Optional counter statistics support

index c1cef8f35e00..399ea52171fe 100644
@@ -342,7 +342,7 @@ struct mlx5_ifc_flow_table_fields_supported_bits {
 	u8         outer_geneve_oam[0x1];
 	u8         outer_geneve_protocol_type[0x1];
 	u8         outer_geneve_opt_len[0x1];
-	u8         reserved_at_1e[0x1];
+	u8         source_vhca_port[0x1];
 	u8         source_eswitch_port[0x1];
 
 	u8         inner_dmac[0x1];

@@ -393,6 +393,14 @@ struct mlx5_ifc_flow_table_fields_supported_bits {
 	u8         metadata_reg_c_0[0x1];
 };
 
+struct mlx5_ifc_flow_table_fields_supported_2_bits {
+	u8         reserved_at_0[0xe];
+	u8         bth_opcode[0x1];
+	u8         reserved_at_f[0x11];
+
+	u8         reserved_at_20[0x60];
+};
+
 struct mlx5_ifc_flow_table_prop_layout_bits {
 	u8         ft_support[0x1];
 	u8         reserved_at_1[0x1];

@@ -539,7 +547,7 @@ struct mlx5_ifc_fte_match_set_misc_bits {
 	union mlx5_ifc_gre_key_bits gre_key;
 
 	u8         vxlan_vni[0x18];
-	u8         reserved_at_b8[0x8];
+	u8         bth_opcode[0x8];
 
 	u8         geneve_vni[0x18];
 	u8         reserved_at_d8[0x7];

@@ -756,7 +764,15 @@ struct mlx5_ifc_flow_table_nic_cap_bits {
 
 	struct mlx5_ifc_flow_table_prop_layout_bits flow_table_properties_nic_transmit_sniffer;
 
-	u8         reserved_at_e00[0x1200];
+	u8         reserved_at_e00[0x700];
+
+	struct mlx5_ifc_flow_table_fields_supported_2_bits ft_field_support_2_nic_receive_rdma;
+
+	u8         reserved_at_1580[0x280];
+
+	struct mlx5_ifc_flow_table_fields_supported_2_bits ft_field_support_2_nic_transmit_rdma;
+
+	u8         reserved_at_1880[0x780];
 
 	u8         sw_steering_nic_rx_action_drop_icm_address[0x40];
 

index fe501ba88bea..71a08f84d49d 100644
@@ -99,6 +99,9 @@
 #define LEFTOVERS_NUM_LEVELS 1
 #define LEFTOVERS_NUM_PRIOS 1
 
+#define RDMA_RX_COUNTERS_PRIO_NUM_LEVELS 1
+#define RDMA_TX_COUNTERS_PRIO_NUM_LEVELS 1
+
 #define BY_PASS_PRIO_NUM_LEVELS 1
 #define BY_PASS_MIN_LEVEL (ETHTOOL_MIN_LEVEL + MLX5_BY_PASS_NUM_PRIOS +\
 			   LEFTOVERS_NUM_PRIOS)

@@ -206,34 +209,63 @@ static struct init_tree_node egress_root_fs = {
 	}
 };
 
-#define RDMA_RX_BYPASS_PRIO 0
-#define RDMA_RX_KERNEL_PRIO 1
+enum {
+	RDMA_RX_COUNTERS_PRIO,
+	RDMA_RX_BYPASS_PRIO,
+	RDMA_RX_KERNEL_PRIO,
+};
+
+#define RDMA_RX_BYPASS_MIN_LEVEL MLX5_BY_PASS_NUM_REGULAR_PRIOS
+#define RDMA_RX_KERNEL_MIN_LEVEL (RDMA_RX_BYPASS_MIN_LEVEL + 1)
+#define RDMA_RX_COUNTERS_MIN_LEVEL (RDMA_RX_KERNEL_MIN_LEVEL + 2)
+
 static struct init_tree_node rdma_rx_root_fs = {
 	.type = FS_TYPE_NAMESPACE,
-	.ar_size = 2,
+	.ar_size = 3,
 	.children = (struct init_tree_node[]) {
+		[RDMA_RX_COUNTERS_PRIO] =
+		ADD_PRIO(0, RDMA_RX_COUNTERS_MIN_LEVEL, 0,
+			 FS_CHAINING_CAPS,
+			 ADD_NS(MLX5_FLOW_TABLE_MISS_ACTION_DEF,
+				ADD_MULTIPLE_PRIO(MLX5_RDMA_RX_NUM_COUNTERS_PRIOS,
+						  RDMA_RX_COUNTERS_PRIO_NUM_LEVELS))),
 		[RDMA_RX_BYPASS_PRIO] =
-		ADD_PRIO(0, MLX5_BY_PASS_NUM_REGULAR_PRIOS, 0,
+		ADD_PRIO(0, RDMA_RX_BYPASS_MIN_LEVEL, 0,
 			 FS_CHAINING_CAPS,
 			 ADD_NS(MLX5_FLOW_TABLE_MISS_ACTION_DEF,
 				ADD_MULTIPLE_PRIO(MLX5_BY_PASS_NUM_REGULAR_PRIOS,
 						  BY_PASS_PRIO_NUM_LEVELS))),
 		[RDMA_RX_KERNEL_PRIO] =
-		ADD_PRIO(0, MLX5_BY_PASS_NUM_REGULAR_PRIOS + 1, 0,
+		ADD_PRIO(0, RDMA_RX_KERNEL_MIN_LEVEL, 0,
 			 FS_CHAINING_CAPS,
 			 ADD_NS(MLX5_FLOW_TABLE_MISS_ACTION_SWITCH_DOMAIN,
 				ADD_MULTIPLE_PRIO(1, 1))),
 	}
 };
 
+enum {
+	RDMA_TX_COUNTERS_PRIO,
+	RDMA_TX_BYPASS_PRIO,
+};
+
+#define RDMA_TX_BYPASS_MIN_LEVEL MLX5_BY_PASS_NUM_PRIOS
+#define RDMA_TX_COUNTERS_MIN_LEVEL (RDMA_TX_BYPASS_MIN_LEVEL + 1)
+
 static struct init_tree_node rdma_tx_root_fs = {
 	.type = FS_TYPE_NAMESPACE,
-	.ar_size = 1,
+	.ar_size = 2,
 	.children = (struct init_tree_node[]) {
-		ADD_PRIO(0, MLX5_BY_PASS_NUM_PRIOS, 0,
+		[RDMA_TX_COUNTERS_PRIO] =
+		ADD_PRIO(0, RDMA_TX_COUNTERS_MIN_LEVEL, 0,
+			 FS_CHAINING_CAPS,
+			 ADD_NS(MLX5_FLOW_TABLE_MISS_ACTION_DEF,
+				ADD_MULTIPLE_PRIO(MLX5_RDMA_TX_NUM_COUNTERS_PRIOS,
+						  RDMA_TX_COUNTERS_PRIO_NUM_LEVELS))),
+		[RDMA_TX_BYPASS_PRIO] =
+		ADD_PRIO(0, RDMA_TX_BYPASS_MIN_LEVEL, 0,
 			 FS_CHAINING_CAPS_RDMA_TX,
 			 ADD_NS(MLX5_FLOW_TABLE_MISS_ACTION_DEF,
-				ADD_MULTIPLE_PRIO(MLX5_BY_PASS_NUM_PRIOS,
+				ADD_MULTIPLE_PRIO(RDMA_TX_BYPASS_MIN_LEVEL,
 						  BY_PASS_PRIO_NUM_LEVELS))),
 	}
 };

@@ -2215,6 +2247,12 @@ struct mlx5_flow_namespace *mlx5_get_flow_namespace(struct mlx5_core_dev *dev,
 		prio = RDMA_RX_KERNEL_PRIO;
 	} else if (type == MLX5_FLOW_NAMESPACE_RDMA_TX) {
 		root_ns = steering->rdma_tx_root_ns;
+	} else if (type == MLX5_FLOW_NAMESPACE_RDMA_RX_COUNTERS) {
+		root_ns = steering->rdma_rx_root_ns;
+		prio = RDMA_RX_COUNTERS_PRIO;
+	} else if (type == MLX5_FLOW_NAMESPACE_RDMA_TX_COUNTERS) {
+		root_ns = steering->rdma_tx_root_ns;
+		prio = RDMA_TX_COUNTERS_PRIO;
 	} else { /* Must be NIC RX */
 		root_ns = steering->root_ns;
 		prio = type;

index 66eaf0aa7f69..ed0230ff9422 100644
@@ -1456,6 +1456,8 @@ static inline u16 mlx5_to_sw_pkey_sz(int pkey_sz)
 	return MLX5_MIN_PKEY_TABLE_SIZE << pkey_sz;
 }
 
+#define MLX5_RDMA_RX_NUM_COUNTERS_PRIOS 2
+#define MLX5_RDMA_TX_NUM_COUNTERS_PRIOS 1
 #define MLX5_BY_PASS_NUM_REGULAR_PRIOS 16
 #define MLX5_BY_PASS_NUM_DONT_TRAP_PRIOS 16
 #define MLX5_BY_PASS_NUM_MULTICAST_PRIOS 1

index 0106c67e8ccb..f2c3da2006d9 100644
@@ -83,6 +83,8 @@ enum mlx5_flow_namespace_type {
 	MLX5_FLOW_NAMESPACE_RDMA_RX,
 	MLX5_FLOW_NAMESPACE_RDMA_RX_KERNEL,
 	MLX5_FLOW_NAMESPACE_RDMA_TX,
+	MLX5_FLOW_NAMESPACE_RDMA_RX_COUNTERS,
+	MLX5_FLOW_NAMESPACE_RDMA_TX_COUNTERS,
 };
 
 enum {

index e9b4b2cccaa0..3f6b98a87566 100644
@@ -969,7 +969,8 @@ static int fill_stat_counter_hwcounters(struct sk_buff *msg,
 		return -EMSGSIZE;
 
 	for (i = 0; i < st->num_counters; i++)
-		if (rdma_nl_stat_hwcounter_entry(msg, st->names[i], st->value[i]))
+		if (rdma_nl_stat_hwcounter_entry(msg, st->descs[i].name,
+						 st->value[i]))
 			goto err;
 
 	nla_nest_end(msg, table_attr);

