[PATCH v2 17/49] ath11k: add dp_rx.c

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



(patches split for easier review, see the cover letter for more)

Signed-off-by: Kalle Valo <kvalo@xxxxxxxxxxxxxx>
---
 drivers/net/wireless/ath/ath11k/dp_rx.c | 4143 +++++++++++++++++++++++++++++++
 1 file changed, 4143 insertions(+)

diff --git a/drivers/net/wireless/ath/ath11k/dp_rx.c b/drivers/net/wireless/ath/ath11k/dp_rx.c
new file mode 100644
index 000000000000..9491a479fab5
--- /dev/null
+++ b/drivers/net/wireless/ath/ath11k/dp_rx.c
@@ -0,0 +1,4143 @@
+// SPDX-License-Identifier: BSD-3-Clause-Clear
+/*
+ * Copyright (c) 2018-2019 The Linux Foundation. All rights reserved.
+ */
+
+#include <linux/ieee80211.h>
+#include "core.h"
+#include "debug.h"
+#include "hal_desc.h"
+#include "hw.h"
+#include "dp_rx.h"
+#include "hal_rx.h"
+#include "dp_tx.h"
+#include "peer.h"
+
+static u8 *ath11k_dp_rx_h_80211_hdr(struct hal_rx_desc *desc)
+{
+	return desc->hdr_status;
+}
+
+static enum hal_encrypt_type ath11k_dp_rx_h_mpdu_start_enctype(struct hal_rx_desc *desc)
+{
+	if (!(__le32_to_cpu(desc->mpdu_start.info1) &
+	    RX_MPDU_START_INFO1_ENCRYPT_INFO_VALID))
+		return HAL_ENCRYPT_TYPE_OPEN;
+
+	return FIELD_GET(RX_MPDU_START_INFO2_ENC_TYPE,
+			 __le32_to_cpu(desc->mpdu_start.info2));
+}
+
+static u8 ath11k_dp_rx_h_mpdu_start_decap_type(struct hal_rx_desc *desc)
+{
+	return FIELD_GET(RX_MPDU_START_INFO5_DECAP_TYPE,
+			 __le32_to_cpu(desc->mpdu_start.info5));
+}
+
+static bool ath11k_dp_rx_h_attn_msdu_done(struct hal_rx_desc *desc)
+{
+	return !!FIELD_GET(RX_ATTENTION_INFO2_MSDU_DONE,
+			   __le32_to_cpu(desc->attention.info2));
+}
+
+static bool ath11k_dp_rx_h_attn_first_mpdu(struct hal_rx_desc *desc)
+{
+	return !!FIELD_GET(RX_ATTENTION_INFO1_FIRST_MPDU,
+			   __le32_to_cpu(desc->attention.info1));
+}
+
+static bool ath11k_dp_rx_h_attn_l4_cksum_fail(struct hal_rx_desc *desc)
+{
+	return !!FIELD_GET(RX_ATTENTION_INFO1_TCP_UDP_CKSUM_FAIL,
+			   __le32_to_cpu(desc->attention.info1));
+}
+
+static bool ath11k_dp_rx_h_attn_ip_cksum_fail(struct hal_rx_desc *desc)
+{
+	return !!FIELD_GET(RX_ATTENTION_INFO1_IP_CKSUM_FAIL,
+			   __le32_to_cpu(desc->attention.info1));
+}
+
+static bool ath11k_dp_rx_h_attn_is_decrypted(struct hal_rx_desc *desc)
+{
+	return (FIELD_GET(RX_ATTENTION_INFO2_DCRYPT_STATUS_CODE,
+			  __le32_to_cpu(desc->attention.info2)) ==
+		RX_DESC_DECRYPT_STATUS_CODE_OK);
+}
+
+static u32 ath11k_dp_rx_h_attn_mpdu_err(struct hal_rx_desc *desc)
+{
+	u32 info = __le32_to_cpu(desc->attention.info1);
+	u32 errmap = 0;
+
+	if (info & RX_ATTENTION_INFO1_FCS_ERR)
+		errmap |= DP_RX_MPDU_ERR_FCS;
+
+	if (info & RX_ATTENTION_INFO1_DECRYPT_ERR)
+		errmap |= DP_RX_MPDU_ERR_DECRYPT;
+
+	if (info & RX_ATTENTION_INFO1_TKIP_MIC_ERR)
