);
+
+static int objagg_stats_info_sort_cmp_func(const void *a, const void *b)
+{
+	const struct objagg_obj_stats_info *stats_info1 = a;
+	const struct objagg_obj_stats_info *stats_info2 = b;
+
+	if (stats_info1->is_root != stats_info2->is_root)
+		return stats_info2->is_root - stats_info1->is_root;
+	if (stats_info1->stats.delta_user_count !=
+	    stats_info2->stats.delta_user_count)
+		return stats_info2->stats.delta_user_count -
+		       stats_info1->stats.delta_user_count;
+	return stats_info2->stats.user_count - stats_info1->stats.user_count;
+}
+
+/**
+ * objagg_stats_get - obtains stats of the objagg instance
+ * @objagg:	objagg instance
+ *
+ * Note: all locking must be provided by the caller.
+ *
+ * The returned structure contains statistics of all object
+ * currently in use, ordered by following rules:
+ * 1) Root objects are always on lower indexes than the rest.
+ * 2) Objects with higher delta user count are always on lower
+ *    indexes.
+ * 3) In case more objects have the same delta user count,
+ *    the objects are ordered by user count.
+ *
+ * Returns a pointer to stats instance in case of success,
+ * otherwise it returns pointer error using ERR_PTR macro.
+ */
+const struct objagg_stats *objagg_stats_get(struct objagg *objagg)
+{
+	struct objagg_stats *objagg_stats;
+	struct objagg_obj *objagg_obj;
+	size_t alloc_size;
+	int i;
+
+	alloc_size = sizeof(*objagg_stats) +
+		     sizeof(objagg_stats->stats_info[0]) * objagg->obj_count;
+	objagg_stats = kzalloc(alloc_size, GFP_KERNEL);
+	if (!objagg_stats)
+		return ERR_PTR(-ENOMEM);
+
+	i = 0;
+	list_for_each_entry(objagg_obj, &objagg->obj_list, list) {
+		memcpy(&objagg_stats->stats_info[i].stats, &objagg_obj->stats,
+		       sizeof(objagg_stats->stats_info[0].stats));
+		objagg_stats->stats_info[i].objagg_obj = objagg_obj;
+		objagg_stats->stats_info[i].is_root =
+					objagg_obj_is_root(objagg_obj);
+		i++;
+	}
+	objagg_stats->stats_info_count = i;
+
+	sort(objagg_stats->stats_info, objagg_stats->stats_info_count,
+	     sizeof(struct objagg_obj_stats_info),
+	     objagg_stats_info_sort_cmp_func, NULL);
+
+	return objagg_stats;
+}
+EXPORT_SYMBOL(objagg_stats_get);
+
+/**
+ * objagg_stats_puts - puts stats of the objagg instance
+ * @objagg_stats:	objagg instance stats
+ *
+ * Note: all locking must be provided by the caller.
+ */
+void objagg_stats_put(const struct objagg_stats *objagg_stats)
+{
+	kfree(objagg_stats);
+}
+EXPORT_SYMBOL(objagg_stats_put);
+
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_AUTHOR("Jiri Pirko <jiri@mellanox.com>");
+MODULE_DESCRIPTION("Object aggregation manager");

lib/test_objagg.c

+// SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0
+/* Copyright (c) 2018 Mellanox Technologies. All rights reserved */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/random.h>
+#include <linux/objagg.h>
+
+struct tokey {
+	unsigned int id;
+};
+
+#define NUM_KEYS 32
+
+static int key_id_index(unsigned int key_id)
+{
+	if (key_id >= NUM_KEYS) {
+		WARN_ON(1);
+		return 0;
+	}
+	return key_id;
+}
+
+#define BUF_LEN 128
+
+struct world {
+	unsigned int root_count;
+	unsigned int delta_count;
+	char next_root_buf[BUF_LEN];
+	struct objagg_obj *objagg_objs[NUM_KEYS];
+	unsigned int key_refs[NUM_KEYS];
+};
+
+struct root {
+	struct tokey key;
+	char buf[BUF_LEN];
+};
+
+struct delta {
+	unsigned int key_id_diff;
+};
+
+static struct objagg_obj *world_obj_get(struct world *world,