@@ -2105,7 +2106,8 @@ static int stat_get_doit_default_counter(struct sk_buff *skb,
 	for (i = 0; i < num_cnts; i++) {
 		v = stats->value[i] +
 			rdma_counter_get_hwstat_value(device, port, i);
-		if (rdma_nl_stat_hwcounter_entry(msg, stats->names[i], v)) {
+		if (rdma_nl_stat_hwcounter_entry(msg,
+						 stats->descs[i].name, v)) {
 			ret = -EMSGSIZE;
 			goto err_table;
 		}

index 6146c3c1cbe5..c3663cfdcd52 100644
@@ -895,7 +895,7 @@ alloc_hw_stats_device(struct ib_device *ibdev)
 	stats = ibdev->ops.alloc_hw_device_stats(ibdev);
 	if (!stats)
 		return ERR_PTR(-ENOMEM);
-	if (!stats->names || stats->num_counters <= 0)
+	if (!stats->descs || stats->num_counters <= 0)
 		goto err_free_stats;
 
 	/*

@@ -957,7 +957,7 @@ int ib_setup_device_attrs(struct ib_device *ibdev)
 	for (i = 0; i < data->stats->num_counters; i++) {
 		attr = &data->attrs[i];
 		sysfs_attr_init(&attr->attr.attr);
-		attr->attr.attr.name = data->stats->names[i];
+		attr->attr.attr.name = data->stats->descs[i].name;
 		attr->attr.attr.mode = 0444;
 		attr->attr.show = hw_stat_device_show;
 		attr->show = show_hw_stats;

@@ -994,7 +994,7 @@ alloc_hw_stats_port(struct ib_port *port, struct attribute_group *group)
 	stats = ibdev->ops.alloc_hw_port_stats(port->ibdev, port->port_num);
 	if (!stats)
 		return ERR_PTR(-ENOMEM);
-	if (!stats->names || stats->num_counters <= 0)
+	if (!stats->descs || stats->num_counters <= 0)
 		goto err_free_stats;
 
 	/*

@@ -1047,7 +1047,7 @@ static int setup_hw_port_stats(struct ib_port *port,
 	for (i = 0; i < data->stats->num_counters; i++) {
 		attr = &data->attrs[i];
 		sysfs_attr_init(&attr->attr.attr);
-		attr->attr.attr.name = data->stats->names[i];
+		attr->attr.attr.name = data->stats->descs[i].name;
 		attr->attr.attr.mode = 0444;
 		attr->attr.show = hw_stat_port_show;
 		attr->show = show_hw_stats;

148 lines not on this page · raw message

index e7337662aff8..0c8fd5a85fcb 100644
@@ -366,23 +366,23 @@ enum counters {
 	NR_COUNTERS
 };
 
-static const char * const names[] = {
-	[IP4INSEGS] = "ip4InSegs",
-	[IP4OUTSEGS] = "ip4OutSegs",
-	[IP4RETRANSSEGS] = "ip4RetransSegs",
-	[IP4OUTRSTS] = "ip4OutRsts",
-	[IP6INSEGS] = "ip6InSegs",
-	[IP6OUTSEGS] = "ip6OutSegs",
-	[IP6RETRANSSEGS] = "ip6RetransSegs",
-	[IP6OUTRSTS] = "ip6OutRsts"
+static const struct rdma_stat_desc cxgb4_descs[] = {
+	[IP4INSEGS].name = "ip4InSegs",
+	[IP4OUTSEGS].name = "ip4OutSegs",
+	[IP4RETRANSSEGS].name = "ip4RetransSegs",
+	[IP4OUTRSTS].name = "ip4OutRsts",
+	[IP6INSEGS].name = "ip6InSegs",
+	[IP6OUTSEGS].name = "ip6OutSegs",
+	[IP6RETRANSSEGS].name = "ip6RetransSegs",
+	[IP6OUTRSTS].name = "ip6OutRsts"
 };
 
 static struct rdma_hw_stats *c4iw_alloc_device_stats(struct ib_device *ibdev)
 {
-	BUILD_BUG_ON(ARRAY_SIZE(names) != NR_COUNTERS);
+	BUILD_BUG_ON(ARRAY_SIZE(cxgb4_descs) != NR_COUNTERS);
 
 	/* FIXME: these look like port stats */
-	return rdma_alloc_hw_stats_struct(names, NR_COUNTERS,
+	return rdma_alloc_hw_stats_struct(cxgb4_descs, NR_COUNTERS,
 					  RDMA_HW_STATS_DEFAULT_LIFESPAN);
 }
 

index e5f9d90aad5e..35d818b38e77 100644
@@ -60,13 +60,14 @@ struct efa_user_mmap_entry {
 	op(EFA_RDMA_READ_RESP_BYTES, "rdma_read_resp_bytes") \
 
 #define EFA_STATS_ENUM(ename, name) ename,
-#define EFA_STATS_STR(ename, name) [ename] = name,
+#define EFA_STATS_STR(ename, nam) \
+	[ename].name = nam,
 
 enum efa_hw_device_stats {
 	EFA_DEFINE_DEVICE_STATS(EFA_STATS_ENUM)
 };
 
-static const char *const efa_device_stats_names[] = {
+static const struct rdma_stat_desc efa_device_stats_descs[] = {
 	EFA_DEFINE_DEVICE_STATS(EFA_STATS_STR)
 };
 

@@ -74,7 +75,7 @@ enum efa_hw_port_stats {
 	EFA_DEFINE_PORT_STATS(EFA_STATS_ENUM)
 };
 
-static const char *const efa_port_stats_names[] = {
+static const struct rdma_stat_desc efa_port_stats_descs[] = {
 	EFA_DEFINE_PORT_STATS(EFA_STATS_STR)
 };
 

@@ -1906,15 +1907,15 @@ int efa_destroy_ah(struct ib_ah *ibah, u32 flags)
 struct rdma_hw_stats *efa_alloc_hw_port_stats(struct ib_device *ibdev,
 					      u32 port_num)
 {
-	return rdma_alloc_hw_stats_struct(efa_port_stats_names,
-					  ARRAY_SIZE(efa_port_stats_names),
+	return rdma_alloc_hw_stats_struct(efa_port_stats_descs,
+					  ARRAY_SIZE(efa_port_stats_descs),
 					  RDMA_HW_STATS_DEFAULT_LIFESPAN);
 }
 
 struct rdma_hw_stats *efa_alloc_hw_device_stats(struct ib_device *ibdev)
 {
-	return rdma_alloc_hw_stats_struct(efa_device_stats_names,
-					  ARRAY_SIZE(efa_device_stats_names),
+	return rdma_alloc_hw_stats_struct(efa_device_stats_descs,
+					  ARRAY_SIZE(efa_device_stats_descs),
 					  RDMA_HW_STATS_DEFAULT_LIFESPAN);
 }
 

@@ -1939,7 +1940,7 @@ static int efa_fill_device_stats(struct efa_dev *dev,
 	stats->value[EFA_CREATE_AH_ERR] = atomic64_read(&s->create_ah_err);
 	stats->value[EFA_MMAP_ERR] = atomic64_read(&s->mmap_err);
 
-	return ARRAY_SIZE(efa_device_stats_names);
+	return ARRAY_SIZE(efa_device_stats_descs);
 }
 
 static int efa_fill_port_stats(struct efa_dev *dev, struct rdma_hw_stats *stats,

@@ -1988,7 +1989,7 @@ static int efa_fill_port_stats(struct efa_dev *dev, struct rdma_hw_stats *stats,
 	stats->value[EFA_RDMA_READ_WR_ERR] = rrs->read_wr_err;
 	stats->value[EFA_RDMA_READ_RESP_BYTES] = rrs->read_resp_bytes;
 
-	return ARRAY_SIZE(efa_port_stats_names);
+	return ARRAY_SIZE(efa_port_stats_descs);
 }
 
 int efa_get_hw_stats(struct ib_device *ibdev, struct rdma_hw_stats *stats,

116 lines not on this page · raw message

149 lines not on this page · raw message

124 lines not on this page · raw message

index c60f6e9ac640..d84023b4b1b8 100644
@@ -601,7 +601,7 @@ struct mlx4_ib_counters {
 #define MLX4_DIAG_COUNTERS_TYPES 2
 
 struct mlx4_ib_diag_counters {
-	const char **name;
+	struct rdma_stat_desc *descs;
 	u32 *offset;
 	u32 num_counters;
 };

132 lines not on this page · raw message

index bf20a388eabe..6f5451d96dd7 100644
@@ -798,7 +798,7 @@ struct mlx5_ib_resources {
 };
 
 struct mlx5_ib_counters {
-	const char **names;
+	struct rdma_stat_desc *descs;
 	size_t *offsets;
 	u32 num_q_counters;
 	u32 num_cong_counters;

61 lines not on this page · raw message

index 4b50d9a3018a..aa1e1029b736 100644
@@ -545,6 +545,14 @@ enum ib_port_speed {
 	IB_SPEED_NDR	= 128,
 };
 
+/**
+ * struct rdma_stat_desc
+ * @name - The name of the counter
+ */
+struct rdma_stat_desc {
+	const char *name;
+};
+
 /**
  * struct rdma_hw_stats
  * @lock - Mutex to protect parallel write access to lifespan and values

@@ -555,8 +563,8 @@ enum ib_port_speed {
  *   should be before being updated again.  Stored in jiffies, defaults
  *   to 10 milliseconds, drivers can override the default be specifying
  *   their own value during their allocation routine.
- * @name - Array of pointers to static names used for the counters in
- *   directory.
+ * @descs - Array of pointers to static descriptors used for the counters
+ *   in directory.
  * @num_counters - How many hardware counters there are.  If name is
  *   shorter than this number, a kernel oops will result.  Driver authors
  *   are encouraged to leave BUILD_BUG_ON(ARRAY_SIZE(@name) < num_counters)