+		errmap |= DP_RX_MPDU_ERR_TKIP_MIC;
+
+	if (info & RX_ATTENTION_INFO1_A_MSDU_ERROR)
+		errmap |= DP_RX_MPDU_ERR_AMSDU_ERR;
+
+	if (info & RX_ATTENTION_INFO1_OVERFLOW_ERR)
+		errmap |= DP_RX_MPDU_ERR_OVERFLOW;
+
+	if (info & RX_ATTENTION_INFO1_MSDU_LEN_ERR)
+		errmap |= DP_RX_MPDU_ERR_MSDU_LEN;
+
+	if (info & RX_ATTENTION_INFO1_MPDU_LEN_ERR)
+		errmap |= DP_RX_MPDU_ERR_MPDU_LEN;
+
+	return errmap;
+}
+
+static u16 ath11k_dp_rx_h_msdu_start_msdu_len(struct hal_rx_desc *desc)
+{
+	return FIELD_GET(RX_MSDU_START_INFO1_MSDU_LENGTH,
+			 __le32_to_cpu(desc->msdu_start.info1));
+}
+
+static u8 ath11k_dp_rx_h_msdu_start_sgi(struct hal_rx_desc *desc)
+{
+	return FIELD_GET(RX_MSDU_START_INFO3_SGI,
+			 __le32_to_cpu(desc->msdu_start.info3));
+}
+
+static u8 ath11k_dp_rx_h_msdu_start_rate_mcs(struct hal_rx_desc *desc)
+{
+	return FIELD_GET(RX_MSDU_START_INFO3_RATE_MCS,
+			 __le32_to_cpu(desc->msdu_start.info3));
+}
+
+static u8 ath11k_dp_rx_h_msdu_start_rx_bw(struct hal_rx_desc *desc)
+{
+	return FIELD_GET(RX_MSDU_START_INFO3_RECV_BW,
+			 __le32_to_cpu(desc->msdu_start.info3));
+}
+
+static u32 ath11k_dp_rx_h_msdu_start_freq(struct hal_rx_desc *desc)
+{
+	return __le32_to_cpu(desc->msdu_start.phy_meta_data);
+}
+
+static u8 ath11k_dp_rx_h_msdu_start_pkt_type(struct hal_rx_desc *desc)
+{
+	return FIELD_GET(RX_MSDU_START_INFO3_PKT_TYPE,
+			 __le32_to_cpu(desc->msdu_start.info3));
+}
+
+static u8 ath11k_dp_rx_h_msdu_start_nss(struct hal_rx_desc *desc)
+{
+	u8 mimo_ss_bitmap = FIELD_GET(RX_MSDU_START_INFO3_MIMO_SS_BITMAP,
+				      __le32_to_cpu(desc->msdu_start.info3));
+
+	return hweight8(mimo_ss_bitmap);
+}
+
+static u8 ath11k_dp_rx_h_msdu_end_l3pad(struct hal_rx_desc *desc)
+{
+	return FIELD_GET(RX_MSDU_END_INFO2_L3_HDR_PADDING,
+			 __le32_to_cpu(desc->msdu_end.info2));
+}
+
+static bool ath11k_dp_rx_h_msdu_end_first_msdu(struct hal_rx_desc *desc)
+{
+	return !!FIELD_GET(RX_MSDU_END_INFO2_FIRST_MSDU,
+			   __le32_to_cpu(desc->msdu_end.info2));
+}
+
+static bool ath11k_dp_rx_h_msdu_end_last_msdu(struct hal_rx_desc *desc)
+{
+	return !!FIELD_GET(RX_MSDU_END_INFO2_LAST_MSDU,
+			   __le32_to_cpu(desc->msdu_end.info2));
+}
+
+static void ath11k_dp_rx_desc_end_tlv_copy(struct hal_rx_desc *fdesc,
+					   struct hal_rx_desc *ldesc)
+{
+	memcpy((u8 *)&fdesc->msdu_end, (u8 *)&ldesc->msdu_end,
+	       sizeof(struct rx_msdu_end));
+	memcpy((u8 *)&fdesc->attention, (u8 *)&ldesc->attention,
+	       sizeof(struct rx_attention));
+	memcpy((u8 *)&fdesc->mpdu_end, (u8 *)&ldesc->mpdu_end,
+	       sizeof(struct rx_mpdu_end));
+}
+
+static u32 ath11k_dp_rxdesc_get_mpdulen_err(struct hal_rx_desc *rx_desc)
+{
+	struct rx_attention *rx_attn;
+
+	rx_attn = &rx_desc->attention;
+
+	return FIELD_GET(RX_ATTENTION_INFO1_MPDU_LEN_ERR,