+					struct objagg *objagg,
+					unsigned int key_id)
+{
+	struct objagg_obj *objagg_obj;
+	struct tokey key;
+	int err;
+
+	key.id = key_id;
+	objagg_obj = objagg_obj_get(objagg, &key);
+	if (IS_ERR(objagg_obj)) {
+		pr_err("Key %u: Failed to get object.\n", key_id);
+		return objagg_obj;
+	}
+	if (!world->key_refs[key_id_index(key_id)]) {
+		world->objagg_objs[key_id_index(key_id)] = objagg_obj;
+	} else if (world->objagg_objs[key_id_index(key_id)] != objagg_obj) {
+		pr_err("Key %u: God another object for the same key.\n",
+		       key_id);
+		err = -EINVAL;
+		goto err_key_id_check;
+	}
+	world->key_refs[key_id_index(key_id)]++;
+	return objagg_obj;
+
+err_key_id_check:
+	objagg_obj_put(objagg, objagg_obj);
+	return ERR_PTR(err);
+}
+
+static void world_obj_put(struct world *world, struct objagg *objagg,
+			  unsigned int key_id)
+{
+	struct objagg_obj *objagg_obj;
+
+	if (!world->key_refs[key_id_index(key_id)])
+		return;
+	objagg_obj = world->objagg_objs[key_id_index(key_id)];
+	objagg_obj_put(objagg, objagg_obj);
+	world->key_refs[key_id_index(key_id)]--;
+}
+
+#define MAX_KEY_ID_DIFF 5
+
+static void *delta_create(void *priv, void *parent_obj, void *obj)
+{
+	struct tokey *parent_key = parent_obj;
+	struct world *world = priv;
+	struct tokey *key = obj;
+	int diff = key->id - parent_key->id;
+	struct delta *delta;
+
+	if (diff < 0 || diff > MAX_KEY_ID_DIFF)
+		return ERR_PTR(-EINVAL);
+
+	delta = kzalloc(sizeof(*delta), GFP_KERNEL);
+	if (!delta)
+		return ERR_PTR(-ENOMEM);
+	delta->key_id_diff = diff;
+	world->delta_count++;
+	return delta;
+}
+
+static void delta_destroy(void *priv, void *delta_priv)
+{
+	struct delta *delta = delta_priv;
+	struct world *world = priv;
+
+	world->delta_count--;
+	kfree(delta);
+}
+
+static void *root_create(void *priv, void *obj)
+{
+	struct world *world = priv;
+	struct tokey *key = obj;
+	struct root *root;
+
+	root = kzalloc(sizeof(*root), GFP_KERNEL);
+	if (!root)
+		return ERR_PTR(-ENOMEM);
+	memcpy(&root->key, key, sizeof(root->key));
+	memcpy(root->buf, world->next_root_buf, sizeof(root->buf));
+	world->root_count++;
+	return root;
+}
+
+static void root_destroy(void *priv, void *root_priv)
+{
+	struct root *root = root_priv;
+	struct world *world = priv;
+
+	world->root_count--;
+	kfree(root);
+}
+
+static int test_nodelta_obj_get(struct world *world, struct objagg *objagg,
+				unsigned int key_id, bool should_create_root)
+{
+	unsigned int orig_root_count = world->root_count;
+	struct objagg_obj *objagg_obj;
+	const struct root *root;
+	int err;
+
+	if (should_create_root)
+		prandom_bytes(world->next_root_buf,
+			      sizeof(world->next_root_buf));
+
+	objagg_obj = world_obj_get(world, objagg, key_id);
+	if (IS_ERR(objagg_obj)) {
+		pr_err("Key %u: Failed to get object.\n", key_id);
+		return PTR_ERR(objagg_obj);
+	}
+	if (should_create_root) {
+		if (world->root_count != orig_root_count + 1) {
+			pr_err("Key %u: Root was not created\n", key_id);
+			err = -EINVAL;
+			goto err_check_root_count;
+		}
+	} else {
+		if (world->root_count != orig_root_count) {
+			pr_err("Key %u: Root was incorrectly created\n",
+			       key_id);
+			err = -EINVAL;
+			goto err_check_root_count;
+		}
+	}
+	root = objagg_obj_root_priv(objagg_obj);
+	if (root->key.id != key_id) {
+		pr_err("Key %u: Root has unexpected key id\n", key_id);
+		err = -EINVAL;
+		goto err_check_key_id;
+	}
+	if (should_create_root &&
+	    memcmp(world->next_root_buf, root->buf, sizeof(root->buf))) {