@@ -568,7 +576,7 @@ struct rdma_hw_stats {
 	struct mutex	lock; /* Protect lifespan and values[] */
 	unsigned long	timestamp;
 	unsigned long	lifespan;
-	const char * const *names;
+	const struct rdma_stat_desc *descs;
 	int		num_counters;
 	u64		value[];
 };

@@ -577,12 +585,12 @@ struct rdma_hw_stats {
 /**
  * rdma_alloc_hw_stats_struct - Helper function to allocate dynamic struct
  *   for drivers.
- * @names - Array of static const char *
+ * @descs - Array of static descriptors
  * @num_counters - How many elements in array
  * @lifespan - How many milliseconds between updates
  */
 static inline struct rdma_hw_stats *rdma_alloc_hw_stats_struct(
-		const char * const *names, int num_counters,
+		const struct rdma_stat_desc *descs, int num_counters,
 		unsigned long lifespan)
 {
 	struct rdma_hw_stats *stats;

@@ -591,7 +599,8 @@ static inline struct rdma_hw_stats *rdma_alloc_hw_stats_struct(
 			GFP_KERNEL);
 	if (!stats)
 		return NULL;
-	stats->names = names;
+
+	stats->descs = descs;
 	stats->num_counters = num_counters;
 	stats->lifespan = msecs_to_jiffies(lifespan);
 

index df9e6c5e4ddf..331cd29f0d61 100644
@@ -165,7 +165,7 @@ static struct rdma_counter *alloc_and_bind(struct ib_device *dev, u32 port,
 	return counter;
 
 err_mode:
-	kfree(counter->stats);
+	rdma_free_hw_stats_struct(counter->stats);
 err_stats:
 	rdma_restrack_put(&counter->res);
 	kfree(counter);

@@ -186,7 +186,7 @@ static void rdma_counter_free(struct rdma_counter *counter)
 	mutex_unlock(&port_counter->lock);
 
 	rdma_restrack_del(&counter->res);
-	kfree(counter->stats);
+	rdma_free_hw_stats_struct(counter->stats);
 	kfree(counter);
 }
 

@@ -618,7 +618,7 @@ void rdma_counter_init(struct ib_device *dev)
 fail:
 	for (i = port; i >= rdma_start_port(dev); i--) {
 		port_counter = &dev->port_data[port].port_counter;
-		kfree(port_counter->hstats);
+		rdma_free_hw_stats_struct(port_counter->hstats);
 		port_counter->hstats = NULL;
 		mutex_destroy(&port_counter->lock);
 	}

@@ -631,7 +631,7 @@ void rdma_counter_release(struct ib_device *dev)
 
 	rdma_for_each_port(dev, port) {
 		port_counter = &dev->port_data[port].port_counter;
-		kfree(port_counter->hstats);
+		rdma_free_hw_stats_struct(port_counter->hstats);
 		mutex_destroy(&port_counter->lock);
 	}
 }

index c3663cfdcd52..8d831d4fd2ad 100644
@@ -755,7 +755,7 @@ static void ib_port_release(struct kobject *kobj)
 	for (i = 0; i != ARRAY_SIZE(port->groups); i++)
 		kfree(port->groups[i].attrs);
 	if (port->hw_stats_data)
-		kfree(port->hw_stats_data->stats);
+		rdma_free_hw_stats_struct(port->hw_stats_data->stats);
 	kfree(port->hw_stats_data);
 	kfree(port);
 }

@@ -919,14 +919,14 @@ alloc_hw_stats_device(struct ib_device *ibdev)
 err_free_data:
 	kfree(data);
 err_free_stats:
-	kfree(stats);
+	rdma_free_hw_stats_struct(stats);
 	return ERR_PTR(-ENOMEM);
 }
 
 void ib_device_release_hw_stats(struct hw_stats_device_data *data)
 {
 	kfree(data->group.attrs);
-	kfree(data->stats);
+	rdma_free_hw_stats_struct(data->stats);
 	kfree(data);
 }
 

@@ -1018,7 +1018,7 @@ alloc_hw_stats_port(struct ib_port *port, struct attribute_group *group)
 err_free_data:
 	kfree(data);
 err_free_stats:
-	kfree(stats);
+	rdma_free_hw_stats_struct(stats);
 	return ERR_PTR(-ENOMEM);
 }
 

index 89a2b21976d6..c3319a7584a2 100644
@@ -2976,3 +2976,38 @@ bool __rdma_block_iter_next(struct ib_block_iter *biter)
 	return true;
 }
 EXPORT_SYMBOL(__rdma_block_iter_next);
+
+/**
+ * rdma_alloc_hw_stats_struct - Helper function to allocate dynamic struct
+ *   for the drivers.
+ * @descs: array of static descriptors
+ * @num_counters: number of elments in array
+ * @lifespan: milliseconds between updates
+ */
+struct rdma_hw_stats *rdma_alloc_hw_stats_struct(
+	const struct rdma_stat_desc *descs, int num_counters,
+	unsigned long lifespan)
+{
+	struct rdma_hw_stats *stats;
+
+	stats = kzalloc(struct_size(stats, value, num_counters), GFP_KERNEL);
+	if (!stats)
+		return NULL;
+
+	stats->descs = descs;
+	stats->num_counters = num_counters;
+	stats->lifespan = msecs_to_jiffies(lifespan);
+
+	return stats;
+}
+EXPORT_SYMBOL(rdma_alloc_hw_stats_struct);
+
+/**
+ * rdma_free_hw_stats_struct - Helper function to release rdma_hw_stats
+ * @stats: statistics to release
+ */
+void rdma_free_hw_stats_struct(struct rdma_hw_stats *stats)
+{
+	kfree(stats);
+}
+EXPORT_SYMBOL(rdma_free_hw_stats_struct);

index aa1e1029b736..938c0c0a1c19 100644
@@ -582,31 +582,12 @@ struct rdma_hw_stats {
 };
 
 #define RDMA_HW_STATS_DEFAULT_LIFESPAN 10
-/**
- * rdma_alloc_hw_stats_struct - Helper function to allocate dynamic struct
- *   for drivers.
- * @descs - Array of static descriptors
- * @num_counters - How many elements in array
- * @lifespan - How many milliseconds between updates
- */
-static inline struct rdma_hw_stats *rdma_alloc_hw_stats_struct(
-		const struct rdma_stat_desc *descs, int num_counters,
-		unsigned long lifespan)
-{
-	struct rdma_hw_stats *stats;
 
-	stats = kzalloc(sizeof(*stats) + num_counters * sizeof(u64),
-			GFP_KERNEL);
-	if (!stats)
-		return NULL;
-
-	stats->descs = descs;
-	stats->num_counters = num_counters;
-	stats->lifespan = msecs_to_jiffies(lifespan);
-
-	return stats;
-}
+struct rdma_hw_stats *rdma_alloc_hw_stats_struct(
+	const struct rdma_stat_desc *descs, int num_counters,
+	unsigned long lifespan);
 
+void rdma_free_hw_stats_struct(struct rdma_hw_stats *stats);
 
 /* Define bits for the various functionality this port needs to be supported by
  * the core.

index 3f6b98a87566..67519730b1ac 100644
@@ -968,15 +968,21 @@ static int fill_stat_counter_hwcounters(struct sk_buff *msg,
 	if (!table_attr)
 		return -EMSGSIZE;
 
-	for (i = 0; i < st->num_counters; i++)
+	mutex_lock(&st->lock);
+	for (i = 0; i < st->num_counters; i++) {
+		if (test_bit(i, st->is_disabled))
+			continue;
 		if (rdma_nl_stat_hwcounter_entry(msg, st->descs[i].name,
 						 st->value[i]))
 			goto err;
+	}
+	mutex_unlock(&st->lock);
 
 	nla_nest_end(msg, table_attr);
 	return 0;
 
 err:
+	mutex_unlock(&st->lock);
 	nla_nest_cancel(msg, table_attr);
 	return -EMSGSIZE;
 }

@@ -2104,6 +2110,9 @@ static int stat_get_doit_default_counter(struct sk_buff *skb,
 		goto err_stats;
 	}
 	for (i = 0; i < num_cnts; i++) {
+		if (test_bit(i, stats->is_disabled))
+			continue;
+
 		v = stats->value[i] +
 			rdma_counter_get_hwstat_value(device, port, i);
 		if (rdma_nl_stat_hwcounter_entry(msg,

index c3319a7584a2..d957b23a0e23 100644
@@ -2994,11 +2994,20 @@ struct rdma_hw_stats *rdma_alloc_hw_stats_struct(
 	if (!stats)
 		return NULL;
 
+	stats->is_disabled = kcalloc(BITS_TO_LONGS(num_counters),
+				     sizeof(*stats->is_disabled), GFP_KERNEL);
+	if (!stats->is_disabled)
+		goto err;
+
 	stats->descs = descs;
 	stats->num_counters = num_counters;
 	stats->lifespan = msecs_to_jiffies(lifespan);
 
 	return stats;
+
+err:
+	kfree(stats);
+	return NULL;
 }
 EXPORT_SYMBOL(rdma_alloc_hw_stats_struct);
 

@@ -3008,6 +3017,7 @@ EXPORT_SYMBOL(rdma_alloc_hw_stats_struct);
  */
 void rdma_free_hw_stats_struct(struct rdma_hw_stats *stats)
 {
+	kfree(stats->is_disabled);
 	kfree(stats);
 }
 EXPORT_SYMBOL(rdma_free_hw_stats_struct);

index 938c0c0a1c19..ae467365706b 100644
@@ -565,6 +565,8 @@ struct rdma_stat_desc {
  *   their own value during their allocation routine.
  * @descs - Array of pointers to static descriptors used for the counters
  *   in directory.
+ * @is_disabled - A bitmap to indicate each counter is currently disabled
+ *   or not.
  * @num_counters - How many hardware counters there are.  If name is
  *   shorter than this number, a kernel oops will result.  Driver authors
  *   are encouraged to leave BUILD_BUG_ON(ARRAY_SIZE(@name) < num_counters)