+			 __le32_to_cpu(rx_attn->info1));
+}
+
+static u32 ath11k_dp_rxdesc_get_decap_format(struct hal_rx_desc *rx_desc)
+{
+	struct rx_msdu_start *rx_msdu_start;
+
+	rx_msdu_start = &rx_desc->msdu_start;
+
+	return FIELD_GET(RX_MSDU_START_INFO2_DECAP_FORMAT,
+			 __le32_to_cpu(rx_msdu_start->info2));
+}
+
+static u8 *ath11k_dp_rxdesc_get_80211hdr(struct hal_rx_desc *rx_desc)
+{
+	u8 *rx_pkt_hdr;
+
+	rx_pkt_hdr = &rx_desc->msdu_payload[0];
+
+	return rx_pkt_hdr;
+}
+
+static bool ath11k_dp_rxdesc_mpdu_valid(struct hal_rx_desc *rx_desc)
+{
+	u32 tlv_tag;
+
+	tlv_tag = FIELD_GET(HAL_TLV_HDR_TAG,
+			    __le32_to_cpu(rx_desc->mpdu_start_tag));
+
+	return tlv_tag == HAL_RX_MPDU_START ? true : false;
+}
+
+static u32 ath11k_dp_rxdesc_get_ppduid(struct hal_rx_desc *rx_desc)
+{
+	return __le16_to_cpu(rx_desc->mpdu_start.phy_ppdu_id);
+}
+
+/* Returns number of Rx buffers replenished */
+int ath11k_dp_rxbufs_replenish(struct ath11k_base *ab, int mac_id,
+			       struct dp_rxdma_ring *rx_ring,
+			       int req_entries,
+			       enum hal_rx_buf_return_buf_manager mgr,
+			       gfp_t gfp)
+{
+	struct hal_srng *srng;
+	u32 *desc;
+	struct sk_buff *skb;
+	int num_free;
+	int num_remain;
+	int buf_id;
+	u32 cookie;
+	dma_addr_t paddr;
+
+	req_entries = min(req_entries, rx_ring->bufs_max);
+
+	srng = &ab->hal.srng_list[rx_ring->refill_buf_ring.ring_id];
+
+	spin_lock_bh(&srng->lock);
+
+	ath11k_hal_srng_access_begin(ab, srng);
+
+	num_free = ath11k_hal_srng_src_num_free(ab, srng, true);
+	if (!req_entries && (num_free > (rx_ring->bufs_max * 3) / 4))
+		req_entries = num_free;
+
+	req_entries = min(num_free, req_entries);
+	num_remain = req_entries;
+
+	while (num_remain > 0) {
+		skb = dev_alloc_skb(DP_RX_BUFFER_SIZE +
+				    DP_RX_BUFFER_ALIGN_SIZE);
+		if (!skb)
+			break;
+
+		if (!IS_ALIGNED((unsigned long)skb->data,
+				DP_RX_BUFFER_ALIGN_SIZE)) {
+			skb_pull(skb,
+				 PTR_ALIGN(skb->data, DP_RX_BUFFER_ALIGN_SIZE) -
+				 skb->data);
+		}
+
+		paddr = dma_map_single(ab->dev, skb->data,
+				       skb->len + skb_tailroom(skb),
+				       DMA_FROM_DEVICE);
+		if (dma_mapping_error(ab->dev, paddr))
+			goto fail_free_skb;
+
+		spin_lock_bh(&rx_ring->idr_lock);
+		buf_id = idr_alloc(&rx_ring->bufs_idr, skb, 0,
+				   rx_ring->bufs_max * 3, gfp);
+		spin_unlock_bh(&rx_ring->idr_lock);
+		if (buf_id < 0)
+			goto fail_dma_unmap;
+
+		desc = ath11k_hal_srng_src_get_next_entry(ab, srng);
+		if (!desc)
+			goto fail_idr_remove;
+
+		ATH11K_SKB_RXCB(skb)->paddr = paddr;
+
+		cookie = FIELD_PREP(DP_RXDMA_BUF_COOKIE_PDEV_ID, mac_id) |
+			 FIELD_PREP(DP_RXDMA_BUF_COOKIE_BUF_ID, buf_id);
+
+		num_remain--;
+
+		ath11k_hal_rx_buf_addr_info_set(desc, paddr, cookie, mgr);
+	}
+
+	ath11k_hal_srng_access_end(ab, srng);
+