+		pr_err("Key %u: Buffer does not match the expected content\n",
+		       key_id);
+		err = -EINVAL;
+		goto err_check_buf;
+	}
+	return 0;
+
+err_check_buf:
+err_check_key_id:
+err_check_root_count:
+	objagg_obj_put(objagg, objagg_obj);
+	return err;
+}
+
+static int test_nodelta_obj_put(struct world *world, struct objagg *objagg,
+				unsigned int key_id, bool should_destroy_root)
+{
+	unsigned int orig_root_count = world->root_count;
+
+	world_obj_put(world, objagg, key_id);
+
+	if (should_destroy_root) {
+		if (world->root_count != orig_root_count - 1) {
+			pr_err("Key %u: Root was not destroyed\n", key_id);
+			return -EINVAL;
+		}
+	} else {
+		if (world->root_count != orig_root_count) {
+			pr_err("Key %u: Root was incorrectly destroyed\n",
+			       key_id);
+			return -EINVAL;
+		}
+	}
+	return 0;
+}
+
+static int check_stats_zero(struct objagg *objagg)
+{
+	const struct objagg_stats *stats;
+	int err = 0;
+
+	stats = objagg_stats_get(objagg);
+	if (IS_ERR(stats))
+		return PTR_ERR(stats);
+
+	if (stats->stats_info_count != 0) {
+		pr_err("Stats: Object count is not zero while it should be\n");
+		err = -EINVAL;
+	}
+
+	objagg_stats_put(stats);
+	return err;
+}
+
+static int check_stats_nodelta(struct objagg *objagg)
+{
+	const struct objagg_stats *stats;
+	int i;
+	int err;
+
+	stats = objagg_stats_get(objagg);
+	if (IS_ERR(stats))
+		return PTR_ERR(stats);
+
+	if (stats->stats_info_count != NUM_KEYS) {
+		pr_err("Stats: Unexpected object count (%u expected, %u returned)\n",
+		       NUM_KEYS, stats->stats_info_count);
+		err = -EINVAL;
+		goto stats_put;
+	}
+
+	for (i = 0; i < stats->stats_info_count; i++) {
+		if (stats->stats_info[i].stats.user_count != 2) {
+			pr_err("Stats: incorrect user count\n");
+			err = -EINVAL;
+			goto stats_put;
+		}
+		if (stats->stats_info[i].stats.delta_user_count != 2) {
+			pr_err("Stats: incorrect delta user count\n");
+			err = -EINVAL;
+			goto stats_put;
+		}
+	}
+	err = 0;
+
+stats_put:
+	objagg_stats_put(stats);
+	return err;
+}
+
+static void *delta_create_dummy(void *priv, void *parent_obj, void *obj)
+{
+	return ERR_PTR(-EOPNOTSUPP);
+}
+
+static void delta_destroy_dummy(void *priv, void *delta_priv)
+{
+}
+
+static const struct objagg_ops nodelta_ops = {
+	.obj_size = sizeof(struct tokey),
+	.delta_create = delta_create_dummy,
+	.delta_destroy = delta_destroy_dummy,
+	.root_create = root_create,
+	.root_destroy = root_destroy,
+};
+
+static int test_nodelta(void)
+{
+	struct world world = {};
+	struct objagg *objagg;
+	int i;
+	int err;
+
+	objagg = objagg_create(&nodelta_ops, &world);
+	if (IS_ERR(objagg))
+		return PTR_ERR(objagg);
+
+	err = check_stats_zero(objagg);
+	if (err)
+		goto err_stats_first_zero;
+
+	/* First round of gets, the root objects should be created */
+	for (i = 0; i < NUM_KEYS; i++) {
+		err = test_nodelta_obj_get(&world, objagg, i, true);
+		if (err)
+			goto err_obj_first_get;
+	}
+
+	/* Do the second round of gets, all roots are already created,
+	 * make sure that no new root is created
+	 */
+	for (i = 0; i < NUM_KEYS; i++) {
+		err = test_nodelta_obj_get(&world, objagg, i, false);
+		if (err)
+			goto err_obj_second_get;
+	}
+
+	err = check_stats_nodelta(objagg);
+	if (err)
+		goto err_stats_nodelta;
+
+	for (i = NUM_KEYS - 1; i >= 0; i--) {
+		err = test_nodelta_obj_put(&world, objagg, i, false);
+		if (err)
+			goto err_obj_first_put;
+	}
+	for (i = NUM_KEYS - 1; i >= 0; i--) {
