[2/3] wifi: ath12k: Optimize the lock contention of used list in Rx data path

Commit Message

Karthikeyan Periyasamy Feb. 26, 2024, 4:23 p.m. UTC

When a packet arrives in Rx rings, the RX descriptor moves from the used
list to the free list. Then, the rxdma ring gets replenished, where the Rx
descriptor again moves from the free list to the used list. At the end, the
descriptor came to the used list with unnecessary list movement. The
descriptor used list is maintained in the Rxdma ring structure, which
creates lock contention for the list operations (add, delete) in the Rx
data path. Optimize the Rx data path by removing the used list from the
common Rxdma ring and maintain as a local variable in the Rx ring handler
itself, which avoid lock contention. Now, to find the used list descriptor
during descriptor cleanup, we need to check the in_use flag for each Rx
descriptor.

Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1

Signed-off-by: Karthikeyan Periyasamy <[e-mail]>
---
 drivers/net/wireless/ath/ath12k/dp.c    |  31 +++---
 drivers/net/wireless/ath/ath12k/dp.h    |   5 +-
 drivers/net/wireless/ath/ath12k/dp_rx.c | 131 ++++++++++++++++--------
 drivers/net/wireless/ath/ath12k/dp_rx.h |   1 +
 4 files changed, 114 insertions(+), 54 deletions(-)

Comments

Jeff Johnson Feb. 26, 2024, 6:44 p.m. UTC | #1

On 2/26/2024 8:23 AM, Karthikeyan Periyasamy wrote:
> When a packet arrives in Rx rings, the RX descriptor moves from the used
> list to the free list. Then, the rxdma ring gets replenished, where the Rx
> descriptor again moves from the free list to the used list. At the end, the
> descriptor came to the used list with unnecessary list movement. The
> descriptor used list is maintained in the Rxdma ring structure, which
> creates lock contention for the list operations (add, delete) in the Rx
> data path. Optimize the Rx data path by removing the used list from the
> common Rxdma ring and maintain as a local variable in the Rx ring handler
> itself, which avoid lock contention. Now, to find the used list descriptor
> during descriptor cleanup, we need to check the in_use flag for each Rx
> descriptor.
> 
> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
> 
> Signed-off-by: Karthikeyan Periyasamy <[e-mail]>
Acked-by: Jeff Johnson <[e-mail]>

Kalle Valo March 11, 2024, 1:05 p.m. UTC | #2

Karthikeyan Periyasamy <[e-mail]> writes:

> When a packet arrives in Rx rings, the RX descriptor moves from the used
> list to the free list. Then, the rxdma ring gets replenished, where the Rx
> descriptor again moves from the free list to the used list. At the end, the
> descriptor came to the used list with unnecessary list movement. The
> descriptor used list is maintained in the Rxdma ring structure, which
> creates lock contention for the list operations (add, delete) in the Rx
> data path. Optimize the Rx data path by removing the used list from the
> common Rxdma ring and maintain as a local variable in the Rx ring handler
> itself, which avoid lock contention. Now, to find the used list descriptor
> during descriptor cleanup, we need to check the in_use flag for each Rx
> descriptor.
>
> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
>
> Signed-off-by: Karthikeyan Periyasamy <[e-mail]>

Before looking at this in detail: When optimising something it would be
good to have before and after numbers showing the improvement. Otherwise
it's just wishful thinking.

Karthikeyan Periyasamy March 11, 2024, 1:55 p.m. UTC | #3

On 3/11/2024 6:35 PM, Kalle Valo wrote:
> Karthikeyan Periyasamy <[e-mail]> writes:
>
>> When a packet arrives in Rx rings, the RX descriptor moves from the used
>> list to the free list. Then, the rxdma ring gets replenished, where the Rx
>> descriptor again moves from the free list to the used list. At the end, the
>> descriptor came to the used list with unnecessary list movement. The
>> descriptor used list is maintained in the Rxdma ring structure, which
>> creates lock contention for the list operations (add, delete) in the Rx
>> data path. Optimize the Rx data path by removing the used list from the
>> common Rxdma ring and maintain as a local variable in the Rx ring handler
>> itself, which avoid lock contention. Now, to find the used list descriptor
>> during descriptor cleanup, we need to check the in_use flag for each Rx
>> descriptor.
>>
>> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
>>
>> Signed-off-by: Karthikeyan Periyasamy <[e-mail]>
> Before looking at this in detail: When optimising something it would be
> good to have before and after numbers showing the improvement. Otherwise
> it's just wishful thinking.