@@ -577,6 +579,7 @@ struct rdma_hw_stats {
 	unsigned long	timestamp;
 	unsigned long	lifespan;
 	const struct rdma_stat_desc *descs;
+	unsigned long	*is_disabled;
 	int		num_counters;
 	u64		value[];
 };

index 331cd29f0d61..af59486fe418 100644
@@ -106,6 +106,38 @@ static int __rdma_counter_bind_qp(struct rdma_counter *counter,
 	return ret;
 }
 
+int rdma_counter_modify(struct ib_device *dev, u32 port,
+			unsigned int index, bool enable)
+{
+	struct rdma_hw_stats *stats;
+	int ret = 0;
+
+	if (!dev->ops.modify_hw_stat)
+		return -EOPNOTSUPP;
+
+	stats = ib_get_hw_stats_port(dev, port);
+	if (!stats || index >= stats->num_counters ||
+	    !(stats->descs[index].flags & IB_STAT_FLAG_OPTIONAL))
+		return -EINVAL;
+
+	mutex_lock(&stats->lock);
+
+	if (enable != test_bit(index, stats->is_disabled))
+		goto out;
+
+	ret = dev->ops.modify_hw_stat(dev, port, index, enable);
+	if (ret)
+		goto out;
+
+	if (enable)
+		clear_bit(index, stats->is_disabled);
+	else
+		set_bit(index, stats->is_disabled);
+out:
+	mutex_unlock(&stats->lock);
+	return ret;
+}
+
 static struct rdma_counter *alloc_and_bind(struct ib_device *dev, u32 port,
 					   struct ib_qp *qp,
 					   enum rdma_nl_counter_mode mode)

index f4814bb7f082..22a4adda7981 100644
@@ -2676,6 +2676,7 @@ void ib_set_device_ops(struct ib_device *dev, const struct ib_device_ops *ops)
 	SET_DEVICE_OP(dev_ops, modify_cq);
 	SET_DEVICE_OP(dev_ops, modify_device);
 	SET_DEVICE_OP(dev_ops, modify_flow_action_esp);
+	SET_DEVICE_OP(dev_ops, modify_hw_stat);
 	SET_DEVICE_OP(dev_ops, modify_port);
 	SET_DEVICE_OP(dev_ops, modify_qp);
 	SET_DEVICE_OP(dev_ops, modify_srq);

index 8d831d4fd2ad..1bf3aea4b71e 100644
@@ -934,7 +934,8 @@ int ib_setup_device_attrs(struct ib_device *ibdev)
 {
 	struct hw_stats_device_attribute *attr;
 	struct hw_stats_device_data *data;
-	int i, ret;
+	bool opstat_skipped = false;
+	int i, ret, pos = 0;
 
 	data = alloc_hw_stats_device(ibdev);
 	if (IS_ERR(data)) {

@@ -955,16 +956,23 @@ int ib_setup_device_attrs(struct ib_device *ibdev)
 	data->stats->timestamp = jiffies;
 
 	for (i = 0; i < data->stats->num_counters; i++) {
-		attr = &data->attrs[i];
+		if (data->stats->descs[i].flags & IB_STAT_FLAG_OPTIONAL) {
+			opstat_skipped = true;
+			continue;
+		}
+
+		WARN_ON(opstat_skipped);
+		attr = &data->attrs[pos];
 		sysfs_attr_init(&attr->attr.attr);
 		attr->attr.attr.name = data->stats->descs[i].name;
 		attr->attr.attr.mode = 0444;
 		attr->attr.show = hw_stat_device_show;
 		attr->show = show_hw_stats;
-		data->group.attrs[i] = &attr->attr.attr;
+		data->group.attrs[pos] = &attr->attr.attr;
+		pos++;
 	}
 
-	attr = &data->attrs[i];
+	attr = &data->attrs[pos];
 	sysfs_attr_init(&attr->attr.attr);
 	attr->attr.attr.name = "lifespan";
 	attr->attr.attr.mode = 0644;

@@ -972,7 +980,7 @@ int ib_setup_device_attrs(struct ib_device *ibdev)
 	attr->show = show_stats_lifespan;
 	attr->attr.store = hw_stat_device_store;
 	attr->store = set_stats_lifespan;
-	data->group.attrs[i] = &attr->attr.attr;
+	data->group.attrs[pos] = &attr->attr.attr;
 	for (i = 0; i != ARRAY_SIZE(ibdev->groups); i++)
 		if (!ibdev->groups[i]) {
 			ibdev->groups[i] = &data->group;

@@ -1027,7 +1035,8 @@ static int setup_hw_port_stats(struct ib_port *port,
 {
 	struct hw_stats_port_attribute *attr;
 	struct hw_stats_port_data *data;
-	int i, ret;
+	bool opstat_skipped = false;
+	int i, ret, pos = 0;
 
 	data = alloc_hw_stats_port(port, group);
 	if (IS_ERR(data))

@@ -1045,16 +1054,23 @@ static int setup_hw_port_stats(struct ib_port *port,
 	data->stats->timestamp = jiffies;
 
 	for (i = 0; i < data->stats->num_counters; i++) {
-		attr = &data->attrs[i];
+		if (data->stats->descs[i].flags & IB_STAT_FLAG_OPTIONAL) {
+			opstat_skipped = true;
+			continue;
+		}
+
+		WARN_ON(opstat_skipped);
+		attr = &data->attrs[pos];
 		sysfs_attr_init(&attr->attr.attr);
 		attr->attr.attr.name = data->stats->descs[i].name;
 		attr->attr.attr.mode = 0444;
 		attr->attr.show = hw_stat_port_show;
 		attr->show = show_hw_stats;
-		group->attrs[i] = &attr->attr.attr;
+		group->attrs[pos] = &attr->attr.attr;
+		pos++;
 	}
 
-	attr = &data->attrs[i];
+	attr = &data->attrs[pos];
 	sysfs_attr_init(&attr->attr.attr);
 	attr->attr.attr.name = "lifespan";
 	attr->attr.attr.mode = 0644;

@@ -1062,7 +1078,7 @@ static int setup_hw_port_stats(struct ib_port *port,
 	attr->show = show_stats_lifespan;
 	attr->attr.store = hw_stat_port_store;
 	attr->store = set_stats_lifespan;
-	group->attrs[i] = &attr->attr.attr;
+	group->attrs[pos] = &attr->attr.attr;
 
 	port->hw_stats_data = data;
 	return 0;

index ae467365706b..2207f60b002f 100644
@@ -545,12 +545,18 @@ enum ib_port_speed {
 	IB_SPEED_NDR	= 128,
 };
 
+enum ib_stat_flag {
+	IB_STAT_FLAG_OPTIONAL = 1 << 0,
+};
+
 /**
  * struct rdma_stat_desc
  * @name - The name of the counter
+ * @flags - Flags of the counter; For example, IB_STAT_FLAG_OPTIONAL
  */
 struct rdma_stat_desc {
 	const char *name;
+	unsigned int flags;
 };
 
 /**

@@ -2562,6 +2568,13 @@ struct ib_device_ops {
 	int (*get_hw_stats)(struct ib_device *device,
 			    struct rdma_hw_stats *stats, u32 port, int index);
 
+	/**
+	 * modify_hw_stat - Modify the counter configuration
+	 * @enable: true/false when enable/disable a counter
+	 * Return codes - 0 on success or error code otherwise.
+	 */
+	int (*modify_hw_stat)(struct ib_device *device, u32 port,
+			      unsigned int counter_index, bool enable);
 	/**
 	 * Allows rdma drivers to add their own restrack attributes.
 	 */

index 0295b22cd1cd..45d5481a7846 100644
@@ -63,4 +63,6 @@ int rdma_counter_get_mode(struct ib_device *dev, u32 port,
 			  enum rdma_nl_counter_mode *mode,
 			  enum rdma_nl_counter_mask *mask);
 
+int rdma_counter_modify(struct ib_device *dev, u32 port,
+			unsigned int index, bool enable);
 #endif /* _RDMA_COUNTER_H_ */

index 67519730b1ac..210057fef7bd 100644
@@ -154,6 +154,8 @@ static const struct nla_policy nldev_policy[RDMA_NLDEV_ATTR_MAX] = {
 	[RDMA_NLDEV_NET_NS_FD]			= { .type = NLA_U32 },
 	[RDMA_NLDEV_SYS_ATTR_NETNS_MODE]	= { .type = NLA_U8 },
 	[RDMA_NLDEV_SYS_ATTR_COPY_ON_FORK]	= { .type = NLA_U8 },
+	[RDMA_NLDEV_ATTR_STAT_HWCOUNTER_INDEX]	= { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_STAT_HWCOUNTER_DYNAMIC] = { .type = NLA_U8 },
 };
 
 static int put_driver_name_print_type(struct sk_buff *msg, const char *name,

@@ -2264,6 +2266,99 @@ static int nldev_stat_get_dumpit(struct sk_buff *skb,
 	return ret;
 }
 