+	spin_unlock_bh(&srng->lock);
+
+	return req_entries - num_remain;
+
+fail_idr_remove:
+	spin_lock_bh(&rx_ring->idr_lock);
+	idr_remove(&rx_ring->bufs_idr, buf_id);
+	spin_unlock_bh(&rx_ring->idr_lock);
+fail_dma_unmap:
+	dma_unmap_single(ab->dev, paddr, skb->len + skb_tailroom(skb),
+			 DMA_FROM_DEVICE);
+fail_free_skb:
+	dev_kfree_skb_any(skb);
+
+	ath11k_hal_srng_access_end(ab, srng);
+
+	spin_unlock_bh(&srng->lock);
+
+	return req_entries - num_remain;
+}
+
+static int ath11k_dp_rxdma_buf_ring_free(struct ath11k *ar,
+					 struct dp_rxdma_ring *rx_ring)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct sk_buff *skb;
+	int buf_id;
+
+	spin_lock_bh(&rx_ring->idr_lock);
+	idr_for_each_entry(&rx_ring->bufs_idr, skb, buf_id) {
+		idr_remove(&rx_ring->bufs_idr, buf_id);
+		/* TODO: Understand where internal driver does this dma_unmap of
+		 * of rxdma_buffer.
+		 */
+		dma_unmap_single(ar->ab->dev, ATH11K_SKB_RXCB(skb)->paddr,
+				 skb->len + skb_tailroom(skb), DMA_FROM_DEVICE);
+		dev_kfree_skb_any(skb);
+	}
+
+	idr_destroy(&rx_ring->bufs_idr);
+	spin_unlock_bh(&rx_ring->idr_lock);
+
+	rx_ring = &dp->rx_mon_status_refill_ring;
+
+	spin_lock_bh(&rx_ring->idr_lock);
+	idr_for_each_entry(&rx_ring->bufs_idr, skb, buf_id) {
+		idr_remove(&rx_ring->bufs_idr, buf_id);
+		/* XXX: Understand where internal driver does this dma_unmap of
+		 * of rxdma_buffer.
+		 */
+		dma_unmap_single(ar->ab->dev, ATH11K_SKB_RXCB(skb)->paddr,
+				 skb->len + skb_tailroom(skb), DMA_BIDIRECTIONAL);
+		dev_kfree_skb_any(skb);
+	}
+
+	idr_destroy(&rx_ring->bufs_idr);
+	spin_unlock_bh(&rx_ring->idr_lock);
+	return 0;
+}
+
+static int ath11k_dp_rxdma_pdev_buf_free(struct ath11k *ar)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct dp_rxdma_ring *rx_ring = &dp->rx_refill_buf_ring;
+
+	ath11k_dp_rxdma_buf_ring_free(ar, rx_ring);
+
+	rx_ring = &dp->rxdma_mon_buf_ring;
+	ath11k_dp_rxdma_buf_ring_free(ar, rx_ring);
+
+	rx_ring = &dp->rx_mon_status_refill_ring;
+	ath11k_dp_rxdma_buf_ring_free(ar, rx_ring);
+	return 0;
+}
+
+static int ath11k_dp_rxdma_ring_buf_setup(struct ath11k *ar,
+					  struct dp_rxdma_ring *rx_ring,
+					  u32 ringtype)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	int num_entries;
+
+	num_entries = rx_ring->refill_buf_ring.size /
+		      ath11k_hal_srng_get_entrysize(ringtype);
+
+	rx_ring->bufs_max = num_entries;
+	ath11k_dp_rxbufs_replenish(ar->ab, dp->mac_id, rx_ring, num_entries,
+				   HAL_RX_BUF_RBM_SW3_BM, GFP_KERNEL);
+	return 0;
+}
+
+static int ath11k_dp_rxdma_pdev_buf_setup(struct ath11k *ar)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct dp_rxdma_ring *rx_ring = &dp->rx_refill_buf_ring;
+
+	ath11k_dp_rxdma_ring_buf_setup(ar, rx_ring, HAL_RXDMA_BUF);
+
+	rx_ring = &dp->rxdma_mon_buf_ring;