I don't have numbers. Like you said, Its just a wishful think.

Kalle Valo March 18, 2024, 6:09 p.m. UTC | #4

Karthikeyan Periyasamy <[e-mail]> writes:

> On 3/11/2024 6:35 PM, Kalle Valo wrote:
>> Karthikeyan Periyasamy <[e-mail]> writes:
>>
>>> When a packet arrives in Rx rings, the RX descriptor moves from the used
>>> list to the free list. Then, the rxdma ring gets replenished, where the Rx
>>> descriptor again moves from the free list to the used list. At the end, the
>>> descriptor came to the used list with unnecessary list movement. The
>>> descriptor used list is maintained in the Rxdma ring structure, which
>>> creates lock contention for the list operations (add, delete) in the Rx
>>> data path. Optimize the Rx data path by removing the used list from the
>>> common Rxdma ring and maintain as a local variable in the Rx ring handler
>>> itself, which avoid lock contention. Now, to find the used list descriptor
>>> during descriptor cleanup, we need to check the in_use flag for each Rx
>>> descriptor.
>>>
>>> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
>>>
>>> Signed-off-by: Karthikeyan Periyasamy <[e-mail]>
>> Before looking at this in detail: When optimising something it would be
>> good to have before and after numbers showing the improvement. Otherwise
>> it's just wishful thinking.
>
> I don't have numbers. Like you said, Its just a wishful think.

So do you still want us to take this?

In the future please do provide numbers to show that the optimisation
really helps as intended. Otherwise we might even go backwards.

Karthikeyan Periyasamy March 19, 2024, 9:37 a.m. UTC | #5

On 3/18/2024 11:39 PM, Kalle Valo wrote:
> Karthikeyan Periyasamy <[e-mail]> writes:
>
>> On 3/11/2024 6:35 PM, Kalle Valo wrote:
>>> Karthikeyan Periyasamy <[e-mail]> writes:
>>>
>>>> When a packet arrives in Rx rings, the RX descriptor moves from the used
>>>> list to the free list. Then, the rxdma ring gets replenished, where the Rx
>>>> descriptor again moves from the free list to the used list. At the end, the
>>>> descriptor came to the used list with unnecessary list movement. The
>>>> descriptor used list is maintained in the Rxdma ring structure, which
>>>> creates lock contention for the list operations (add, delete) in the Rx
>>>> data path. Optimize the Rx data path by removing the used list from the
>>>> common Rxdma ring and maintain as a local variable in the Rx ring handler
>>>> itself, which avoid lock contention. Now, to find the used list descriptor
>>>> during descriptor cleanup, we need to check the in_use flag for each Rx
>>>> descriptor.
>>>>
>>>> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
>>>>
>>>> Signed-off-by: Karthikeyan Periyasamy <[e-mail]>
>>> Before looking at this in detail: When optimising something it would be
>>> good to have before and after numbers showing the improvement. Otherwise
>>> it's just wishful thinking.
>> I don't have numbers. Like you said, Its just a wishful think.
> So do you still want us to take this?
>
> In the future please do provide numbers to show that the optimisation
> really helps as intended. Otherwise we might even go backwards.

This is a simple UDP UL throughput test case results on x86+NUC device
with QCN9274 card

Before:
Average:     CPU    %usr   %nice    %sys %iowait    %irq   %soft %steal  
%guest  %gnice   %idle
Average:     all    0.24    0.00   12.54    0.08    0.00 23.33    
0.00    0.00    0.00   63.81

After:
Average:     CPU    %usr   %nice    %sys %iowait    %irq   %soft %steal  
%guest  %gnice   %idle
Average:     all    0.34    0.00    4.60    0.00    0.00 19.59    
0.00    0.00    0.00   75.47

is it fine to capture in the commit log ?