+static int nldev_stat_get_counter_status_doit(struct sk_buff *skb,
+					      struct nlmsghdr *nlh,
+					      struct netlink_ext_ack *extack)
+{
+	struct nlattr *tb[RDMA_NLDEV_ATTR_MAX], *table, *entry;
+	struct rdma_hw_stats *stats;
+	struct ib_device *device;
+	struct sk_buff *msg;
+	u32 devid, port;
+	int ret, i;
+
+	ret = nlmsg_parse(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1,
+			  nldev_policy, extack);
+	if (ret || !tb[RDMA_NLDEV_ATTR_DEV_INDEX] ||
+	    !tb[RDMA_NLDEV_ATTR_PORT_INDEX])
+		return -EINVAL;
+
+	devid = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]);
+	device = ib_device_get_by_index(sock_net(skb->sk), devid);
+	if (!device)
+		return -EINVAL;
+
+	port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]);
+	if (!rdma_is_port_valid(device, port)) {
+		ret = -EINVAL;
+		goto err;
+	}
+
+	stats = ib_get_hw_stats_port(device, port);
+	if (!stats) {
+		ret = -EINVAL;
+		goto err;
+	}
+
+	msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
+	if (!msg) {
+		ret = -ENOMEM;
+		goto err;
+	}
+
+	nlh = nlmsg_put(
+		msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq,
+		RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_STAT_GET_STATUS),
+		0, 0);
+
+	ret = -EMSGSIZE;
+	if (fill_nldev_handle(msg, device) ||
+	    nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, port))
+		goto err_msg;
+
+	table = nla_nest_start(msg, RDMA_NLDEV_ATTR_STAT_HWCOUNTERS);
+	if (!table)
+		goto err_msg;
+
+	mutex_lock(&stats->lock);
+	for (i = 0; i < stats->num_counters; i++) {
+		entry = nla_nest_start(msg,
+				       RDMA_NLDEV_ATTR_STAT_HWCOUNTER_ENTRY);
+		if (!entry)
+			goto err_msg_table;
+
+		if (nla_put_string(msg,
+				   RDMA_NLDEV_ATTR_STAT_HWCOUNTER_ENTRY_NAME,
+				   stats->descs[i].name) ||
+		    nla_put_u32(msg, RDMA_NLDEV_ATTR_STAT_HWCOUNTER_INDEX, i))
+			goto err_msg_entry;
+
+		if ((stats->descs[i].flags & IB_STAT_FLAG_OPTIONAL) &&
+		    (nla_put_u8(msg, RDMA_NLDEV_ATTR_STAT_HWCOUNTER_DYNAMIC,
+				!test_bit(i, stats->is_disabled))))
+			goto err_msg_entry;
+
+		nla_nest_end(msg, entry);
+	}
+	mutex_unlock(&stats->lock);
+
+	nla_nest_end(msg, table);
+	nlmsg_end(msg, nlh);
+	ib_device_put(device);
+	return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid);
+
+err_msg_entry:
+	nla_nest_cancel(msg, entry);
+err_msg_table:
+	mutex_unlock(&stats->lock);
+	nla_nest_cancel(msg, table);
+err_msg:
+	nlmsg_free(msg);
+err:
+	ib_device_put(device);
+	return ret;
+}
+
 static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
 	[RDMA_NLDEV_CMD_GET] = {
 		.doit = nldev_get_doit,

@@ -2353,6 +2448,9 @@ static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
 		.dump = nldev_res_get_mr_raw_dumpit,
 		.flags = RDMA_NL_ADMIN_PERM,
 	},
+	[RDMA_NLDEV_CMD_STAT_GET_STATUS] = {
+		.doit = nldev_stat_get_counter_status_doit,
+	},
 };
 
 void __init nldev_init(void)

index 75a1ae2311d8..e50c357367db 100644
@@ -297,6 +297,8 @@ enum rdma_nldev_command {
 
 	RDMA_NLDEV_CMD_RES_SRQ_GET, /* can dump */
 
+	RDMA_NLDEV_CMD_STAT_GET_STATUS,
+
 	RDMA_NLDEV_NUM_OPS
 };
 

@@ -549,6 +551,9 @@ enum rdma_nldev_attr {
 
 	RDMA_NLDEV_SYS_ATTR_COPY_ON_FORK,	/* u8 */
 
+	RDMA_NLDEV_ATTR_STAT_HWCOUNTER_INDEX,	/* u32 */
+	RDMA_NLDEV_ATTR_STAT_HWCOUNTER_DYNAMIC, /* u8 */
+
 	/*
 	 * Always the end
 	 */

index 210057fef7bd..bbe3dfca3d98 100644
@@ -1897,42 +1897,23 @@ static int nldev_set_sys_set_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 	return err;
 }
 
-static int nldev_stat_set_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
-			       struct netlink_ext_ack *extack)
+static int nldev_stat_set_mode_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
+				    struct netlink_ext_ack *extack,
+				    struct nlattr *tb[],
+				    struct ib_device *device, u32 port)
 {
-	u32 index, port, mode, mask = 0, qpn, cntn = 0;
-	struct nlattr *tb[RDMA_NLDEV_ATTR_MAX];
-	struct ib_device *device;
+	u32 mode, mask = 0, qpn, cntn = 0;
 	struct sk_buff *msg;
 	int ret;
 
-	ret = nlmsg_parse(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1,
-			  nldev_policy, extack);
 	/* Currently only counter for QP is supported */
-	if (ret || !tb[RDMA_NLDEV_ATTR_STAT_RES] ||
-	    !tb[RDMA_NLDEV_ATTR_DEV_INDEX] ||
-	    !tb[RDMA_NLDEV_ATTR_PORT_INDEX] || !tb[RDMA_NLDEV_ATTR_STAT_MODE])
-		return -EINVAL;
-
 	if (nla_get_u32(tb[RDMA_NLDEV_ATTR_STAT_RES]) != RDMA_NLDEV_ATTR_RES_QP)
 		return -EINVAL;
 
-	index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]);
-	device = ib_device_get_by_index(sock_net(skb->sk), index);
-	if (!device)
-		return -EINVAL;
-
-	port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]);
-	if (!rdma_is_port_valid(device, port)) {
-		ret = -EINVAL;
-		goto err;
-	}
-
 	msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
-	if (!msg) {
-		ret = -ENOMEM;
-		goto err;
-	}
+	if (!msg)
+		return -ENOMEM;
+
 	nlh = nlmsg_put(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq,
 			RDMA_NL_GET_TYPE(RDMA_NL_NLDEV,
 					 RDMA_NLDEV_CMD_STAT_SET),

@@ -1947,8 +1928,10 @@ static int nldev_stat_set_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 		if (ret)
 			goto err_msg;
 	} else {
-		if (!tb[RDMA_NLDEV_ATTR_RES_LQPN])
+		if (!tb[RDMA_NLDEV_ATTR_RES_LQPN]) {
+			ret = -EINVAL;
 			goto err_msg;
+		}
 		qpn = nla_get_u32(tb[RDMA_NLDEV_ATTR_RES_LQPN]);
 		if (tb[RDMA_NLDEV_ATTR_STAT_COUNTER_ID]) {
 			cntn = nla_get_u32(tb[RDMA_NLDEV_ATTR_STAT_COUNTER_ID]);

@@ -1970,14 +1953,45 @@ static int nldev_stat_set_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 	}
 
 	nlmsg_end(msg, nlh);
-	ib_device_put(device);
 	return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid);
 
 err_fill:
 	rdma_counter_unbind_qpn(device, port, qpn, cntn);
 err_msg:
 	nlmsg_free(msg);
-err:
+	return ret;
+}
+
+static int nldev_stat_set_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
+			       struct netlink_ext_ack *extack)
+{
+	struct nlattr *tb[RDMA_NLDEV_ATTR_MAX];
+	struct ib_device *device;
+	u32 index, port;
+	int ret;
+
+	ret = nlmsg_parse(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, nldev_policy,
+			  extack);
+	if (ret)
+		return -EINVAL;
+
+	index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]);
+	device = ib_device_get_by_index(sock_net(skb->sk), index);
+	if (!device)
+		return -EINVAL;
+
+	port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]);
+	if (!rdma_is_port_valid(device, port)) {
+		ret = -EINVAL;
+		goto end;
+	}
+
+	if (!tb[RDMA_NLDEV_ATTR_STAT_MODE]) {
+		ret = -EINVAL;
+		goto end;
+	}
+	ret = nldev_stat_set_mode_doit(skb, nlh, extack, tb, device, port);
+end:
 	ib_device_put(device);
 	return ret;
 }

index bbe3dfca3d98..7bd4e08268a6 100644
@@ -1962,6 +1962,50 @@ static int nldev_stat_set_mode_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 	return ret;
 }
 
+static int nldev_stat_set_counter_dynamic_doit(struct nlattr *tb[],
+					       struct ib_device *device,
+					       u32 port)
+{
+	struct rdma_hw_stats *stats;
+	int rem, i, index, ret = 0;
+	struct nlattr *entry_attr;
+	unsigned long *target;
+
+	stats = ib_get_hw_stats_port(device, port);
+	if (!stats)
+		return -EINVAL;
+
+	target = kcalloc(BITS_TO_LONGS(stats->num_counters),
+			 sizeof(*stats->is_disabled), GFP_KERNEL);
+	if (!target)
+		return -ENOMEM;
+
+	nla_for_each_nested(entry_attr, tb[RDMA_NLDEV_ATTR_STAT_HWCOUNTERS],
+			    rem) {
+		index = nla_get_u32(entry_attr);
+		if ((index >= stats->num_counters) ||
+		    !(stats->descs[index].flags & IB_STAT_FLAG_OPTIONAL)) {
+			ret = -EINVAL;
+			goto out;
+		}
+
+		set_bit(index, target);
+	}
+
+	for (i = 0; i < stats->num_counters; i++) {
+		if (!(stats->descs[i].flags & IB_STAT_FLAG_OPTIONAL))
+			continue;
+
+		ret = rdma_counter_modify(device, port, i, test_bit(i, target));
+		if (ret)
+			goto out;
+	}
+
+out:
+	kfree(target);
+	return ret;
+}
+
 static int nldev_stat_set_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 			       struct netlink_ext_ack *extack)
 {

@@ -1986,11 +2030,13 @@ static int nldev_stat_set_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 		goto end;
 	}
 
-	if (!tb[RDMA_NLDEV_ATTR_STAT_MODE]) {
+	if (tb[RDMA_NLDEV_ATTR_STAT_MODE])
+		ret = nldev_stat_set_mode_doit(skb, nlh, extack, tb, device,
+					       port);
+	else if (tb[RDMA_NLDEV_ATTR_STAT_HWCOUNTER_DYNAMIC])
+		ret = nldev_stat_set_counter_dynamic_doit(tb, device, port);
+	else
 		ret = -EINVAL;
-		goto end;
-	}
-	ret = nldev_stat_set_mode_doit(skb, nlh, extack, tb, device, port);
 end:
 	ib_device_put(device);
 	return ret;

index caa35ea14b48..8fe7900b29f0 100644
@@ -75,6 +75,21 @@ static const struct mlx5_ib_counter ext_ppcnt_cnts[] = {
 	INIT_EXT_PPCNT_COUNTER(rx_icrc_encapsulated),
 };
 
