diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index c76729122143e23d1203f4249424e22beaffadcb..bbf6da389f86d9aa98e4c511f73b1932e828e875 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -6979,6 +6979,11 @@ static void bnxt_timer(unsigned long data)
 		set_bit(BNXT_PERIODIC_STATS_SP_EVENT, &bp->sp_event);
 		bnxt_queue_sp_work(bp);
 	}
+
+	if (bnxt_tc_flower_enabled(bp)) {
+		set_bit(BNXT_FLOW_STATS_SP_EVENT, &bp->sp_event);
+		bnxt_queue_sp_work(bp);
+	}
 bnxt_restart_timer:
 	mod_timer(&bp->timer, jiffies + bp->current_interval);
 }
@@ -7069,6 +7074,10 @@ static void bnxt_sp_task(struct work_struct *work)
 		bnxt_get_port_module_status(bp);
 		mutex_unlock(&bp->link_lock);
 	}
+
+	if (test_and_clear_bit(BNXT_FLOW_STATS_SP_EVENT, &bp->sp_event))
+		bnxt_tc_flow_stats_work(bp);
+
 	/* These functions below will clear BNXT_STATE_IN_SP_TASK.  They
 	 * must be the last functions to be called before exiting.
 	 */
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index d88d864db7d49e14f00388ad0c8fa40f73c3a5e0..b8343ee4182c8ed8a71b5a418f5dfb5d5f662a1e 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -955,6 +955,11 @@ struct bnxt_coal {
 	u8			budget;
 };
 
+struct bnxt_tc_flow_stats {
+	u64		packets;
+	u64		bytes;
+};
+
 struct bnxt_tc_info {
 	bool				enabled;
 
@@ -980,6 +985,14 @@ struct bnxt_tc_info {
 	 */
 	struct mutex			lock;
 
+	/* Fields used for batching stats query */
+	struct rhashtable_iter		iter;
+#define BNXT_FLOW_STATS_BATCH_MAX	10
+	struct bnxt_tc_stats_batch {
+		void			  *flow_node;
+		struct bnxt_tc_flow_stats hw_stats;
+	} stats_batch[BNXT_FLOW_STATS_BATCH_MAX];
+
 	/* Stat counter mask (width) */
 	u64				bytes_mask;
 	u64				packets_mask;
@@ -1282,6 +1295,7 @@ struct bnxt {
 #define BNXT_GENEVE_ADD_PORT_SP_EVENT	12
 #define BNXT_GENEVE_DEL_PORT_SP_EVENT	13
 #define BNXT_LINK_SPEED_CHNG_SP_EVENT	14
+#define BNXT_FLOW_STATS_SP_EVENT	15
 
 	struct bnxt_pf_info	pf;
 #ifdef CONFIG_BNXT_SRIOV
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.c
index 0d258d303eefac8697293b1cfb668fbd229e2dc5..71828a5beefe29ceb62872ea59ea370beb8a1824 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.c
@@ -504,81 +504,6 @@ static int bnxt_hwrm_cfa_flow_alloc(struct bnxt *bp, struct bnxt_tc_flow *flow,
 	return rc;
 }
 