Kalle Valo March 19, 2024, 4:44 p.m. UTC | #6

Karthikeyan Periyasamy <[e-mail]> writes:

> On 3/18/2024 11:39 PM, Kalle Valo wrote:
>
>> Karthikeyan Periyasamy <[e-mail]> writes:
>>
>>> On 3/11/2024 6:35 PM, Kalle Valo wrote:
>>>> Karthikeyan Periyasamy <[e-mail]> writes:
>>>>
>>>>> When a packet arrives in Rx rings, the RX descriptor moves from the used
>>>>> list to the free list. Then, the rxdma ring gets replenished, where the Rx
>>>>> descriptor again moves from the free list to the used list. At the end, the
>>>>> descriptor came to the used list with unnecessary list movement. The
>>>>> descriptor used list is maintained in the Rxdma ring structure, which
>>>>> creates lock contention for the list operations (add, delete) in the Rx
>>>>> data path. Optimize the Rx data path by removing the used list from the
>>>>> common Rxdma ring and maintain as a local variable in the Rx ring handler
>>>>> itself, which avoid lock contention. Now, to find the used list descriptor
>>>>> during descriptor cleanup, we need to check the in_use flag for each Rx
>>>>> descriptor.
>>>>>
>>>>> Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
>>>>>
>>>>> Signed-off-by: Karthikeyan Periyasamy <[e-mail]>
>>>> Before looking at this in detail: When optimising something it would be
>>>> good to have before and after numbers showing the improvement. Otherwise
>>>> it's just wishful thinking.
>>> I don't have numbers. Like you said, Its just a wishful think.
>> So do you still want us to take this?
>>
>> In the future please do provide numbers to show that the optimisation
>> really helps as intended. Otherwise we might even go backwards.
>
> This is a simple UDP UL throughput test case results on x86+NUC device
> with QCN9274 card
>
> Before:
> Average:     CPU    %usr   %nice    %sys %iowait    %irq   %soft
> %steal  %guest  %gnice   %idle
> Average:     all    0.24    0.00   12.54    0.08    0.00 23.33   
> 0.00    0.00    0.00   63.81
>
> After:
> Average:     CPU    %usr   %nice    %sys %iowait    %irq   %soft
> %steal  %guest  %gnice   %idle
> Average:     all    0.34    0.00    4.60    0.00    0.00 19.59   
> 0.00    0.00    0.00   75.47
>
> is it fine to capture in the commit log ?

Thanks, exactly something like that helps to show that the optimisation
is helping. Please also include your analysis of the results in the
commit message.

Patch

diff --git a/drivers/net/wireless/ath/ath12k/dp.c b/drivers/net/wireless/ath/ath12k/dp.c
index c8e1b244b69e..1006eef8ff0c 100644
--- a/drivers/net/wireless/ath/ath12k/dp.c
+++ b/drivers/net/wireless/ath/ath12k/dp.c
@@ -960,8 +960,9 @@  int ath12k_dp_service_srng(struct ath12k_base *ab,
 	if (ab->hw_params->ring_mask->host2rxdma[grp_id]) {
 		struct ath12k_dp *dp = &ab->dp;
 		struct dp_rxdma_ring *rx_ring = &dp->rx_refill_buf_ring;
+		LIST_HEAD(list);
 
-		ath12k_dp_rx_bufs_replenish(ab, rx_ring, 0);
+		ath12k_dp_rx_bufs_replenish(ab, rx_ring, &list, 0);
 	}
 
 	/* TODO: Implement handler for other interrupts */
@@ -1146,11 +1147,11 @@  void ath12k_dp_vdev_tx_attach(struct ath12k *ar, struct ath12k_vif *arvif)
 
 static void ath12k_dp_cc_cleanup(struct ath12k_base *ab)
 {
-	struct ath12k_rx_desc_info *desc_info, *tmp;
+	struct ath12k_rx_desc_info *desc_info;
 	struct ath12k_tx_desc_info *tx_desc_info, *tmp1;
 	struct ath12k_dp *dp = &ab->dp;
 	struct sk_buff *skb;
-	int i;
+	int i, j;
 	u32 pool_id, tx_spt_page;
 
 	if (!dp->spt_info)
@@ -1159,16 +1160,23 @@  static void ath12k_dp_cc_cleanup(struct ath12k_base *ab)
 	/* RX Descriptor cleanup */
 	spin_lock_bh(&dp->rx_desc_lock);
 
-	list_for_each_entry_safe(desc_info, tmp, &dp->rx_desc_used_list, list) {
-		list_del(&desc_info->list);
-		skb = desc_info->skb;
+	for (i = 0; i < ATH12K_NUM_RX_SPT_PAGES; i++) {
+		desc_info = dp->spt_info->rxbaddr[i];
 