+#define INIT_OP_COUNTER(_name)                                          \
+	{ .name = #_name }
+
+static const struct mlx5_ib_counter basic_op_cnts[] = {
+	INIT_OP_COUNTER(cc_rx_ce_pkts),
+};
+
+static const struct mlx5_ib_counter rdmarx_cnp_op_cnts[] = {
+	INIT_OP_COUNTER(cc_rx_cnp_pkts),
+};
+
+static const struct mlx5_ib_counter rdmatx_cnp_op_cnts[] = {
+	INIT_OP_COUNTER(cc_tx_cnp_pkts),
+};
+
 static int mlx5_ib_read_counters(struct ib_counters *counters,
 				 struct ib_counters_read_attr *read_attr,
 				 struct uverbs_attr_bundle *attrs)

@@ -161,17 +176,34 @@ u16 mlx5_ib_get_counters_id(struct mlx5_ib_dev *dev, u32 port_num)
 	return cnts->set_id;
 }
 
+static struct rdma_hw_stats *do_alloc_stats(const struct mlx5_ib_counters *cnts)
+{
+	struct rdma_hw_stats *stats;
+	u32 num_hw_counters;
+	int i;
+
+	num_hw_counters = cnts->num_q_counters + cnts->num_cong_counters +
+			  cnts->num_ext_ppcnt_counters;
+	stats = rdma_alloc_hw_stats_struct(cnts->descs,
+					   num_hw_counters +
+					   cnts->num_op_counters,
+					   RDMA_HW_STATS_DEFAULT_LIFESPAN);
+	if (!stats)
+		return NULL;
+
+	for (i = 0; i < cnts->num_op_counters; i++)
+		set_bit(num_hw_counters + i, stats->is_disabled);
+
+	return stats;
+}
+
 static struct rdma_hw_stats *
 mlx5_ib_alloc_hw_device_stats(struct ib_device *ibdev)
 {
 	struct mlx5_ib_dev *dev = to_mdev(ibdev);
 	const struct mlx5_ib_counters *cnts = &dev->port[0].cnts;
 
-	return rdma_alloc_hw_stats_struct(cnts->descs,
-					  cnts->num_q_counters +
-						  cnts->num_cong_counters +
-						  cnts->num_ext_ppcnt_counters,
-					  RDMA_HW_STATS_DEFAULT_LIFESPAN);
+	return do_alloc_stats(cnts);
 }
 
 static struct rdma_hw_stats *

@@ -180,11 +212,7 @@ mlx5_ib_alloc_hw_port_stats(struct ib_device *ibdev, u32 port_num)
 	struct mlx5_ib_dev *dev = to_mdev(ibdev);
 	const struct mlx5_ib_counters *cnts = &dev->port[port_num - 1].cnts;
 
-	return rdma_alloc_hw_stats_struct(cnts->descs,
-					  cnts->num_q_counters +
-						  cnts->num_cong_counters +
-						  cnts->num_ext_ppcnt_counters,
-					  RDMA_HW_STATS_DEFAULT_LIFESPAN);
+	return do_alloc_stats(cnts);
 }
 
 static int mlx5_ib_query_q_counters(struct mlx5_core_dev *mdev,

@@ -302,11 +330,7 @@ mlx5_ib_counter_alloc_stats(struct rdma_counter *counter)
 	const struct mlx5_ib_counters *cnts =
 		get_counters(dev, counter->port - 1);
 
-	return rdma_alloc_hw_stats_struct(cnts->descs,
-					  cnts->num_q_counters +
-					  cnts->num_cong_counters +
-					  cnts->num_ext_ppcnt_counters,
-					  RDMA_HW_STATS_DEFAULT_LIFESPAN);
+	return do_alloc_stats(cnts);
 }
 
 static int mlx5_ib_counter_update_stats(struct rdma_counter *counter)

@@ -423,13 +447,34 @@ static void mlx5_ib_fill_counters(struct mlx5_ib_dev *dev,
 			offsets[j] = ext_ppcnt_cnts[i].offset;
 		}
 	}
+
+	for (i = 0; i < ARRAY_SIZE(basic_op_cnts); i++, j++) {
+		descs[j].name = basic_op_cnts[i].name;
+		descs[j].flags |= IB_STAT_FLAG_OPTIONAL;
+	}
+
+	if (MLX5_CAP_FLOWTABLE(dev->mdev,
+			       ft_field_support_2_nic_receive_rdma.bth_opcode)) {
+		for (i = 0; i < ARRAY_SIZE(rdmarx_cnp_op_cnts); i++, j++) {
+			descs[j].name = rdmarx_cnp_op_cnts[i].name;
+			descs[j].flags |= IB_STAT_FLAG_OPTIONAL;
+		}
+	}
+
+	if (MLX5_CAP_FLOWTABLE(dev->mdev,
+			       ft_field_support_2_nic_transmit_rdma.bth_opcode)) {
+		for (i = 0; i < ARRAY_SIZE(rdmatx_cnp_op_cnts); i++, j++) {
+			descs[j].name = rdmatx_cnp_op_cnts[i].name;
+			descs[j].flags |= IB_STAT_FLAG_OPTIONAL;
+		}
+	}
 }
 
 
 static int __mlx5_ib_alloc_counters(struct mlx5_ib_dev *dev,
 				    struct mlx5_ib_counters *cnts)
 {
-	u32 num_counters;
+	u32 num_counters, num_op_counters;
 
 	num_counters = ARRAY_SIZE(basic_q_cnts);
 

@@ -455,6 +500,19 @@ static int __mlx5_ib_alloc_counters(struct mlx5_ib_dev *dev,
 		cnts->num_ext_ppcnt_counters = ARRAY_SIZE(ext_ppcnt_cnts);
 		num_counters += ARRAY_SIZE(ext_ppcnt_cnts);
 	}
+
+	num_op_counters = ARRAY_SIZE(basic_op_cnts);
+
+	if (MLX5_CAP_FLOWTABLE(dev->mdev,
+			       ft_field_support_2_nic_receive_rdma.bth_opcode))
+		num_op_counters += ARRAY_SIZE(rdmarx_cnp_op_cnts);
+
+	if (MLX5_CAP_FLOWTABLE(dev->mdev,
+			       ft_field_support_2_nic_transmit_rdma.bth_opcode))
+		num_op_counters += ARRAY_SIZE(rdmatx_cnp_op_cnts);
+
+	cnts->num_op_counters = num_op_counters;
+	num_counters += num_op_counters;
 	cnts->descs = kcalloc(num_counters,
 			      sizeof(struct rdma_stat_desc), GFP_KERNEL);
 	if (!cnts->descs)

index 6f5451d96dd7..8215d7ab579d 100644
@@ -803,6 +803,7 @@ struct mlx5_ib_counters {
 	u32 num_q_counters;
 	u32 num_cong_counters;
 	u32 num_ext_ppcnt_counters;
+	u32 num_op_counters;
 	u16 set_id;
 };
 

index 5fbc0a8454b9..b780185d9dc6 100644
@@ -10,12 +10,14 @@
 #include <rdma/uverbs_std_types.h>
 #include <rdma/mlx5_user_ioctl_cmds.h>
 #include <rdma/mlx5_user_ioctl_verbs.h>
+#include <rdma/ib_hdrs.h>
 #include <rdma/ib_umem.h>
 #include <linux/mlx5/driver.h>
 #include <linux/mlx5/fs.h>
 #include <linux/mlx5/fs_helpers.h>
 #include <linux/mlx5/accel.h>
 #include <linux/mlx5/eswitch.h>
+#include <net/inet_ecn.h>
 #include "mlx5_ib.h"
 #include "counters.h"
 #include "devx.h"

@@ -847,6 +849,191 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev,
 	return prio;
 }
 