+	ath11k_dp_rxdma_ring_buf_setup(ar, rx_ring, HAL_RXDMA_MONITOR_BUF);
+
+	rx_ring = &dp->rx_mon_status_refill_ring;
+	ath11k_dp_rxdma_ring_buf_setup(ar, rx_ring, HAL_RXDMA_MONITOR_STATUS);
+
+	return 0;
+}
+
+static void ath11k_dp_rx_pdev_srng_free(struct ath11k *ar)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+
+	ath11k_dp_srng_cleanup(ar->ab, &dp->rx_refill_buf_ring.refill_buf_ring);
+	ath11k_dp_srng_cleanup(ar->ab, &dp->reo_dst_ring);
+	ath11k_dp_srng_cleanup(ar->ab, &dp->rxdma_err_dst_ring);
+	ath11k_dp_srng_cleanup(ar->ab, &dp->rx_mon_status_refill_ring.refill_buf_ring);
+	ath11k_dp_srng_cleanup(ar->ab, &dp->rxdma_mon_buf_ring.refill_buf_ring);
+}
+
+static int ath11k_dp_rx_pdev_srng_alloc(struct ath11k *ar)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct dp_srng *srng = NULL;
+	int ret;
+
+	ret = ath11k_dp_srng_setup(ar->ab,
+				   &dp->rx_refill_buf_ring.refill_buf_ring,
+				   HAL_RXDMA_BUF, 0,
+				   dp->mac_id, DP_RXDMA_BUF_RING_SIZE);
+	if (ret) {
+		ath11k_warn(ar->ab, "failed to setup rx_refill_buf_ring\n");
+		return ret;
+	}
+
+	ret = ath11k_dp_srng_setup(ar->ab, &dp->reo_dst_ring, HAL_REO_DST,
+				   dp->mac_id, dp->mac_id,
+				   DP_REO_DST_RING_SIZE);
+	if (ret) {
+		ath11k_warn(ar->ab, "failed to setup reo_dst_ring\n");
+		return ret;
+	}
+
+	ret = ath11k_dp_srng_setup(ar->ab, &dp->rxdma_err_dst_ring,
+				   HAL_RXDMA_DST, 0, dp->mac_id,
+				   DP_RXDMA_ERR_DST_RING_SIZE);
+	if (ret) {
+		ath11k_warn(ar->ab, "failed to setup rxdma_err_dst_ring\n");
+		return ret;
+	}
+
+	srng = &dp->rx_mon_status_refill_ring.refill_buf_ring;
+	ret = ath11k_dp_srng_setup(ar->ab,
+				   srng,
+				   HAL_RXDMA_MONITOR_STATUS, 0, dp->mac_id,
+				   DP_RXDMA_MON_STATUS_RING_SIZE);
+	if (ret) {
+		ath11k_warn(ar->ab,
+			    "failed to setup rx_mon_status_refill_ring\n");
+		return ret;
+	}
+	ret = ath11k_dp_srng_setup(ar->ab,
+				   &dp->rxdma_mon_buf_ring.refill_buf_ring,
+				   HAL_RXDMA_MONITOR_BUF, 0, dp->mac_id,
+				   DP_RXDMA_MONITOR_BUF_RING_SIZE);
+	if (ret) {
+		ath11k_warn(ar->ab,
+			    "failed to setup HAL_RXDMA_MONITOR_BUF\n");
+		return ret;
+	}
+
+	ret = ath11k_dp_srng_setup(ar->ab, &dp->rxdma_mon_dst_ring,
+				   HAL_RXDMA_MONITOR_DST, 0, dp->mac_id,
+				   DP_RXDMA_MONITOR_DST_RING_SIZE);
+	if (ret) {
+		ath11k_warn(ar->ab,
+			    "failed to setup HAL_RXDMA_MONITOR_DST\n");
+		return ret;
+	}
+
+	ret = ath11k_dp_srng_setup(ar->ab, &dp->rxdma_mon_desc_ring,
+				   HAL_RXDMA_MONITOR_DESC, 0, dp->mac_id,
+				   DP_RXDMA_MONITOR_DESC_RING_SIZE);
+	if (ret) {
+		ath11k_warn(ar->ab,
+			    "failed to setup HAL_RXDMA_MONITOR_DESC\n");
+		return ret;
+	}
+
+	return 0;
+}
+
+void ath11k_dp_reo_cmd_list_cleanup(struct ath11k_base *ab)
+{
+	struct ath11k_dp *dp = &ab->dp;