-		if (!skb)
-			continue;
+		for (j = 0; j < ATH12K_MAX_SPT_ENTRIES; j++) {
+			if (!desc_info[j].in_use) {
+				list_del(&desc_info[j].list);
+				continue;
+			}
 
-		dma_unmap_single(ab->dev, ATH12K_SKB_RXCB(skb)->paddr,
-				 skb->len + skb_tailroom(skb), DMA_FROM_DEVICE);
-		dev_kfree_skb_any(skb);
+			skb = desc_info[j].skb;
+			if (!skb)
+				continue;
+
+			dma_unmap_single(ab->dev, ATH12K_SKB_RXCB(skb)->paddr,
+					 skb->len + skb_tailroom(skb), DMA_FROM_DEVICE);
+			dev_kfree_skb_any(skb);
+		}
 	}
 
 	for (i = 0; i < ATH12K_NUM_RX_SPT_PAGES; i++) {
@@ -1444,7 +1452,6 @@  static int ath12k_dp_cc_init(struct ath12k_base *ab)
 	u32 cmem_base;
 
 	INIT_LIST_HEAD(&dp->rx_desc_free_list);
-	INIT_LIST_HEAD(&dp->rx_desc_used_list);
 	spin_lock_init(&dp->rx_desc_lock);
 
 	for (i = 0; i < ATH12K_HW_MAX_QUEUES; i++) {
diff --git a/drivers/net/wireless/ath/ath12k/dp.h b/drivers/net/wireless/ath/ath12k/dp.h
index eb2dd408e081..0d59910a00d9 100644
--- a/drivers/net/wireless/ath/ath12k/dp.h
+++ b/drivers/net/wireless/ath/ath12k/dp.h
@@ -282,6 +282,8 @@  struct ath12k_rx_desc_info {
 	struct sk_buff *skb;
 	u32 cookie;
 	u32 magic;
+	u8 in_use	: 1,
+	   reserved	: 7;
 };
 
 struct ath12k_tx_desc_info {
@@ -347,8 +349,7 @@  struct ath12k_dp {
 	struct ath12k_spt_info *spt_info;
 	u32 num_spt_pages;
 	struct list_head rx_desc_free_list;
-	struct list_head rx_desc_used_list;
-	/* protects the free and used desc list */
+	/* protects the free desc list */
 	spinlock_t rx_desc_lock;
 
 	struct list_head tx_desc_free_list[ATH12K_HW_MAX_QUEUES];
diff --git a/drivers/net/wireless/ath/ath12k/dp_rx.c b/drivers/net/wireless/ath/ath12k/dp_rx.c
index 45646d987139..2a0e4faadcf1 100644
--- a/drivers/net/wireless/ath/ath12k/dp_rx.c
+++ b/drivers/net/wireless/ath/ath12k/dp_rx.c
@@ -261,9 +261,53 @@  static int ath12k_dp_purge_mon_ring(struct ath12k_base *ab)
 	return -ETIMEDOUT;
 }
 
+static size_t ath12k_dp_list_cut_nodes(struct list_head *list,
+				       struct list_head *head,
+				       size_t count)
+{
+	struct list_head *cur;
+	struct ath12k_rx_desc_info *rx_desc;
+	size_t nodes = 0;
+
+	if (!count) {
+		INIT_LIST_HEAD(list);
+		goto out;
+	}
+
+	list_for_each(cur, head) {
+		if (!count)
+			break;
+
+		rx_desc = list_entry(cur, struct ath12k_rx_desc_info, list);
+		rx_desc->in_use = true;
+
+		count--;
+		nodes++;
+	}
+
+	list_cut_before(list, head, cur);
+out:
+	return nodes;
+}
+
+static void ath12k_dp_rx_enqueue_free(struct ath12k_dp *dp,
+				      struct list_head *used_list)
+{
+	struct ath12k_rx_desc_info *rx_desc, *safe;
+
+	/* Reset the use flag */
+	list_for_each_entry_safe(rx_desc, safe, used_list, list)
+		rx_desc->in_use = false;
+
+	spin_lock_bh(&dp->rx_desc_lock);
+	list_splice_tail(used_list, &dp->rx_desc_free_list);
+	spin_unlock_bh(&dp->rx_desc_lock);
+}
+
 /* Returns number of Rx buffers replenished */
 int ath12k_dp_rx_bufs_replenish(struct ath12k_base *ab,
 				struct dp_rxdma_ring *rx_ring,
+				struct list_head *used_list,
 				int req_entries)
 {
 	struct ath12k_buffer_addr *desc;
@@ -292,6 +336,19 @@  int ath12k_dp_rx_bufs_replenish(struct ath12k_base *ab,
 	req_entries = min(num_free, req_entries);
 	num_remain = req_entries;
 