+		err = test_nodelta_obj_put(&world, objagg, i, true);
+		if (err)
+			goto err_obj_second_put;
+	}
+
+	err = check_stats_zero(objagg);
+	if (err)
+		goto err_stats_second_zero;
+
+	objagg_destroy(objagg);
+	return 0;
+
+err_stats_nodelta:
+err_obj_first_put:
+err_obj_second_get:
+	for (i--; i >= 0; i--)
+		world_obj_put(&world, objagg, i);
+
+	i = NUM_KEYS;
+err_obj_first_get:
+err_obj_second_put:
+	for (i--; i >= 0; i--)
+		world_obj_put(&world, objagg, i);
+err_stats_first_zero:
+err_stats_second_zero:
+	objagg_destroy(objagg);
+	return err;
+}
+
+static const struct objagg_ops delta_ops = {
+	.obj_size = sizeof(struct tokey),
+	.delta_create = delta_create,
+	.delta_destroy = delta_destroy,
+	.root_create = root_create,
+	.root_destroy = root_destroy,
+};
+
+enum action {
+	ACTION_GET,
+	ACTION_PUT,
+};
+
+enum expect_delta {
+	EXPECT_DELTA_SAME,
+	EXPECT_DELTA_INC,
+	EXPECT_DELTA_DEC,
+};
+
+enum expect_root {
+	EXPECT_ROOT_SAME,
+	EXPECT_ROOT_INC,
+	EXPECT_ROOT_DEC,
+};
+
+struct expect_stats_info {
+	struct objagg_obj_stats stats;
+	bool is_root;
+	unsigned int key_id;
+};
+
+struct expect_stats {
+	unsigned int info_count;
+	struct expect_stats_info info[NUM_KEYS];
+};
+
+struct action_item {
+	unsigned int key_id;
+	enum action action;
+	enum expect_delta expect_delta;
+	enum expect_root expect_root;
+	struct expect_stats expect_stats;
+};
+
+#define EXPECT_STATS(count, ...)		\
+{						\
+	.info_count = count,			\
+	.info = { __VA_ARGS__ }			\
+}
+
+#define ROOT(key_id, user_count, delta_user_count)	\
+	{{user_count, delta_user_count}, true, key_id}
+
+#define DELTA(key_id, user_count)			\
+	{{user_count, user_count}, false, key_id}
+
+static const struct action_item action_items[] = {
+	{
+		1, ACTION_GET, EXPECT_DELTA_SAME, EXPECT_ROOT_INC,
+		EXPECT_STATS(1, ROOT(1, 1, 1)),
+	},	/* r: 1			d: */
+	{
+		7, ACTION_GET, EXPECT_DELTA_SAME, EXPECT_ROOT_INC,
+		EXPECT_STATS(2, ROOT(1, 1, 1), ROOT(7, 1, 1)),
+	},	/* r: 1, 7		d: */
+	{
+		3, ACTION_GET, EXPECT_DELTA_INC, EXPECT_ROOT_SAME,
+		EXPECT_STATS(3, ROOT(1, 1, 2), ROOT(7, 1, 1),
+				DELTA(3, 1)),
+	},	/* r: 1, 7		d: 3^1 */
+	{
+		5, ACTION_GET, EXPECT_DELTA_INC, EXPECT_ROOT_SAME,
+		EXPECT_STATS(4, ROOT(1, 1, 3), ROOT(7, 1, 1),
+				DELTA(3, 1), DELTA(5, 1)),
+	},	/* r: 1, 7		d: 3^1, 5^1 */
+	{
+		3, ACTION_GET, EXPECT_DELTA_SAME, EXPECT_ROOT_SAME,
+		EXPECT_STATS(4, ROOT(1, 1, 4), ROOT(7, 1, 1),
+				DELTA(3, 2), DELTA(5, 1)),
+	},	/* r: 1, 7		d: 3^1, 3^1, 5^1 */
+	{
+		1, ACTION_GET, EXPECT_DELTA_SAME, EXPECT_ROOT_SAME,
+		EXPECT_STATS(4, ROOT(1, 2, 5), ROOT(7, 1, 1),
+				DELTA(3, 2), DELTA(5, 1)),
+	},	/* r: 1, 1, 7		d: 3^1, 3^1, 5^1 */
+	{
+		30, ACTION_GET, EXPECT_DELTA_SAME, EXPECT_ROOT_INC,
+		EXPECT_STATS(5, ROOT(1, 2, 5), ROOT(7, 1, 1), ROOT(30, 1, 1),
+				DELTA(3, 2), DELTA(5, 1)),
+	},	/* r: 1, 1, 7, 30	d: 3^1, 3^1, 5^1 */
+	{
+		8, ACTION_GET, EXPECT_DELTA_INC, EXPECT_ROOT_SAME,
+		EXPECT_STATS(6, ROOT(1, 2, 5), ROOT(7, 1, 2), ROOT(30, 1, 1),
+				DELTA(3, 2), DELTA(5, 1), DELTA(8, 1)),
+	},	/* r: 1, 1, 7, 30	d: 3^1, 3^1, 5^1, 8^7 */
+	{
+		8, ACTION_GET, EXPECT_DELTA_SAME, EXPECT_ROOT_SAME,
+		EXPECT_STATS(6, ROOT(1, 2, 5), ROOT(7, 1, 3), ROOT(30, 1, 1),
+				DELTA(3, 2), DELTA(8, 2), DELTA(5, 1)),
+	},	/* r: 1, 1, 7, 30	d: 3^1, 3^1, 5^1, 8^7, 8^7 */
+	{
+		3, ACTION_PUT, EXPECT_DELTA_SAME, EXPECT_ROOT_SAME,
+		EXPECT_STATS(6, ROOT(1, 2, 4), ROOT(7, 1, 3), ROOT(30, 1, 1),
+				DELTA(8, 2), DELTA(3, 1), DELTA(5, 1)),