-/* Add val to accum while handling a possible wraparound
- * of val. Eventhough val is of type u64, its actual width
- * is denoted by mask and will wrap-around beyond that width.
- */
-static void accumulate_val(u64 *accum, u64 val, u64 mask)
-{
-#define low_bits(x, mask)		((x) & (mask))
-#define high_bits(x, mask)		((x) & ~(mask))
-	bool wrapped = val < low_bits(*accum, mask);
-
-	*accum = high_bits(*accum, mask) + val;
-	if (wrapped)
-		*accum += (mask + 1);
-}
-
-/* The HW counters' width is much less than 64bits.
- * Handle possible wrap-around while updating the stat counters
- */
-static void bnxt_flow_stats_fix_wraparound(struct bnxt_tc_info *tc_info,
-					   struct bnxt_tc_flow_stats *stats,
-					   struct bnxt_tc_flow_stats *hw_stats)
-{
-	accumulate_val(&stats->bytes, hw_stats->bytes, tc_info->bytes_mask);
-	accumulate_val(&stats->packets, hw_stats->packets,
-		       tc_info->packets_mask);
-}
-
-/* Fix possible wraparound of the stats queried from HW, calculate
- * the delta from prev_stats, and also update the prev_stats.
- * The HW flow stats are fetched under the hwrm_cmd_lock mutex.
- * This routine is best called while under the mutex so that the
- * stats processing happens atomically.
- */
-static void bnxt_flow_stats_calc(struct bnxt_tc_info *tc_info,
-				 struct bnxt_tc_flow *flow,
-				 struct bnxt_tc_flow_stats *stats)
-{
-	struct bnxt_tc_flow_stats *acc_stats, *prev_stats;
-
-	acc_stats = &flow->stats;
-	bnxt_flow_stats_fix_wraparound(tc_info, acc_stats, stats);
-
-	prev_stats = &flow->prev_stats;
-	stats->bytes = acc_stats->bytes - prev_stats->bytes;
-	stats->packets = acc_stats->packets - prev_stats->packets;
-	*prev_stats = *acc_stats;
-}
-
-static int bnxt_hwrm_cfa_flow_stats_get(struct bnxt *bp,
-					__le16 flow_handle,
-					struct bnxt_tc_flow *flow,
-					struct bnxt_tc_flow_stats *stats)
-{
-	struct hwrm_cfa_flow_stats_output *resp = bp->hwrm_cmd_resp_addr;
-	struct hwrm_cfa_flow_stats_input req = { 0 };
-	int rc;
-
-	bnxt_hwrm_cmd_hdr_init(bp, &req, HWRM_CFA_FLOW_STATS, -1, -1);
-	req.num_flows = cpu_to_le16(1);
-	req.flow_handle_0 = flow_handle;
-
-	mutex_lock(&bp->hwrm_cmd_lock);
-	rc = _hwrm_send_message(bp, &req, sizeof(req), HWRM_CMD_TIMEOUT);
-	if (!rc) {
-		stats->packets = le64_to_cpu(resp->packet_0);
-		stats->bytes = le64_to_cpu(resp->byte_0);
-		bnxt_flow_stats_calc(&bp->tc_info, flow, stats);
-	} else {
-		netdev_info(bp->dev, "error rc=%d", rc);
-	}
-
-	mutex_unlock(&bp->hwrm_cmd_lock);
-	return rc;
-}
-
 static int hwrm_cfa_decap_filter_alloc(struct bnxt *bp,
 				       struct bnxt_tc_flow *flow,
 				       struct bnxt_tc_l2_key *l2_info,
@@ -1306,6 +1231,8 @@ static int bnxt_tc_add_flow(struct bnxt *bp, u16 src_fid,
 	if (rc)
 		goto put_tunnel;
 
+	flow->lastused = jiffies;
+	spin_lock_init(&flow->stats_lock);
 	/* add new flow to flow-table */
 	rc = rhashtable_insert_fast(&tc_info->flow_table, &new_node->node,
 				    tc_info->flow_ht_params);
@@ -1352,10 +1279,11 @@ static int bnxt_tc_del_flow(struct bnxt *bp,
 static int bnxt_tc_get_flow_stats(struct bnxt *bp,
 				  struct tc_cls_flower_offload *tc_flow_cmd)
 {
+	struct bnxt_tc_flow_stats stats, *curr_stats, *prev_stats;
 	struct bnxt_tc_info *tc_info = &bp->tc_info;
 	struct bnxt_tc_flow_node *flow_node;
-	struct bnxt_tc_flow_stats stats;
-	int rc;
+	struct bnxt_tc_flow *flow;
+	unsigned long lastused;
 
 	flow_node = rhashtable_lookup_fast(&tc_info->flow_table,
 					   &tc_flow_cmd->cookie,
@@ -1366,15 +1294,183 @@ static int bnxt_tc_get_flow_stats(struct bnxt *bp,
 		return -1;
 	}
 