+	if (!num_remain)
+		goto skip_replenish;
+
+	/* Get the descriptor from free list */
+	if (list_empty(used_list)) {
+		spin_lock_bh(&dp->rx_desc_lock);
+		req_entries = ath12k_dp_list_cut_nodes(used_list,
+						       &dp->rx_desc_free_list,
+						       num_remain);
+		spin_unlock_bh(&dp->rx_desc_lock);
+		num_remain = req_entries;
+	}
+
 	while (num_remain > 0) {
 		skb = dev_alloc_skb(DP_RX_BUFFER_SIZE +
 				    DP_RX_BUFFER_ALIGN_SIZE);
@@ -311,33 +368,20 @@  int ath12k_dp_rx_bufs_replenish(struct ath12k_base *ab,
 		if (dma_mapping_error(ab->dev, paddr))
 			goto fail_free_skb;
 
-		spin_lock_bh(&dp->rx_desc_lock);
-
-		/* Get desc from free list and store in used list
-		 * for cleanup purposes
-		 *
-		 * TODO: pass the removed descs rather than
-		 * add/read to optimize
-		 */
-		rx_desc = list_first_entry_or_null(&dp->rx_desc_free_list,
+		rx_desc = list_first_entry_or_null(used_list,
 						   struct ath12k_rx_desc_info,
 						   list);
-		if (!rx_desc) {
-			spin_unlock_bh(&dp->rx_desc_lock);
+		if (!rx_desc)
 			goto fail_dma_unmap;
-		}
 
 		rx_desc->skb = skb;
 		cookie = rx_desc->cookie;
-		list_del(&rx_desc->list);
-		list_add_tail(&rx_desc->list, &dp->rx_desc_used_list);
-
-		spin_unlock_bh(&dp->rx_desc_lock);
 
 		desc = ath12k_hal_srng_src_get_next_entry(ab, srng);
 		if (!desc)
-			goto fail_buf_unassign;
+			goto fail_dma_unmap;
 
+		list_del(&rx_desc->list);
 		ATH12K_SKB_RXCB(skb)->paddr = paddr;
 
 		num_remain--;
@@ -345,18 +389,16 @@  int ath12k_dp_rx_bufs_replenish(struct ath12k_base *ab,
 		ath12k_hal_rx_buf_addr_info_set(desc, paddr, cookie, mgr);
 	}
 
+skip_replenish:
 	ath12k_hal_srng_access_end(ab, srng);
 
+	if (!list_empty(used_list))
+		ath12k_dp_rx_enqueue_free(dp, used_list);
+
 	spin_unlock_bh(&srng->lock);
 
 	return req_entries - num_remain;
 
-fail_buf_unassign:
-	spin_lock_bh(&dp->rx_desc_lock);
-	list_del(&rx_desc->list);
-	list_add_tail(&rx_desc->list, &dp->rx_desc_free_list);
-	rx_desc->skb = NULL;
-	spin_unlock_bh(&dp->rx_desc_lock);
 fail_dma_unmap:
 	dma_unmap_single(ab->dev, paddr, skb->len + skb_tailroom(skb),
 			 DMA_FROM_DEVICE);
@@ -365,6 +407,9 @@  int ath12k_dp_rx_bufs_replenish(struct ath12k_base *ab,
 
 	ath12k_hal_srng_access_end(ab, srng);
 
+	if (!list_empty(used_list))
+		ath12k_dp_rx_enqueue_free(dp, used_list);
+
 	spin_unlock_bh(&srng->lock);
 
 	return req_entries - num_remain;
@@ -422,10 +467,12 @@  static int ath12k_dp_rxdma_mon_ring_buf_setup(struct ath12k_base *ab,
 static int ath12k_dp_rxdma_ring_buf_setup(struct ath12k_base *ab,
 					  struct dp_rxdma_ring *rx_ring)
 {
+	LIST_HEAD(list);
+
 	rx_ring->bufs_max = rx_ring->refill_buf_ring.size /
 			ath12k_hal_srng_get_entrysize(ab, HAL_RXDMA_BUF);
 