+	},	/* r: 1, 1, 7, 30	d: 3^1, 5^1, 8^7, 8^7 */
+	{
+		3, ACTION_PUT, EXPECT_DELTA_DEC, EXPECT_ROOT_SAME,
+		EXPECT_STATS(5, ROOT(1, 2, 3), ROOT(7, 1, 3), ROOT(30, 1, 1),
+				DELTA(8, 2), DELTA(5, 1)),
+	},	/* r: 1, 1, 7, 30	d: 5^1, 8^7, 8^7 */
+	{
+		1, ACTION_PUT, EXPECT_DELTA_SAME, EXPECT_ROOT_SAME,
+		EXPECT_STATS(5, ROOT(7, 1, 3), ROOT(1, 1, 2), ROOT(30, 1, 1),
+				DELTA(8, 2), DELTA(5, 1)),
+	},	/* r: 1, 7, 30		d: 5^1, 8^7, 8^7 */
+	{
+		1, ACTION_PUT, EXPECT_DELTA_SAME, EXPECT_ROOT_SAME,
+		EXPECT_STATS(5, ROOT(7, 1, 3), ROOT(30, 1, 1), ROOT(1, 0, 1),
+				DELTA(8, 2), DELTA(5, 1)),
+	},	/* r: 7, 30		d: 5^1, 8^7, 8^7 */
+	{
+		5, ACTION_PUT, EXPECT_DELTA_DEC, EXPECT_ROOT_DEC,
+		EXPECT_STATS(3, ROOT(7, 1, 3), ROOT(30, 1, 1),
+				DELTA(8, 2)),
+	},	/* r: 7, 30		d: 8^7, 8^7 */
+	{
+		5, ACTION_GET, EXPECT_DELTA_SAME, EXPECT_ROOT_INC,
+		EXPECT_STATS(4, ROOT(7, 1, 3), ROOT(30, 1, 1), ROOT(5, 1, 1),
+				DELTA(8, 2)),
+	},	/* r: 7, 30, 5		d: 8^7, 8^7 */
+	{
+		6, ACTION_GET, EXPECT_DELTA_INC, EXPECT_ROOT_SAME,
+		EXPECT_STATS(5, ROOT(7, 1, 3), ROOT(5, 1, 2), ROOT(30, 1, 1),
+				DELTA(8, 2), DELTA(6, 1)),
+	},	/* r: 7, 30, 5		d: 8^7, 8^7, 6^5 */
+	{
+		8, ACTION_GET, EXPECT_DELTA_SAME, EXPECT_ROOT_SAME,
+		EXPECT_STATS(5, ROOT(7, 1, 4), ROOT(5, 1, 2), ROOT(30, 1, 1),
+				DELTA(8, 3), DELTA(6, 1)),
+	},	/* r: 7, 30, 5		d: 8^7, 8^7, 8^7, 6^5 */
+	{
+		8, ACTION_PUT, EXPECT_DELTA_SAME, EXPECT_ROOT_SAME,
+		EXPECT_STATS(5, ROOT(7, 1, 3), ROOT(5, 1, 2), ROOT(30, 1, 1),
+				DELTA(8, 2), DELTA(6, 1)),
+	},	/* r: 7, 30, 5		d: 8^7, 8^7, 6^5 */
+	{
+		8, ACTION_PUT, EXPECT_DELTA_SAME, EXPECT_ROOT_SAME,
+		EXPECT_STATS(5, ROOT(7, 1, 2), ROOT(5, 1, 2), ROOT(30, 1, 1),
+				DELTA(8, 1), DELTA(6, 1)),
+	},	/* r: 7, 30, 5		d: 8^7, 6^5 */
+	{
+		8, ACTION_PUT, EXPECT_DELTA_DEC, EXPECT_ROOT_SAME,
+		EXPECT_STATS(4, ROOT(5, 1, 2), ROOT(7, 1, 1), ROOT(30, 1, 1),
+				DELTA(6, 1)),
+	},	/* r: 7, 30, 5		d: 6^5 */
+	{
+		8, ACTION_GET, EXPECT_DELTA_INC, EXPECT_ROOT_SAME,
+		EXPECT_STATS(5, ROOT(5, 1, 3), ROOT(7, 1, 1), ROOT(30, 1, 1),
+				DELTA(6, 1), DELTA(8, 1)),
+	},	/* r: 7, 30, 5		d: 6^5, 8^5 */
+	{
+		7, ACTION_PUT, EXPECT_DELTA_SAME, EXPECT_ROOT_DEC,
+		EXPECT_STATS(4, ROOT(5, 1, 3), ROOT(30, 1, 1),
+				DELTA(6, 1), DELTA(8, 1)),
+	},	/* r: 30, 5		d: 6^5, 8^5 */
+	{
+		30, ACTION_PUT, EXPECT_DELTA_SAME, EXPECT_ROOT_DEC,
+		EXPECT_STATS(3, ROOT(5, 1, 3),
+				DELTA(6, 1), DELTA(8, 1)),
+	},	/* r: 5			d: 6^5, 8^5 */
+	{
+		5, ACTION_PUT, EXPECT_DELTA_SAME, EXPECT_ROOT_SAME,
+		EXPECT_STATS(3, ROOT(5, 0, 2),
+				DELTA(6, 1), DELTA(8, 1)),
+	},	/* r:			d: 6^5, 8^5 */
+	{
+		6, ACTION_PUT, EXPECT_DELTA_DEC, EXPECT_ROOT_SAME,
+		EXPECT_STATS(2, ROOT(5, 0, 1),
+				DELTA(8, 1)),
+	},	/* r:			d: 6^5 */
+	{
+		8, ACTION_PUT, EXPECT_DELTA_DEC, EXPECT_ROOT_DEC,
+		EXPECT_STATS(0, ),
+	},	/* r:			d: */
+};
+
+static int check_expect(struct world *world,
+			const struct action_item *action_item,
+			unsigned int orig_delta_count,
+			unsigned int orig_root_count)
+{
+	unsigned int key_id = action_item->key_id;
+
+	switch (action_item->expect_delta) {
+	case EXPECT_DELTA_SAME:
+		if (orig_delta_count != world->delta_count) {
+			pr_err("Key %u: Delta count changed while expected to remain the same.\n",
+			       key_id);
+			return -EINVAL;
+		}
+		break;
+	case EXPECT_DELTA_INC:
+		if (WARN_ON(action_item->action == ACTION_PUT))
+			return -EINVAL;
+		if (orig_delta_count + 1 != world->delta_count) {