-	rc = bnxt_hwrm_cfa_flow_stats_get(bp, flow_node->flow_handle,
-					  &flow_node->flow, &stats);
+	flow = &flow_node->flow;
+	curr_stats = &flow->stats;
+	prev_stats = &flow->prev_stats;
+
+	spin_lock(&flow->stats_lock);
+	stats.packets = curr_stats->packets - prev_stats->packets;
+	stats.bytes = curr_stats->bytes - prev_stats->bytes;
+	*prev_stats = *curr_stats;
+	lastused = flow->lastused;
+	spin_unlock(&flow->stats_lock);
+
+	tcf_exts_stats_update(tc_flow_cmd->exts, stats.bytes, stats.packets,
+			      lastused);
+	return 0;
+}
+
+static int
+bnxt_hwrm_cfa_flow_stats_get(struct bnxt *bp, int num_flows,
+			     struct bnxt_tc_stats_batch stats_batch[])
+{
+	struct hwrm_cfa_flow_stats_output *resp = bp->hwrm_cmd_resp_addr;
+	struct hwrm_cfa_flow_stats_input req = { 0 };
+	__le16 *req_flow_handles = &req.flow_handle_0;
+	int rc, i;
+
+	bnxt_hwrm_cmd_hdr_init(bp, &req, HWRM_CFA_FLOW_STATS, -1, -1);
+	req.num_flows = cpu_to_le16(num_flows);
+	for (i = 0; i < num_flows; i++) {
+		struct bnxt_tc_flow_node *flow_node = stats_batch[i].flow_node;
+
+		req_flow_handles[i] = flow_node->flow_handle;
+	}
+
+	mutex_lock(&bp->hwrm_cmd_lock);
+	rc = _hwrm_send_message(bp, &req, sizeof(req), HWRM_CMD_TIMEOUT);
+	if (!rc) {
+		__le64 *resp_packets = &resp->packet_0;
+		__le64 *resp_bytes = &resp->byte_0;
+
+		for (i = 0; i < num_flows; i++) {
+			stats_batch[i].hw_stats.packets =
+						le64_to_cpu(resp_packets[i]);
+			stats_batch[i].hw_stats.bytes =
+						le64_to_cpu(resp_bytes[i]);
+		}
+	} else {
+		netdev_info(bp->dev, "error rc=%d", rc);
+	}
+
+	mutex_unlock(&bp->hwrm_cmd_lock);
+	return rc;
+}
+
+/* Add val to accum while handling a possible wraparound
+ * of val. Eventhough val is of type u64, its actual width
+ * is denoted by mask and will wrap-around beyond that width.
+ */
+static void accumulate_val(u64 *accum, u64 val, u64 mask)
+{
+#define low_bits(x, mask)		((x) & (mask))
+#define high_bits(x, mask)		((x) & ~(mask))
+	bool wrapped = val < low_bits(*accum, mask);
+
+	*accum = high_bits(*accum, mask) + val;
+	if (wrapped)
+		*accum += (mask + 1);
+}
+
+/* The HW counters' width is much less than 64bits.
+ * Handle possible wrap-around while updating the stat counters
+ */
+static void bnxt_flow_stats_accum(struct bnxt_tc_info *tc_info,
+				  struct bnxt_tc_flow_stats *acc_stats,
+				  struct bnxt_tc_flow_stats *hw_stats)
+{
+	accumulate_val(&acc_stats->bytes, hw_stats->bytes, tc_info->bytes_mask);
+	accumulate_val(&acc_stats->packets, hw_stats->packets,
+		       tc_info->packets_mask);
+}
+
+static int
+bnxt_tc_flow_stats_batch_update(struct bnxt *bp, int num_flows,
+				struct bnxt_tc_stats_batch stats_batch[])
+{
+	struct bnxt_tc_info *tc_info = &bp->tc_info;
+	int rc, i;
+
+	rc = bnxt_hwrm_cfa_flow_stats_get(bp, num_flows, stats_batch);
 	if (rc)
 		return rc;
 
-	tcf_exts_stats_update(tc_flow_cmd->exts, stats.bytes, stats.packets, 0);
+	for (i = 0; i < num_flows; i++) {
+		struct bnxt_tc_flow_node *flow_node = stats_batch[i].flow_node;
+		struct bnxt_tc_flow *flow = &flow_node->flow;
+
+		spin_lock(&flow->stats_lock);
+		bnxt_flow_stats_accum(tc_info, &flow->stats,
+				      &stats_batch[i].hw_stats);
+		if (flow->stats.packets != flow->prev_stats.packets)
+			flow->lastused = jiffies;
+		spin_unlock(&flow->stats_lock);
+	}
+
 	return 0;
 }
 