-	ath12k_dp_rx_bufs_replenish(ab, rx_ring, 0);
+	ath12k_dp_rx_bufs_replenish(ab, rx_ring, &list, 0);
 
 	return 0;
 }
@@ -2582,6 +2629,7 @@  static void ath12k_dp_rx_process_received_packets(struct ath12k_base *ab,
 int ath12k_dp_rx_process(struct ath12k_base *ab, int ring_id,
 			 struct napi_struct *napi, int budget)
 {
+	LIST_HEAD(rx_desc_used_list);
 	struct ath12k_rx_desc_info *desc_info;
 	struct ath12k_dp *dp = &ab->dp;
 	struct dp_rxdma_ring *rx_ring = &dp->rx_refill_buf_ring;
@@ -2634,9 +2682,7 @@  int ath12k_dp_rx_process(struct ath12k_base *ab, int ring_id,
 		msdu = desc_info->skb;
 		desc_info->skb = NULL;
 
-		spin_lock_bh(&dp->rx_desc_lock);
-		list_move_tail(&desc_info->list, &dp->rx_desc_free_list);
-		spin_unlock_bh(&dp->rx_desc_lock);
+		list_add_tail(&desc_info->list, &rx_desc_used_list);
 
 		rxcb = ATH12K_SKB_RXCB(msdu);
 		dma_unmap_single(ab->dev, rxcb->paddr,
@@ -2697,7 +2743,8 @@  int ath12k_dp_rx_process(struct ath12k_base *ab, int ring_id,
 	if (!total_msdu_reaped)
 		goto exit;
 
-	ath12k_dp_rx_bufs_replenish(ab, rx_ring, num_buffs_reaped);
+	ath12k_dp_rx_bufs_replenish(ab, rx_ring, &rx_desc_used_list,
+				    num_buffs_reaped);
 
 	ath12k_dp_rx_process_received_packets(ab, napi, &msdu_list,
 					      ring_id);
@@ -3018,9 +3065,9 @@  static int ath12k_dp_rx_h_defrag_reo_reinject(struct ath12k *ar,
 	}
 
 	desc_info->skb = defrag_skb;
+	desc_info->in_use = true;
 
 	list_del(&desc_info->list);
-	list_add_tail(&desc_info->list, &dp->rx_desc_used_list);
 	spin_unlock_bh(&dp->rx_desc_lock);
 
 	ATH12K_SKB_RXCB(defrag_skb)->paddr = buf_paddr;
@@ -3082,9 +3129,9 @@  static int ath12k_dp_rx_h_defrag_reo_reinject(struct ath12k *ar,
 
 err_free_desc:
 	spin_lock_bh(&dp->rx_desc_lock);
-	list_del(&desc_info->list);
-	list_add_tail(&desc_info->list, &dp->rx_desc_free_list);
+	desc_info->in_use = false;
 	desc_info->skb = NULL;
+	list_add_tail(&desc_info->list, &dp->rx_desc_free_list);
 	spin_unlock_bh(&dp->rx_desc_lock);
 err_unmap_dma:
 	dma_unmap_single(ab->dev, buf_paddr, defrag_skb->len + skb_tailroom(defrag_skb),
@@ -3301,6 +3348,7 @@  static int ath12k_dp_rx_frag_h_mpdu(struct ath12k *ar,
 
 static int
 ath12k_dp_process_rx_err_buf(struct ath12k *ar, struct hal_reo_dest_ring *desc,
+			     struct list_head *used_list,
 			     bool drop, u32 cookie)
 {
 	struct ath12k_base *ab = ar->ab;
@@ -3330,9 +3378,8 @@  ath12k_dp_process_rx_err_buf(struct ath12k *ar, struct hal_reo_dest_ring *desc,
 
 	msdu = desc_info->skb;
 	desc_info->skb = NULL;
-	spin_lock_bh(&ab->dp.rx_desc_lock);
-	list_move_tail(&desc_info->list, &ab->dp.rx_desc_free_list);
-	spin_unlock_bh(&ab->dp.rx_desc_lock);
+
+	list_add_tail(&desc_info->list, used_list);
 