+			pr_err("Key %u: Delta count was not incremented.\n",
+			       key_id);
+			return -EINVAL;
+		}
+		break;
+	case EXPECT_DELTA_DEC:
+		if (WARN_ON(action_item->action == ACTION_GET))
+			return -EINVAL;
+		if (orig_delta_count - 1 != world->delta_count) {
+			pr_err("Key %u: Delta count was not decremented.\n",
+			       key_id);
+			return -EINVAL;
+		}
+		break;
+	}
+
+	switch (action_item->expect_root) {
+	case EXPECT_ROOT_SAME:
+		if (orig_root_count != world->root_count) {
+			pr_err("Key %u: Root count changed while expected to remain the same.\n",
+			       key_id);
+			return -EINVAL;
+		}
+		break;
+	case EXPECT_ROOT_INC:
+		if (WARN_ON(action_item->action == ACTION_PUT))
+			return -EINVAL;
+		if (orig_root_count + 1 != world->root_count) {
+			pr_err("Key %u: Root count was not incremented.\n",
+			       key_id);
+			return -EINVAL;
+		}
+		break;
+	case EXPECT_ROOT_DEC:
+		if (WARN_ON(action_item->action == ACTION_GET))
+			return -EINVAL;
+		if (orig_root_count - 1 != world->root_count) {
+			pr_err("Key %u: Root count was not decremented.\n",
+			       key_id);
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+
+static unsigned int obj_to_key_id(struct objagg_obj *objagg_obj)
+{
+	const struct tokey *root_key;
+	const struct delta *delta;
+	unsigned int key_id;
+
+	root_key = objagg_obj_root_priv(objagg_obj);
+	key_id = root_key->id;
+	delta = objagg_obj_delta_priv(objagg_obj);
+	if (delta)
+		key_id += delta->key_id_diff;
+	return key_id;
+}
+
+static int
+check_expect_stats_nums(const struct objagg_obj_stats_info *stats_info,
+			const struct expect_stats_info *expect_stats_info,
+			const char **errmsg)
+{
+	if (stats_info->is_root != expect_stats_info->is_root) {
+		if (errmsg)
+			*errmsg = "Incorrect root/delta indication";
+		return -EINVAL;
+	}
+	if (stats_info->stats.user_count !=
+	    expect_stats_info->stats.user_count) {
+		if (errmsg)
+			*errmsg = "Incorrect user count";
+		return -EINVAL;
+	}
+	if (stats_info->stats.delta_user_count !=
+	    expect_stats_info->stats.delta_user_count) {
+		if (errmsg)
+			*errmsg = "Incorrect delta user count";
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int
+check_expect_stats_key_id(const struct objagg_obj_stats_info *stats_info,
+			  const struct expect_stats_info *expect_stats_info,
+			  const char **errmsg)
+{
+	if (obj_to_key_id(stats_info->objagg_obj) !=
+	    expect_stats_info->key_id) {
+		if (errmsg)
+			*errmsg = "incorrect key id";
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static int check_expect_stats_neigh(const struct objagg_stats *stats,
+				    const struct expect_stats *expect_stats,
+				    int pos)
+{
+	int i;
+	int err;
+
+	for (i = pos - 1; i >= 0; i--) {
+		err = check_expect_stats_nums(&stats->stats_info[i],
+					      &expect_stats->info[pos], NULL);
+		if (err)
+			break;
+		err = check_expect_stats_key_id(&stats->stats_info[i],
+						&expect_stats->info[pos], NULL);
+		if (!err)
+			return 0;
+	}
+	for (i = pos + 1; i < stats->stats_info_count; i++) {
+		err = check_expect_stats_nums(&stats->stats_info[i],
+					      &expect_stats->info[pos], NULL);
+		if (err)
+			break;
+		err = check_expect_stats_key_id(&stats->stats_info[i],
+						&expect_stats->info[pos], NULL);
+		if (!err)
+			return 0;
+	}
+	return -EINVAL;
+}
+
+static int __check_expect_stats(const struct objagg_stats *stats,
+				const struct expect_stats *expect_stats,
