wlan-ap-Telecominfraproject/feeds/wifi-ax/mac80211/patches/qca/237-007-ath11k-optimizations-in-ath11k_dp_process_rx.patch

From 0fe4a7f57a44a0aaad6fb1d6861c9fb9597358a2 Mon Sep 17 00:00:00 2001
From: P Praneesh <ppranees@codeaurora.org>
Date: Tue, 15 Dec 2020 11:27:48 +0530
Subject: [PATCH] ath11k: optimizations in ath11k_dp_process_rx()

Incaludes below changes,
	1. Increment total_msdu_reaped only when continuation is not set and update
	drop flag correspondingly. This is to avoid additional access to ath11k_hal_srng_dst_num_free()
	2. Avoid descriptor memcpy as the data is prefetched already
	3. The Active Pdev and CAC check are done for each msdu in ath11k_dp_rx_process_received_packets
	Rather, collect all msdu's in a per mac msdu list, and pass to the function.
	4. Add branch predictors where required in the process rx()
	5. Remove usage of quota variable since we add only max napi budget msdus to the list
	6. Prefetch the initial descriptor part which is used by all the hw ops for msdu info collection.

	About 3-4% improvement on single radio UDP UL on single core was seen.
	More validation/confirmation on perf stats required.

Signed-off-by: Sriram R <srirrama@codeaurora.org>
Signed-off-by: P Praneesh <ppranees@codeaurora.org>
---
 drivers/net/wireless/ath/ath11k/dp_rx.c | 107 ++++++++++++++++----------------
 1 file changed, 55 insertions(+), 52 deletions(-)

--- a/drivers/net/wireless/ath/ath11k/dp_rx.c
+++ b/drivers/net/wireless/ath/ath11k/dp_rx.c
@@ -2856,6 +2856,9 @@ static int ath11k_dp_rx_process_msdu(str
 	int ret;
 	u32 hal_rx_desc_sz = ar->ab->hw_params.hal_desc_sz;
 
+	/* Prefetch the full descriptor range before hdr status into cache */
+	prefetch_range(msdu->data, 256);
+
 	last_buf = ath11k_dp_rx_get_msdu_last_buf(msdu_list, msdu);
 	if (!last_buf) {
 		ath11k_warn(ab,
@@ -2920,10 +2923,8 @@ static int ath11k_dp_rx_process_msdu(str
 	}
 
 	ath11k_dp_rx_h_mpdu(ar, msdu, rx_desc, rx_status, fast_rx);
-	if (*fast_rx) {
-		ab->soc_stats.invalid_rbm++;
+	if (*fast_rx)
 		return 0;
-	}
 
 	ath11k_dp_rx_h_ppdu(ar, rx_desc, rx_status);
 
@@ -2938,13 +2939,12 @@ free_out:
 static void ath11k_dp_rx_process_received_packets(struct ath11k_base *ab,
 						  struct napi_struct *napi,
 						  struct sk_buff_head *msdu_list,
-						  int *quota, int ring_id)
+						  int mac_id)
 {
 	struct ath11k_skb_rxcb *rxcb;
 	struct sk_buff *msdu;
 	struct ath11k *ar;
 	struct ieee80211_rx_status rx_status = {0};
-	u8 mac_id;
 	int ret;
 	bool fast_rx;
 
@@ -2953,26 +2953,28 @@ static void ath11k_dp_rx_process_receive
 
 	rcu_read_lock();
 
-	while (*quota && (msdu = __skb_dequeue(msdu_list))) {
-		rxcb = ATH11K_SKB_RXCB(msdu);
-		mac_id = rxcb->mac_id;
-		ar = ab->pdevs[mac_id].ar;
-		if (!rcu_dereference(ab->pdevs_active[mac_id])) {
-			dev_kfree_skb_any(msdu);
-			continue;
-		}
 
-		if (test_bit(ATH11K_CAC_RUNNING, &ar->dev_flags)) {
-			dev_kfree_skb_any(msdu);
-			continue;
-		}
+	ar = ab->pdevs[mac_id].ar;
+	if (unlikely(!rcu_dereference(ab->pdevs_active[mac_id]))) {
+		__skb_queue_purge(msdu_list);
+		rcu_read_unlock();
+		return;
+	}
 