+static int
+bnxt_tc_flow_stats_batch_prep(struct bnxt *bp,
+			      struct bnxt_tc_stats_batch stats_batch[],
+			      int *num_flows)
+{
+	struct bnxt_tc_info *tc_info = &bp->tc_info;
+	struct rhashtable_iter *iter = &tc_info->iter;
+	void *flow_node;
+	int rc, i;
+
+	rc = rhashtable_walk_start(iter);
+	if (rc && rc != -EAGAIN) {
+		i = 0;
+		goto done;
+	}
+
+	rc = 0;
+	for (i = 0; i < BNXT_FLOW_STATS_BATCH_MAX; i++) {
+		flow_node = rhashtable_walk_next(iter);
+		if (IS_ERR(flow_node)) {
+			i = 0;
+			if (PTR_ERR(flow_node) == -EAGAIN) {
+				continue;
+			} else {
+				rc = PTR_ERR(flow_node);
+				goto done;
+			}
+		}
+
+		/* No more flows */
+		if (!flow_node)
+			goto done;
+
+		stats_batch[i].flow_node = flow_node;
+	}
+done:
+	rhashtable_walk_stop(iter);
+	*num_flows = i;
+	return rc;
+}
+
+void bnxt_tc_flow_stats_work(struct bnxt *bp)
+{
+	struct bnxt_tc_info *tc_info = &bp->tc_info;
+	int num_flows, rc;
+
+	num_flows = atomic_read(&tc_info->flow_table.nelems);
+	if (!num_flows)
+		return;
+
+	rhashtable_walk_enter(&tc_info->flow_table, &tc_info->iter);
+
+	for (;;) {
+		rc = bnxt_tc_flow_stats_batch_prep(bp, tc_info->stats_batch,
+						   &num_flows);
+		if (rc) {
+			if (rc == -EAGAIN)
+				continue;
+			break;
+		}
+
+		if (!num_flows)
+			break;
+
+		bnxt_tc_flow_stats_batch_update(bp, num_flows,
+						tc_info->stats_batch);
+	}
+
+	rhashtable_walk_exit(&tc_info->iter);
+}
+
 int bnxt_tc_setup_flower(struct bnxt *bp, u16 src_fid,
 			 struct tc_cls_flower_offload *cls_flower)
 {
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.h b/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.h
index 2beccd41c886740f0378b9917d09f719a150502a..404f98dea0ab454cc7d59800324ad703da672942 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_tc.h
@@ -78,11 +78,6 @@ struct bnxt_tc_actions {
 	struct ip_tunnel_key		tun_encap_key;
 };
 
-struct bnxt_tc_flow_stats {
-	u64		packets;
-	u64		bytes;
-};
-
 struct bnxt_tc_flow {
 	u32				flags;
 #define BNXT_TC_FLOW_FLAGS_ETH_ADDRS		BIT(1)
@@ -119,6 +114,10 @@ struct bnxt_tc_flow {
 	/* previous snap-shot of stats */
 	struct bnxt_tc_flow_stats	prev_stats;
 	unsigned long			lastused; /* jiffies */
+	/* for calculating delta from prev_stats and
+	 * updating prev_stats atomically.
+	 */
+	spinlock_t			stats_lock;
 };
 
 /* Tunnel encap/decap hash table
@@ -195,6 +194,12 @@ int bnxt_tc_setup_flower(struct bnxt *bp, u16 src_fid,
 			 struct tc_cls_flower_offload *cls_flower);
 int bnxt_init_tc(struct bnxt *bp);
 void bnxt_shutdown_tc(struct bnxt *bp);
+void bnxt_tc_flow_stats_work(struct bnxt *bp);
+
+static inline bool bnxt_tc_flower_enabled(struct bnxt *bp)
+{
+	return bp->tc_info.enabled;
+}
 
 #else /* CONFIG_BNXT_FLOWER_OFFLOAD */
 
@@ -212,5 +217,14 @@ static inline int bnxt_init_tc(struct bnxt *bp)
 static inline void bnxt_shutdown_tc(struct bnxt *bp)
 {
 }
+
+static inline void bnxt_tc_flow_stats_work(struct bnxt *bp)
+{
+}
+
+static inline bool bnxt_tc_flower_enabled(struct bnxt *bp)
+{
+	return false;
+}
 #endif /* CONFIG_BNXT_FLOWER_OFFLOAD */
 #endif /* BNXT_TC_H */