+				const char **errmsg)
+{
+	int i;
+	int err;
+
+	if (stats->stats_info_count != expect_stats->info_count) {
+		*errmsg = "Unexpected object count";
+		return -EINVAL;
+	}
+
+	for (i = 0; i < stats->stats_info_count; i++) {
+		err = check_expect_stats_nums(&stats->stats_info[i],
+					      &expect_stats->info[i], errmsg);
+		if (err)
+			return err;
+		err = check_expect_stats_key_id(&stats->stats_info[i],
+						&expect_stats->info[i], errmsg);
+		if (err) {
+			/* It is possible that one of the neighbor stats with
+			 * same numbers have the correct key id, so check it
+			 */
+			err = check_expect_stats_neigh(stats, expect_stats, i);
+			if (err)
+				return err;
+		}
+	}
+	return 0;
+}
+
+static int check_expect_stats(struct objagg *objagg,
+			      const struct expect_stats *expect_stats,
+			      const char **errmsg)
+{
+	const struct objagg_stats *stats;
+	int err;
+
+	stats = objagg_stats_get(objagg);
+	if (IS_ERR(stats))
+		return PTR_ERR(stats);
+	err = __check_expect_stats(stats, expect_stats, errmsg);
+	objagg_stats_put(stats);
+	return err;
+}
+
+static int test_delta_action_item(struct world *world,
+				  struct objagg *objagg,
+				  const struct action_item *action_item,
+				  bool inverse)
+{
+	unsigned int orig_delta_count = world->delta_count;
+	unsigned int orig_root_count = world->root_count;
+	unsigned int key_id = action_item->key_id;
+	enum action action = action_item->action;
+	struct objagg_obj *objagg_obj;
+	const char *errmsg;
+	int err;
+
+	if (inverse)
+		action = action == ACTION_GET ? ACTION_PUT : ACTION_GET;
+
+	switch (action) {
+	case ACTION_GET:
+		objagg_obj = world_obj_get(world, objagg, key_id);
+		if (IS_ERR(objagg_obj))
+			return PTR_ERR(objagg_obj);
+		break;
+	case ACTION_PUT:
+		world_obj_put(world, objagg, key_id);
+		break;
+	}
+
+	if (inverse)
+		return 0;
+	err = check_expect(world, action_item,
+			   orig_delta_count, orig_root_count);
+	if (err)
+		goto errout;
+
+	err = check_expect_stats(objagg, &action_item->expect_stats, &errmsg);
+	if (err) {
+		pr_err("Key %u: Stats: %s\n", action_item->key_id, errmsg);
+		goto errout;
+	}
+
+	return 0;
+
+errout:
+	/* This can only happen when action is not inversed.
+	 * So in case of an error, cleanup by doing inverse action.
+	 */
+	test_delta_action_item(world, objagg, action_item, true);
+	return err;
+}
+
+static int test_delta(void)
+{
+	struct world world = {};
+	struct objagg *objagg;
+	int i;
+	int err;
+
+	objagg = objagg_create(&delta_ops, &world);
+	if (IS_ERR(objagg))
+		return PTR_ERR(objagg);
+
+	for (i = 0; i < ARRAY_SIZE(action_items); i++) {
+		err = test_delta_action_item(&world, objagg,
+					     &action_items[i], false);
+		if (err)
+			goto err_do_action_item;
+	}
+
+	objagg_destroy(objagg);
+	return 0;
+
+err_do_action_item:
+	for (i--; i >= 0; i--)
+		test_delta_action_item(&world, objagg, &action_items[i], true);
+
+	objagg_destroy(objagg);
+	return err;
+}
+
+static int __init test_objagg_init(void)
+{
+	int err;
+
+	err = test_nodelta();
+	if (err)
+		return err;
+	return test_delta();
+}
+
+static void __exit test_objagg_exit(void)
+{
+}
+
+module_init(test_objagg_init);
+module_exit(test_objagg_exit);
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_AUTHOR("Jiri Pirko <jiri@mellanox.com>");
+MODULE_DESCRIPTION("Test module for objagg");

tools/testing/selftests/drivers/net/mlxsw/spectrum-2/tc_flower.sh

@@ -8,7 +8,7 @@
 lib_dir=$(dirname $0)/../../../../net/forwarding
 