+	if (unlikely(test_bit(ATH11K_CAC_RUNNING, &ar->dev_flags))) {
+		__skb_queue_purge(msdu_list);
+		rcu_read_unlock();
+		return;
+	}
+
+	while ((msdu = __skb_dequeue(msdu_list))) {
+		rxcb = ATH11K_SKB_RXCB(msdu);
 		/* Enable fast rx by default, the value will cahnge based on peer cap
 		 * and packet type */
 		fast_rx = true;
 		rxcb->napi = napi;
 		ret = ath11k_dp_rx_process_msdu(ar, msdu, msdu_list, &rx_status, &fast_rx);
-		if (ret) {
+		if (unlikely(ret)) {
 			ath11k_dbg(ab, ATH11K_DBG_DATA,
 				   "Unable to process msdu %d", ret);
 			dev_kfree_skb_any(msdu);
@@ -2982,8 +2984,6 @@ static void ath11k_dp_rx_process_receive
 		/* msdu is already delivered directectly */
 		if (!fast_rx)
 			ath11k_dp_rx_deliver_msdu(ar, napi, msdu, &rx_status);
-
-		(*quota)--;
 	}
 
 	rcu_read_unlock();
@@ -2995,22 +2995,24 @@ int ath11k_dp_process_rx(struct ath11k_b
 	struct ath11k_dp *dp = &ab->dp;
 	struct dp_rxdma_ring *rx_ring;
 	int num_buffs_reaped[MAX_RADIOS] = {0};
-	struct sk_buff_head msdu_list;
+	struct sk_buff_head msdu_list[MAX_RADIOS];
 	struct ath11k_skb_rxcb *rxcb;
 	int total_msdu_reaped = 0;
 	struct hal_srng *srng;
 	struct sk_buff *msdu;
 	struct ath11k_peer *peer = NULL;
 	struct ath11k_sta *arsta = NULL;
-	int quota = budget;
 	bool done = false;
 	int buf_id, mac_id;
 	struct ath11k *ar;
-	u32 *rx_desc;
+	struct hal_reo_dest_ring *desc;
+	enum hal_reo_dest_ring_push_reason push_reason;
+	u32 cookie;
 	int i;
 
-	__skb_queue_head_init(&msdu_list);
 
+	for (i = 0; i < MAX_RADIOS; i++)
+		__skb_queue_head_init(&msdu_list[i]);
 	srng = &ab->hal.srng_list[dp->reo_dst_ring[ring_id].ring_id];
 
 	spin_lock_bh(&srng->lock);
@@ -3018,13 +3020,9 @@ int ath11k_dp_process_rx(struct ath11k_b
 	ath11k_hal_srng_access_begin(ab, srng);
 
 try_again:
-	while ((rx_desc = ath11k_hal_srng_dst_get_next_entry(ab, srng))) {
-		struct hal_reo_dest_ring desc = *(struct hal_reo_dest_ring *)rx_desc;
-		enum hal_reo_dest_ring_push_reason push_reason;
-		u32 cookie;
-
+	while (likely(desc = (struct hal_reo_dest_ring *)ath11k_hal_srng_dst_get_next_entry(ab, srng))) {
 		cookie = FIELD_GET(BUFFER_ADDR_INFO1_SW_COOKIE,
-				   desc.buf_addr_info.info1);
+				   desc->buf_addr_info.info1);
 		buf_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_BUF_ID,
 				   cookie);
 		mac_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_PDEV_ID, cookie);
@@ -3033,7 +3031,7 @@ try_again:
 		rx_ring = &ar->dp.rx_refill_buf_ring;
 		spin_lock_bh(&rx_ring->idr_lock);
 		msdu = idr_find(&rx_ring->bufs_idr, buf_id);
-		if (!msdu) {
+		if (unlikely(!msdu)) {
 			ath11k_warn(ab, "frame rx with invalid buf_id %d\n",
 				    buf_id);
 			spin_unlock_bh(&rx_ring->idr_lock);
@@ -3049,29 +3047,28 @@ try_again:
 				 DMA_FROM_DEVICE);
 
 		num_buffs_reaped[mac_id]++;
-		total_msdu_reaped++;
 
 		push_reason = FIELD_GET(HAL_REO_DEST_RING_INFO0_PUSH_REASON,
-					desc.info0);
-		if (push_reason !=
-		    HAL_REO_DEST_RING_PUSH_REASON_ROUTING_INSTRUCTION) {
+					desc->info0);
+		if (unlikely(push_reason !=
+		    HAL_REO_DEST_RING_PUSH_REASON_ROUTING_INSTRUCTION)) {
 			dev_kfree_skb_any(msdu);
 			ab->soc_stats.hal_reo_error[dp->reo_dst_ring[ring_id].ring_id]++;
 			continue;
 		}
 