+	struct dp_reo_cmd *cmd, *tmp;
+	struct dp_reo_cache_flush_elem *cmd_cache, *tmp_cache;
+
+	spin_lock_bh(&dp->reo_cmd_lock);
+	list_for_each_entry_safe(cmd, tmp, &dp->reo_cmd_list, list) {
+		list_del(&cmd->list);
+		dma_unmap_single(ab->dev, cmd->data.paddr,
+				 cmd->data.size, DMA_BIDIRECTIONAL);
+		kfree(cmd->data.vaddr);
+		kfree(cmd);
+	}
+
+	list_for_each_entry_safe(cmd_cache, tmp_cache,
+				 &dp->reo_cmd_cache_flush_list, list) {
+		list_del(&cmd_cache->list);
+		dma_unmap_single(ab->dev, cmd_cache->data.paddr,
+				 cmd_cache->data.size, DMA_BIDIRECTIONAL);
+		kfree(cmd_cache->data.vaddr);
+		kfree(cmd_cache);
+	}
+	spin_unlock_bh(&dp->reo_cmd_lock);
+}
+
+static void ath11k_dp_reo_cmd_free(struct ath11k_dp *dp, void *ctx,
+				   enum hal_reo_cmd_status status)
+{
+	struct dp_rx_tid *rx_tid = ctx;
+
+	if (status != HAL_REO_CMD_SUCCESS)
+		ath11k_warn(dp->ab, "failed to flush rx tid hw desc, tid %d status %d\n",
+			    rx_tid->tid, status);
+
+	dma_unmap_single(dp->ab->dev, rx_tid->paddr, rx_tid->size,
+			 DMA_BIDIRECTIONAL);
+	kfree(rx_tid->vaddr);
+}
+
+static void ath11k_dp_reo_cache_flush(struct ath11k_base *ab,
+				      struct dp_rx_tid *rx_tid)
+{
+	struct ath11k_hal_reo_cmd cmd = {0};
+	unsigned long tot_desc_sz, desc_sz;
+	int ret;
+
+	tot_desc_sz = rx_tid->size;
+	desc_sz = ath11k_hal_reo_qdesc_size(0, HAL_DESC_REO_NON_QOS_TID);
+
+	while (tot_desc_sz > desc_sz) {
+		tot_desc_sz -= desc_sz;
+		cmd.addr_lo = lower_32_bits(rx_tid->paddr + tot_desc_sz);
+		cmd.addr_hi = upper_32_bits(rx_tid->paddr);
+		ret = ath11k_dp_tx_send_reo_cmd(ab, rx_tid,
+						HAL_REO_CMD_FLUSH_CACHE, &cmd,
+						NULL);
+		if (ret)
+			ath11k_warn(ab,
+				    "failed to send HAL_REO_CMD_FLUSH_CACHE, tid %d (%d)\n",
+				    rx_tid->tid, ret);
+	}
+
+	memset(&cmd, 0, sizeof(cmd));
+	cmd.addr_lo = lower_32_bits(rx_tid->paddr);
+	cmd.addr_hi = upper_32_bits(rx_tid->paddr);
+	cmd.flag |= HAL_REO_CMD_FLG_NEED_STATUS;
+	ret = ath11k_dp_tx_send_reo_cmd(ab, rx_tid,
+					HAL_REO_CMD_FLUSH_CACHE,
+					&cmd, ath11k_dp_reo_cmd_free);
+	if (ret) {
+		ath11k_err(ab, "failed to send HAL_REO_CMD_FLUSH_CACHE cmd, tid %d (%d)\n",
+			   rx_tid->tid, ret);
+		dma_unmap_single(ab->dev, rx_tid->paddr, rx_tid->size,
+				 DMA_BIDIRECTIONAL);
+		kfree(rx_tid->vaddr);
+	}
+}
+
+static void ath11k_dp_rx_tid_del_func(struct ath11k_dp *dp, void *ctx,
+				      enum hal_reo_cmd_status status)
+{
+	struct ath11k_base *ab = dp->ab;
+	struct dp_rx_tid *rx_tid = ctx;
+	struct dp_reo_cache_flush_elem *elem, *tmp;
+
+	if (status == HAL_REO_CMD_DRAIN) {
+		goto free_desc;
+	} else if (status != HAL_REO_CMD_SUCCESS) {
+		/* Shouldn't happen! Cleanup in case of other failure? */
+		ath11k_warn(ab, "failed to delete rx tid %d hw descriptor %d\n",