 	rxcb = ATH12K_SKB_RXCB(msdu);
 	dma_unmap_single(ar->ab->dev, rxcb->paddr,
@@ -3388,6 +3435,7 @@  int ath12k_dp_rx_process_err(struct ath12k_base *ab, struct napi_struct *napi,
 	struct hal_reo_dest_ring *reo_desc;
 	struct dp_rxdma_ring *rx_ring;
 	struct dp_srng *reo_except;
+	LIST_HEAD(rx_desc_used_list);
 	u32 desc_bank, num_msdus;
 	struct hal_srng *srng;
 	struct ath12k_dp *dp;
@@ -3455,7 +3503,9 @@  int ath12k_dp_rx_process_err(struct ath12k_base *ab, struct napi_struct *napi,
 			pdev_id = ath12k_hw_mac_id_to_pdev_id(ab->hw_params, mac_id);
 			ar = ab->pdevs[pdev_id].ar;
 
-			if (!ath12k_dp_process_rx_err_buf(ar, reo_desc, drop,
+			if (!ath12k_dp_process_rx_err_buf(ar, reo_desc,
+							  &rx_desc_used_list,
+							  drop,
 							  msdu_cookies[i]))
 				tot_n_bufs_reaped++;
 		}
@@ -3475,7 +3525,8 @@  int ath12k_dp_rx_process_err(struct ath12k_base *ab, struct napi_struct *napi,
 
 	rx_ring = &dp->rx_refill_buf_ring;
 
-	ath12k_dp_rx_bufs_replenish(ab, rx_ring, tot_n_bufs_reaped);
+	ath12k_dp_rx_bufs_replenish(ab, rx_ring, &rx_desc_used_list,
+				    tot_n_bufs_reaped);
 
 	return tot_n_bufs_reaped;
 }
@@ -3692,6 +3743,7 @@  static void ath12k_dp_rx_wbm_err(struct ath12k *ar,
 int ath12k_dp_rx_process_wbm_err(struct ath12k_base *ab,
 				 struct napi_struct *napi, int budget)
 {
+	LIST_HEAD(rx_desc_used_list);
 	struct ath12k *ar;
 	struct ath12k_dp *dp = &ab->dp;
 	struct dp_rxdma_ring *rx_ring;
@@ -3745,9 +3797,7 @@  int ath12k_dp_rx_process_wbm_err(struct ath12k_base *ab,
 		msdu = desc_info->skb;
 		desc_info->skb = NULL;
 
-		spin_lock_bh(&dp->rx_desc_lock);
-		list_move_tail(&desc_info->list, &dp->rx_desc_free_list);
-		spin_unlock_bh(&dp->rx_desc_lock);
+		list_add_tail(&desc_info->list, &rx_desc_used_list);
 
 		rxcb = ATH12K_SKB_RXCB(msdu);
 		dma_unmap_single(ab->dev, rxcb->paddr,
@@ -3783,7 +3833,8 @@  int ath12k_dp_rx_process_wbm_err(struct ath12k_base *ab,
 	if (!num_buffs_reaped)
 		goto done;
 
-	ath12k_dp_rx_bufs_replenish(ab, rx_ring, num_buffs_reaped);
+	ath12k_dp_rx_bufs_replenish(ab, rx_ring, &rx_desc_used_list,
+				    num_buffs_reaped);
 
 	rcu_read_lock();
 	while ((msdu = __skb_dequeue(&msdu_list))) {
diff --git a/drivers/net/wireless/ath/ath12k/dp_rx.h b/drivers/net/wireless/ath/ath12k/dp_rx.h
index 05b3d5581dbe..25940061ead5 100644
--- a/drivers/net/wireless/ath/ath12k/dp_rx.h
+++ b/drivers/net/wireless/ath/ath12k/dp_rx.h
@@ -118,6 +118,7 @@  int ath12k_dp_rx_process(struct ath12k_base *ab, int mac_id,
 			 int budget);
 int ath12k_dp_rx_bufs_replenish(struct ath12k_base *ab,
 				struct dp_rxdma_ring *rx_ring,
+				struct list_head *used_list,
 				int req_entries);
 int ath12k_dp_rx_pdev_mon_attach(struct ath12k *ar);
 int ath12k_dp_rx_peer_frag_setup(struct ath12k *ar, const u8 *peer_mac, int vdev_id);