-		rxcb->is_first_msdu = !!(desc.rx_msdu_info.info0 &
+		rxcb->is_first_msdu = !!(desc->rx_msdu_info.info0 &
 					 RX_MSDU_DESC_INFO0_FIRST_MSDU_IN_MPDU);
-		rxcb->is_last_msdu = !!(desc.rx_msdu_info.info0 &
+		rxcb->is_last_msdu = !!(desc->rx_msdu_info.info0 &
 					RX_MSDU_DESC_INFO0_LAST_MSDU_IN_MPDU);
-		rxcb->is_continuation = !!(desc.rx_msdu_info.info0 &
+		rxcb->is_continuation = !!(desc->rx_msdu_info.info0 &
 					   RX_MSDU_DESC_INFO0_MSDU_CONTINUATION);
 		rxcb->peer_id = FIELD_GET(RX_MPDU_DESC_META_DATA_PEER_ID,
-					  desc.rx_mpdu_info.meta_data);
+					  desc->rx_mpdu_info.meta_data);
 		rxcb->seq_no = FIELD_GET(RX_MPDU_DESC_INFO0_SEQ_NUM,
-					 desc.rx_mpdu_info.info0);
+					 desc->rx_mpdu_info.info0);
 		rxcb->tid = FIELD_GET(HAL_REO_DEST_RING_INFO0_RX_QUEUE_NUM,
-				      desc.info0);
+				      desc->info0);
 
 		if (ath11k_debugfs_is_extd_rx_stats_enabled(ar) && rxcb->peer_id) {
 			rcu_read_lock();
@@ -3087,12 +3084,17 @@ try_again:
 		}
 
 		rxcb->mac_id = mac_id;
-		__skb_queue_tail(&msdu_list, msdu);
+		__skb_queue_tail(&msdu_list[mac_id], msdu);
 
-		if (total_msdu_reaped >= quota && !rxcb->is_continuation) {
+		if (!rxcb->is_continuation) {
+			total_msdu_reaped++;
 			done = true;
-			break;
+		} else {
+			done = false;
 		}
+
+		if (total_msdu_reaped >= budget)
+			break;
 	}
 
 	/* Hw might have updated the head pointer after we cached it.
@@ -3101,7 +3103,7 @@ try_again:
 	 * head pointer so that we can reap complete MPDU in the current
 	 * rx processing.
 	 */
-	if (!done && ath11k_hal_srng_dst_num_valid(ab, srng, true)) {
+	if (unlikely(!done && ath11k_hal_srng_dst_num_valid(ab, srng, true))) {
 		ath11k_hal_srng_access_end(ab, srng);
 		goto try_again;
 	}
@@ -3110,13 +3112,14 @@ try_again:
 
 	spin_unlock_bh(&srng->lock);
 
-	if (!total_msdu_reaped)
+	if (unlikely(!total_msdu_reaped))
 		goto exit;
 
 	for (i = 0; i < ab->num_radios; i++) {
 		if (!num_buffs_reaped[i])
 			continue;
 
+		ath11k_dp_rx_process_received_packets(ab, napi, &msdu_list[i], i);
 		ar = ab->pdevs[i].ar;
 		rx_ring = &ar->dp.rx_refill_buf_ring;
 
@@ -3124,11 +3127,8 @@ try_again:
 					   HAL_RX_BUF_RBM_SW3_BM);
 	}
 
-	ath11k_dp_rx_process_received_packets(ab, napi, &msdu_list,
-					      &quota, ring_id);
-
 exit:
-	return budget - quota;
+	return total_msdu_reaped;
 }
 
 static void