+			    rx_tid->tid, status);
+		return;
+	}
+
+	elem = kzalloc(sizeof(*elem), GFP_ATOMIC);
+	if (!elem)
+		goto free_desc;
+
+	elem->ts = jiffies;
+	memcpy(&elem->data, rx_tid, sizeof(*rx_tid));
+
+	spin_lock_bh(&dp->reo_cmd_lock);
+	list_add_tail(&elem->list, &dp->reo_cmd_cache_flush_list);
+	spin_unlock_bh(&dp->reo_cmd_lock);
+
+	/* Flush and invalidate aged REO desc from HW cache */
+	spin_lock_bh(&dp->reo_cmd_lock);
+	list_for_each_entry_safe(elem, tmp, &dp->reo_cmd_cache_flush_list,
+				 list) {
+		if (time_after(jiffies, elem->ts +
+			       msecs_to_jiffies(DP_REO_DESC_FREE_TIMEOUT_MS))) {
+			list_del(&elem->list);
+			spin_unlock_bh(&dp->reo_cmd_lock);
+
+			ath11k_dp_reo_cache_flush(ab, &elem->data);
+			kfree(elem);
+			spin_lock_bh(&dp->reo_cmd_lock);
+		}
+	}
+	spin_unlock_bh(&dp->reo_cmd_lock);
+
+	return;
+free_desc:
+	dma_unmap_single(ab->dev, rx_tid->paddr, rx_tid->size,
+			 DMA_BIDIRECTIONAL);
+	kfree(rx_tid->vaddr);
+}
+
+static void ath11k_peer_rx_tid_delete(struct ath11k *ar,
+				      struct ath11k_peer *peer, u8 tid)
+{
+	struct ath11k_hal_reo_cmd cmd = {0};
+	struct dp_rx_tid *rx_tid = &peer->rx_tid[tid];
+	int ret;
+
+	if (!rx_tid->active)
+		return;
+
+	cmd.flag = HAL_REO_CMD_FLG_NEED_STATUS;
+	cmd.addr_lo = lower_32_bits(rx_tid->paddr);
+	cmd.addr_hi = upper_32_bits(rx_tid->paddr);
+	cmd.upd0 |= HAL_REO_CMD_UPD0_VLD;
+	ret = ath11k_dp_tx_send_reo_cmd(ar->ab, rx_tid,
+					HAL_REO_CMD_UPDATE_RX_QUEUE, &cmd,
+					ath11k_dp_rx_tid_del_func);
+	if (ret) {
+		ath11k_err(ar->ab, "failed to send HAL_REO_CMD_UPDATE_RX_QUEUE cmd, tid %d (%d)\n",
+			   tid, ret);
+		dma_unmap_single(ar->ab->dev, rx_tid->paddr, rx_tid->size,
+				 DMA_BIDIRECTIONAL);
+		kfree(rx_tid->vaddr);
+	}
+
+	rx_tid->active = false;
+}
+
+void ath11k_peer_rx_tid_cleanup(struct ath11k *ar, struct ath11k_peer *peer)
+{
+	int i;
+
+	for (i = 0; i <= IEEE80211_NUM_TIDS; i++)
+		ath11k_peer_rx_tid_delete(ar, peer, i);
+}
+
+static int ath11k_peer_rx_tid_reo_update(struct ath11k *ar,
+					 struct ath11k_peer *peer,
+					 struct dp_rx_tid *rx_tid,
+					 u32 ba_win_sz, u16 ssn)
+{
+	struct ath11k_hal_reo_cmd cmd = {0};
+	int ret;
+
+	cmd.addr_lo = lower_32_bits(rx_tid->paddr);
+	cmd.addr_hi = upper_32_bits(rx_tid->paddr);
+	cmd.flag = HAL_REO_CMD_FLG_NEED_STATUS;
+	cmd.upd0 = HAL_REO_CMD_UPD0_BA_WINDOW_SIZE |
+		   HAL_REO_CMD_UPD0_SSN;
+	cmd.ba_window_size = ba_win_sz;
+	cmd.upd2 = FIELD_PREP(HAL_REO_CMD_UPD2_SSN, ssn);
+
+	ret = ath11k_dp_tx_send_reo_cmd(ar->ab, rx_tid,
+					HAL_REO_CMD_UPDATE_RX_QUEUE, &cmd,
+					NULL);
+	if (ret) {
+		ath11k_warn(ar->ab, "failed to update rx tid queue, tid %d (%d)\n",
+			    rx_tid->tid, ret);
+		return ret;
+	}
+
+	rx_tid->ba_win_sz = ba_win_sz;
+
+	return 0;
+}
+