+enum {
+	RDMA_RX_ECN_OPCOUNTER_PRIO,
+	RDMA_RX_CNP_OPCOUNTER_PRIO,
+};
+
+enum {
+	RDMA_TX_CNP_OPCOUNTER_PRIO,
+};
+
+static int set_vhca_port_spec(struct mlx5_ib_dev *dev, u32 port_num,
+			      struct mlx5_flow_spec *spec)
+{
+	if (!MLX5_CAP_FLOWTABLE_RDMA_RX(dev->mdev,
+					ft_field_support.source_vhca_port) ||
+	    !MLX5_CAP_FLOWTABLE_RDMA_TX(dev->mdev,
+					ft_field_support.source_vhca_port))
+		return -EOPNOTSUPP;
+
+	MLX5_SET_TO_ONES(fte_match_param, &spec->match_criteria,
+			 misc_parameters.source_vhca_port);
+	MLX5_SET(fte_match_param, &spec->match_value,
+		 misc_parameters.source_vhca_port, port_num);
+
+	return 0;
+}
+
+static int set_ecn_ce_spec(struct mlx5_ib_dev *dev, u32 port_num,
+			   struct mlx5_flow_spec *spec, int ipv)
+{
+	if (!MLX5_CAP_FLOWTABLE_RDMA_RX(dev->mdev,
+					ft_field_support.outer_ip_version))
+		return -EOPNOTSUPP;
+
+	if (mlx5_core_mp_enabled(dev->mdev) &&
+	    set_vhca_port_spec(dev, port_num, spec))
+		return -EOPNOTSUPP;
+
+	MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria,
+			 outer_headers.ip_ecn);
+	MLX5_SET(fte_match_param, spec->match_value, outer_headers.ip_ecn,
+		 INET_ECN_CE);
+	MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria,
+			 outer_headers.ip_version);
+	MLX5_SET(fte_match_param, spec->match_value, outer_headers.ip_version,
+		 ipv);
+
+	spec->match_criteria_enable =
+		get_match_criteria_enable(spec->match_criteria);
+
+	return 0;
+}
+
+static int set_cnp_spec(struct mlx5_ib_dev *dev, u32 port_num,
+			struct mlx5_flow_spec *spec)
+{
+	if (mlx5_core_mp_enabled(dev->mdev) &&
+	    set_vhca_port_spec(dev, port_num, spec))
+		return -EOPNOTSUPP;
+
+	MLX5_SET_TO_ONES(fte_match_param, spec->match_criteria,
+			 misc_parameters.bth_opcode);
+	MLX5_SET(fte_match_param, spec->match_value, misc_parameters.bth_opcode,
+		 IB_BTH_OPCODE_CNP);
+
+	spec->match_criteria_enable =
+		get_match_criteria_enable(spec->match_criteria);
+
+	return 0;
+}
+
+int mlx5_ib_fs_add_op_fc(struct mlx5_ib_dev *dev, u32 port_num,
+			 struct mlx5_ib_op_fc *opfc,
+			 enum mlx5_ib_optional_counter_type type)
+{
+	enum mlx5_flow_namespace_type fn_type;
+	int priority, i, err, spec_num;
+	struct mlx5_flow_act flow_act = {};
+	struct mlx5_flow_destination dst;
+	struct mlx5_flow_namespace *ns;
+	struct mlx5_ib_flow_prio *prio;
+	struct mlx5_flow_spec *spec;
+
+	spec = kcalloc(MAX_OPFC_RULES, sizeof(*spec), GFP_KERNEL);
+	if (!spec)
+		return -ENOMEM;
+
+	switch (type) {
+	case MLX5_IB_OPCOUNTER_CC_RX_CE_PKTS:
+		if (set_ecn_ce_spec(dev, port_num, &spec[0],
+				    MLX5_FS_IPV4_VERSION) ||
+		    set_ecn_ce_spec(dev, port_num, &spec[1],
+				    MLX5_FS_IPV6_VERSION)) {
+			err = -EOPNOTSUPP;
+			goto free;
+		}
+		spec_num = 2;
+		fn_type = MLX5_FLOW_NAMESPACE_RDMA_RX_COUNTERS;
+		priority = RDMA_RX_ECN_OPCOUNTER_PRIO;
+		break;
+
+	case MLX5_IB_OPCOUNTER_CC_RX_CNP_PKTS:
+		if (!MLX5_CAP_FLOWTABLE(dev->mdev,
+					ft_field_support_2_nic_receive_rdma.bth_opcode) ||
+		    set_cnp_spec(dev, port_num, &spec[0])) {
+			err = -EOPNOTSUPP;
+			goto free;
+		}
+		spec_num = 1;
+		fn_type = MLX5_FLOW_NAMESPACE_RDMA_RX_COUNTERS;
+		priority = RDMA_RX_CNP_OPCOUNTER_PRIO;
+		break;
+
+	case MLX5_IB_OPCOUNTER_CC_TX_CNP_PKTS:
+		if (!MLX5_CAP_FLOWTABLE(dev->mdev,
+					ft_field_support_2_nic_transmit_rdma.bth_opcode) ||
+		    set_cnp_spec(dev, port_num, &spec[0])) {
+			err = -EOPNOTSUPP;
+			goto free;
+		}
+		spec_num = 1;
+		fn_type = MLX5_FLOW_NAMESPACE_RDMA_TX_COUNTERS;
+		priority = RDMA_TX_CNP_OPCOUNTER_PRIO;
+		break;
+
+	default:
+		err = -EOPNOTSUPP;
+		goto free;
+	}
+
+	ns = mlx5_get_flow_namespace(dev->mdev, fn_type);
+	if (!ns) {
+		err = -EOPNOTSUPP;
+		goto free;
+	}
+
+	prio = &dev->flow_db->opfcs[type];
+	if (!prio->flow_table) {
+		prio = _get_prio(ns, prio, priority,
+				 dev->num_ports * MAX_OPFC_RULES, 1, 0);
+		if (IS_ERR(prio)) {
+			err = PTR_ERR(prio);
+			goto free;
+		}
+	}
+
+	dst.type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
+	dst.counter_id = mlx5_fc_id(opfc->fc);
+
+	flow_act.action =
+		MLX5_FLOW_CONTEXT_ACTION_COUNT | MLX5_FLOW_CONTEXT_ACTION_ALLOW;
+
+	for (i = 0; i < spec_num; i++) {
+		opfc->rule[i] = mlx5_add_flow_rules(prio->flow_table, &spec[i],
+						    &flow_act, &dst, 1);
+		if (IS_ERR(opfc->rule[i])) {
+			err = PTR_ERR(opfc->rule[i]);
+			goto del_rules;
+		}
+	}
+	prio->refcount += spec_num;
+	kfree(spec);
+
+	return 0;
+
+del_rules:
+	for (i -= 1; i >= 0; i--)
+		mlx5_del_flow_rules(opfc->rule[i]);
+	put_flow_table(dev, prio, false);
+free:
+	kfree(spec);
+	return err;
+}
+
+void mlx5_ib_fs_remove_op_fc(struct mlx5_ib_dev *dev,
+			     struct mlx5_ib_op_fc *opfc,
+			     enum mlx5_ib_optional_counter_type type)
+{
+	int i;
+
+	for (i = 0; i < MAX_OPFC_RULES && opfc->rule[i]; i++) {
+		mlx5_del_flow_rules(opfc->rule[i]);
+		put_flow_table(dev, &dev->flow_db->opfcs[type], true);
+	}
+}
+
 static void set_underlay_qp(struct mlx5_ib_dev *dev,
 			    struct mlx5_flow_spec *spec,
 			    u32 underlay_qpn)

index 8215d7ab579d..d81ff5078e5e 100644
@@ -263,6 +263,14 @@ struct mlx5_ib_pp {
 	struct mlx5_core_dev *mdev;
 };
 
+enum mlx5_ib_optional_counter_type {
+	MLX5_IB_OPCOUNTER_CC_RX_CE_PKTS,
+	MLX5_IB_OPCOUNTER_CC_RX_CNP_PKTS,
+	MLX5_IB_OPCOUNTER_CC_TX_CNP_PKTS,
+
+	MLX5_IB_OPCOUNTER_MAX,
+};
+
 struct mlx5_ib_flow_db {
 	struct mlx5_ib_flow_prio	prios[MLX5_IB_NUM_FLOW_FT];
 	struct mlx5_ib_flow_prio	egress_prios[MLX5_IB_NUM_FLOW_FT];

@@ -271,6 +279,7 @@ struct mlx5_ib_flow_db {
 	struct mlx5_ib_flow_prio	fdb;
 	struct mlx5_ib_flow_prio	rdma_rx[MLX5_IB_NUM_FLOW_FT];
 	struct mlx5_ib_flow_prio	rdma_tx[MLX5_IB_NUM_FLOW_FT];
+	struct mlx5_ib_flow_prio	opfcs[MLX5_IB_OPCOUNTER_MAX];
 	struct mlx5_flow_table		*lag_demux_ft;
 	/* Protect flow steering bypass flow tables
 	 * when add/del flow rules.

@@ -797,6 +806,13 @@ struct mlx5_ib_resources {
 	struct mlx5_ib_port_resources ports[2];
 };
 
+#define MAX_OPFC_RULES 2
+
+struct mlx5_ib_op_fc {
+	struct mlx5_fc *fc;
+	struct mlx5_flow_handle *rule[MAX_OPFC_RULES];
+};
+
 struct mlx5_ib_counters {
 	struct rdma_stat_desc *descs;
 	size_t *offsets;

@@ -807,6 +823,14 @@ struct mlx5_ib_counters {
 	u16 set_id;
 };
 
+int mlx5_ib_fs_add_op_fc(struct mlx5_ib_dev *dev, u32 port_num,
+			 struct mlx5_ib_op_fc *opfc,
+			 enum mlx5_ib_optional_counter_type type);
+
+void mlx5_ib_fs_remove_op_fc(struct mlx5_ib_dev *dev,
+			     struct mlx5_ib_op_fc *opfc,
+			     enum mlx5_ib_optional_counter_type type);
+
 struct mlx5_ib_multiport_info;
 
 struct mlx5_ib_multiport {

index 7e542205861c..8ae07c0ecdf7 100644
@@ -232,6 +232,7 @@ static inline u32 ib_get_sqpn(struct ib_other_headers *ohdr)
 #define IB_BTH_SE_SHIFT	23
 #define IB_BTH_TVER_MASK	0xf
 #define IB_BTH_TVER_SHIFT	16
+#define IB_BTH_OPCODE_CNP	0x81
 
 static inline u8 ib_bth_get_pad(struct ib_other_headers *ohdr)
 {

index 8fe7900b29f0..6ee340c63b20 100644
@@ -12,6 +12,7 @@
 struct mlx5_ib_counter {
 	const char *name;
 	size_t offset;
+	u32 type;
 };
 
 #define INIT_Q_COUNTER(_name)		\

@@ -75,19 +76,19 @@ static const struct mlx5_ib_counter ext_ppcnt_cnts[] = {
 	INIT_EXT_PPCNT_COUNTER(rx_icrc_encapsulated),
 };
 
-#define INIT_OP_COUNTER(_name)                                          \
-	{ .name = #_name }
+#define INIT_OP_COUNTER(_name, _type)		\
+	{ .name = #_name, .type = MLX5_IB_OPCOUNTER_##_type}
 
 static const struct mlx5_ib_counter basic_op_cnts[] = {
-	INIT_OP_COUNTER(cc_rx_ce_pkts),
+	INIT_OP_COUNTER(cc_rx_ce_pkts, CC_RX_CE_PKTS),
 };
 
 static const struct mlx5_ib_counter rdmarx_cnp_op_cnts[] = {
-	INIT_OP_COUNTER(cc_rx_cnp_pkts),
+	INIT_OP_COUNTER(cc_rx_cnp_pkts, CC_RX_CNP_PKTS),
 };
 
 static const struct mlx5_ib_counter rdmatx_cnp_op_cnts[] = {
-	INIT_OP_COUNTER(cc_tx_cnp_pkts),
+	INIT_OP_COUNTER(cc_tx_cnp_pkts, CC_TX_CNP_PKTS),
 };
 
 static int mlx5_ib_read_counters(struct ib_counters *counters,

@@ -451,6 +452,7 @@ static void mlx5_ib_fill_counters(struct mlx5_ib_dev *dev,
 	for (i = 0; i < ARRAY_SIZE(basic_op_cnts); i++, j++) {
 		descs[j].name = basic_op_cnts[i].name;
 		descs[j].flags |= IB_STAT_FLAG_OPTIONAL;
+		descs[j].priv = &basic_op_cnts[i].type;
 	}
 
 	if (MLX5_CAP_FLOWTABLE(dev->mdev,

@@ -458,6 +460,7 @@ static void mlx5_ib_fill_counters(struct mlx5_ib_dev *dev,
 		for (i = 0; i < ARRAY_SIZE(rdmarx_cnp_op_cnts); i++, j++) {
 			descs[j].name = rdmarx_cnp_op_cnts[i].name;
 			descs[j].flags |= IB_STAT_FLAG_OPTIONAL;
+			descs[j].priv = &rdmarx_cnp_op_cnts[i].type;
 		}
 	}
 