 ALL_TESTS="single_mask_test identical_filters_test two_masks_test \
-	multiple_masks_test ctcam_edge_cases_test"
+	multiple_masks_test ctcam_edge_cases_test delta_simple_test"
 NUM_NETIFS=2
 source $lib_dir/tc_common.sh
 source $lib_dir/lib.sh
@@ -142,7 +142,7 @@ two_masks_test()
 	tc filter add dev $h2 ingress protocol ip pref 1 handle 101 flower \
 		$tcflags dst_ip 192.0.2.2 action drop
 	tc filter add dev $h2 ingress protocol ip pref 3 handle 103 flower \
-		$tcflags dst_ip 192.0.0.0/16 action drop
+		$tcflags dst_ip 192.0.0.0/8 action drop
 
 	$MZ $h1 -c 1 -p 64 -a $h1mac -b $h2mac -A 192.0.2.1 -B 192.0.2.2 \
 		-t ip -q
@@ -235,7 +235,7 @@ ctcam_two_atcam_masks_test()
 		$tcflags dst_ip 192.0.2.2 action drop
 	# Filter goes into A-TCAM
 	tc filter add dev $h2 ingress protocol ip pref 3 handle 103 flower \
-		$tcflags dst_ip 192.0.2.0/24 action drop
+		$tcflags dst_ip 192.0.0.0/16 action drop
 
 	$MZ $h1 -c 1 -p 64 -a $h1mac -b $h2mac -A 192.0.2.1 -B 192.0.2.2 \
 		-t ip -q
@@ -324,6 +324,86 @@ ctcam_edge_cases_test()
 	ctcam_no_atcam_masks_test
 }
 
+tp_record()
+{
+	local tracepoint=$1
+	local cmd=$2
+
+	perf record -q -e $tracepoint $cmd
+	return $?
+}
+
+tp_check_hits()
+{
+	local tracepoint=$1
+	local count=$2
+
+	perf_output=`perf script -F trace:event,trace`
+	hits=`echo $perf_output | grep "$tracepoint:" | wc -l`
+	if [[ "$count" -ne "$hits" ]]; then
+		return 1
+	fi
+	return 0
+}
+
+delta_simple_test()
+{
+	# The first filter will create eRP, the second filter will fit into
+	# the first eRP with delta. Remove the first rule then and check that
+        # the eRP stays (referenced by the second filter).
+
+	RET=0
+
+	if [[ "$tcflags" != "skip_sw" ]]; then
+		return 0;
+	fi
+
+	tp_record "objagg:*" "tc filter add dev $h2 ingress protocol ip \
+		   pref 1 handle 101 flower $tcflags dst_ip 192.0.0.0/24 \
+		   action drop"
+	tp_check_hits "objagg:objagg_obj_root_create" 1
+	check_err $? "eRP was not created"
+
+	tp_record "objagg:*" "tc filter add dev $h2 ingress protocol ip \
+		   pref 2 handle 102 flower $tcflags dst_ip 192.0.2.2 \
+		   action drop"
+	tp_check_hits "objagg:objagg_obj_root_create" 0
+	check_err $? "eRP was incorrectly created"
+	tp_check_hits "objagg:objagg_obj_parent_assign" 1
+	check_err $? "delta was not created"
+
+	$MZ $h1 -c 1 -p 64 -a $h1mac -b $h2mac -A 192.0.2.1 -B 192.0.2.2 \
+		-t ip -q
+
+	tc_check_packets "dev $h2 ingress" 101 1
+	check_fail $? "Matched a wrong filter"
+
+	tc_check_packets "dev $h2 ingress" 102 1
+	check_err $? "Did not match on correct filter"
+
+	tp_record "objagg:*" "tc filter del dev $h2 ingress protocol ip \
+		   pref 1 handle 101 flower"
+	tp_check_hits "objagg:objagg_obj_root_destroy" 0
+	check_err $? "eRP was incorrectly destroyed"
+	tp_check_hits "objagg:objagg_obj_parent_unassign" 0
+	check_err $? "delta was incorrectly destroyed"
+
+	$MZ $h1 -c 1 -p 64 -a $h1mac -b $h2mac -A 192.0.2.1 -B 192.0.2.2 \
+		-t ip -q
+
+	tc_check_packets "dev $h2 ingress" 102 2
+	check_err $? "Did not match on correct filter after the first was removed"
+
+	tp_record "objagg:*" "tc filter del dev $h2 ingress protocol ip \
+		   pref 2 handle 102 flower"
+	tp_check_hits "objagg:objagg_obj_parent_unassign" 1
+	check_err $? "delta was not destroyed"
+	tp_check_hits "objagg:objagg_obj_root_destroy" 1
+	check_err $? "eRP was not destroyed"
+
+	log_test "delta simple test ($tcflags)"
+}
+
 setup_prepare()
 {
 	h1=${NETIFS[p1]}