@@ -466,6 +469,7 @@ static void mlx5_ib_fill_counters(struct mlx5_ib_dev *dev,
 		for (i = 0; i < ARRAY_SIZE(rdmatx_cnp_op_cnts); i++, j++) {
 			descs[j].name = rdmatx_cnp_op_cnts[i].name;
 			descs[j].flags |= IB_STAT_FLAG_OPTIONAL;
+			descs[j].priv = &rdmatx_cnp_op_cnts[i].type;
 		}
 	}
 }

@@ -535,7 +539,7 @@ static void mlx5_ib_dealloc_counters(struct mlx5_ib_dev *dev)
 {
 	u32 in[MLX5_ST_SZ_DW(dealloc_q_counter_in)] = {};
 	int num_cnt_ports;
-	int i;
+	int i, j;
 
 	num_cnt_ports = is_mdev_switchdev_mode(dev->mdev) ? 1 : dev->num_ports;
 

@@ -550,6 +554,18 @@ static void mlx5_ib_dealloc_counters(struct mlx5_ib_dev *dev)
 		}
 		kfree(dev->port[i].cnts.descs);
 		kfree(dev->port[i].cnts.offsets);
+
+		for (j = 0; j < MLX5_IB_OPCOUNTER_MAX; j++) {
+			if (!dev->port[i].cnts.opfcs[j].fc)
+				continue;
+
+			mlx5_ib_fs_remove_op_fc(dev,
+						&dev->port[i].cnts.opfcs[j],
+						j);
+			mlx5_fc_destroy(dev->mdev,
+					dev->port[i].cnts.opfcs[j].fc);
+			dev->port[i].cnts.opfcs[j].fc = NULL;
+		}
 	}
 }
 

@@ -729,6 +745,56 @@ void mlx5_ib_counters_clear_description(struct ib_counters *counters)
 	mutex_unlock(&mcounters->mcntrs_mutex);
 }
 
+static int mlx5_ib_modify_stat(struct ib_device *device, u32 port,
+			       unsigned int index, bool enable)
+{
+	struct mlx5_ib_dev *dev = to_mdev(device);
+	struct mlx5_ib_counters *cnts;
+	struct mlx5_ib_op_fc *opfc;
+	u32 num_hw_counters, type;
+	int ret;
+
+	cnts = &dev->port[port - 1].cnts;
+	num_hw_counters = cnts->num_q_counters + cnts->num_cong_counters +
+		cnts->num_ext_ppcnt_counters;
+	if (index < num_hw_counters ||
+	    index >= (num_hw_counters + cnts->num_op_counters))
+		return -EINVAL;
+
+	if (!(cnts->descs[index].flags & IB_STAT_FLAG_OPTIONAL))
+		return -EINVAL;
+
+	type = *(u32 *)cnts->descs[index].priv;
+	if (type >= MLX5_IB_OPCOUNTER_MAX)
+		return -EINVAL;
+
+	opfc = &cnts->opfcs[type];
+
+	if (enable) {
+		if (opfc->fc)
+			return -EEXIST;
+
+		opfc->fc = mlx5_fc_create(dev->mdev, false);
+		if (IS_ERR(opfc->fc))
+			return PTR_ERR(opfc->fc);
+
+		ret = mlx5_ib_fs_add_op_fc(dev, port, opfc, type);
+		if (ret) {
+			mlx5_fc_destroy(dev->mdev, opfc->fc);
+			opfc->fc = NULL;
+		}
+		return ret;
+	}
+
+	if (!opfc->fc)
+		return -EINVAL;
+
+	mlx5_ib_fs_remove_op_fc(dev, opfc, type);
+	mlx5_fc_destroy(dev->mdev, opfc->fc);
+	opfc->fc = NULL;
+	return 0;
+}
+
 static const struct ib_device_ops hw_stats_ops = {
 	.alloc_hw_port_stats = mlx5_ib_alloc_hw_port_stats,
 	.get_hw_stats = mlx5_ib_get_hw_stats,

@@ -737,6 +803,7 @@ static const struct ib_device_ops hw_stats_ops = {
 	.counter_dealloc = mlx5_ib_counter_dealloc,
 	.counter_alloc_stats = mlx5_ib_counter_alloc_stats,
 	.counter_update_stats = mlx5_ib_counter_update_stats,
+	.modify_hw_stat = mlx5_ib_modify_stat,
 };
 
 static const struct ib_device_ops hw_switchdev_stats_ops = {

index d81ff5078e5e..cf8b0653f0ce 100644
@@ -821,6 +821,7 @@ struct mlx5_ib_counters {
 	u32 num_ext_ppcnt_counters;
 	u32 num_op_counters;
 	u16 set_id;
+	struct mlx5_ib_op_fc opfcs[MLX5_IB_OPCOUNTER_MAX];
 };
 
 int mlx5_ib_fs_add_op_fc(struct mlx5_ib_dev *dev, u32 port_num,

index 2207f60b002f..7688720411a3 100644
@@ -553,10 +553,12 @@ enum ib_stat_flag {
  * struct rdma_stat_desc
  * @name - The name of the counter
  * @flags - Flags of the counter; For example, IB_STAT_FLAG_OPTIONAL
+ * @priv - Driver private information; Core code should not use
  */
 struct rdma_stat_desc {
 	const char *name;
 	unsigned int flags;
+	const void *priv;
 };
 
 /**

index 6ee340c63b20..6f1c4b57110e 100644
@@ -270,9 +270,9 @@ static int mlx5_ib_query_ext_ppcnt_counters(struct mlx5_ib_dev *dev,
 	return ret;
 }
 
-static int mlx5_ib_get_hw_stats(struct ib_device *ibdev,
-				struct rdma_hw_stats *stats,
-				u32 port_num, int index)
+static int do_get_hw_stats(struct ib_device *ibdev,
+			   struct rdma_hw_stats *stats,
+			   u32 port_num, int index)
 {
 	struct mlx5_ib_dev *dev = to_mdev(ibdev);
 	const struct mlx5_ib_counters *cnts = get_counters(dev, port_num - 1);

@@ -324,6 +324,88 @@ static int mlx5_ib_get_hw_stats(struct ib_device *ibdev,
 	return num_counters;
 }
 
+static int do_get_op_stat(struct ib_device *ibdev,
+			  struct rdma_hw_stats *stats,
+			  u32 port_num, int index)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	const struct mlx5_ib_counters *cnts;
+	const struct mlx5_ib_op_fc *opfcs;
+	u64 packets = 0, bytes;
+	u32 type;
+	int ret;
+
+	cnts = get_counters(dev, port_num - 1);
+	opfcs = cnts->opfcs;
+	type = *(u32 *)cnts->descs[index].priv;
+	if (type >= MLX5_IB_OPCOUNTER_MAX)
+		return -EINVAL;
+
+	if (!opfcs[type].fc)
+		goto out;
+
+	ret = mlx5_fc_query(dev->mdev, opfcs[type].fc,
+			    &packets, &bytes);
+	if (ret)
+		return ret;
+
+out:
+	stats->value[index] = packets;
+	return index;
+}
+
+static int do_get_op_stats(struct ib_device *ibdev,
+			   struct rdma_hw_stats *stats,
+			   u32 port_num)
+{
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	const struct mlx5_ib_counters *cnts;
+	int index, ret, num_hw_counters;
+
+	cnts = get_counters(dev, port_num - 1);
+	num_hw_counters = cnts->num_q_counters + cnts->num_cong_counters +
+			  cnts->num_ext_ppcnt_counters;
+	for (index = num_hw_counters;
+	     index < (num_hw_counters + cnts->num_op_counters); index++) {
+		ret = do_get_op_stat(ibdev, stats, port_num, index);
+		if (ret != index)
+			return ret;
+	}
+
+	return cnts->num_op_counters;
+}
+
+static int mlx5_ib_get_hw_stats(struct ib_device *ibdev,
+				struct rdma_hw_stats *stats,
+				u32 port_num, int index)
+{
+	int num_counters, num_hw_counters, num_op_counters;
+	struct mlx5_ib_dev *dev = to_mdev(ibdev);
+	const struct mlx5_ib_counters *cnts;
+
+	cnts = get_counters(dev, port_num - 1);
+	num_hw_counters = cnts->num_q_counters + cnts->num_cong_counters +
+		cnts->num_ext_ppcnt_counters;
+	num_counters = num_hw_counters + cnts->num_op_counters;
+
+	if (index < 0 || index > num_counters)
+		return -EINVAL;
+	else if (index > 0 && index < num_hw_counters)
+		return do_get_hw_stats(ibdev, stats, port_num, index);
+	else if (index >= num_hw_counters && index < num_counters)
+		return do_get_op_stat(ibdev, stats, port_num, index);
+
+	num_hw_counters = do_get_hw_stats(ibdev, stats, port_num, index);
+	if (num_hw_counters < 0)
+		return num_hw_counters;
+
+	num_op_counters = do_get_op_stats(ibdev, stats, port_num);
+	if (num_op_counters < 0)
+		return num_op_counters;
+
+	return num_hw_counters + num_op_counters;
+}
+
 static struct rdma_hw_stats *
 mlx5_ib_counter_alloc_stats(struct rdma_counter *counter)
 {