+static void ath11k_dp_rx_tid_mem_free(struct ath11k_base *ab,
+				      const u8 *peer_mac, int vdev_id, u8 tid)
+{
+	struct ath11k_peer *peer;
+	struct dp_rx_tid *rx_tid;
+
+	spin_lock_bh(&ab->base_lock);
+
+	peer = ath11k_peer_find(ab, vdev_id, peer_mac);
+	if (!peer) {
+		ath11k_warn(ab, "failed to find the peer to free up rx tid mem\n");
+		goto unlock_exit;
+	}
+
+	rx_tid = &peer->rx_tid[tid];
+	if (!rx_tid->active)
+		goto unlock_exit;
+
+	dma_unmap_single(ab->dev, rx_tid->paddr, rx_tid->size,
+			 DMA_BIDIRECTIONAL);
+	kfree(rx_tid->vaddr);
+
+	rx_tid->active = false;
+
+unlock_exit:
+	spin_unlock_bh(&ab->base_lock);
+}
+
+int ath11k_peer_rx_tid_setup(struct ath11k *ar, const u8 *peer_mac, int vdev_id,
+			     u8 tid, u32 ba_win_sz, u16 ssn)
+{
+	struct ath11k_base *ab = ar->ab;
+	struct ath11k_peer *peer;
+	struct dp_rx_tid *rx_tid;
+	u32 hw_desc_sz;
+	u32 *addr_aligned;
+	void *vaddr;
+	dma_addr_t paddr;
+	int ret;
+
+	spin_lock_bh(&ab->base_lock);
+
+	peer = ath11k_peer_find(ab, vdev_id, peer_mac);
+	if (!peer) {
+		ath11k_warn(ab, "failed to find the peer to set up rx tid\n");
+		spin_unlock_bh(&ab->base_lock);
+		return -ENOENT;
+	}
+
+	rx_tid = &peer->rx_tid[tid];
+	/* Update the tid queue if it is already setup */
+	if (rx_tid->active) {
+		paddr = rx_tid->paddr;
+		ret = ath11k_peer_rx_tid_reo_update(ar, peer, rx_tid,
+						    ba_win_sz, ssn);
+		spin_unlock_bh(&ab->base_lock);
+		if (ret) {
+			ath11k_warn(ab, "failed to update reo for rx tid %d\n", tid);
+			return ret;
+		}
+
+		ret = ath11k_wmi_peer_rx_reorder_queue_setup(ar, vdev_id,
+							     peer_mac, paddr,
+							     tid, 1, ba_win_sz);
+		if (ret)
+			ath11k_warn(ab, "failed to send wmi command to update rx reorder queue, tid :%d (%d)\n",
+				    tid, ret);
+		return ret;
+	}
+
+	rx_tid->tid = tid;
+
+	rx_tid->ba_win_sz = ba_win_sz;
+
+	/* TODO: Optimize the memory allocation for qos tid based on the
+	 * the actual BA window size in REO tid update path.
+	 */
+	if (tid == HAL_DESC_REO_NON_QOS_TID)
+		hw_desc_sz = ath11k_hal_reo_qdesc_size(ba_win_sz, tid);
+	else
+		hw_desc_sz = ath11k_hal_reo_qdesc_size(DP_BA_WIN_SZ_MAX, tid);
+
+	vaddr = kzalloc(hw_desc_sz + HAL_LINK_DESC_ALIGN - 1, GFP_KERNEL);
+	if (!vaddr) {
+		spin_unlock_bh(&ab->base_lock);
+		return -ENOMEM;
+	}
+
+	addr_aligned = PTR_ALIGN(vaddr, HAL_LINK_DESC_ALIGN);
+
+	ath11k_hal_reo_qdesc_setup(addr_aligned, tid, ba_win_sz, ssn);
+
+	paddr = dma_map_single(ab->dev, addr_aligned, hw_desc_sz,
+			       DMA_BIDIRECTIONAL);
+
+	ret = dma_mapping_error(ab->dev, paddr);
+	if (ret) {
+		spin_unlock_bh(&ab->base_lock);
+		goto err_mem_free;
+	}
+
+	rx_tid->vaddr = vaddr;
+	rx_tid->paddr = paddr;
+	rx_tid->size = hw_desc_sz;
+	rx_tid->active = true;
+
+	spin_unlock_bh(&ab->base_lock);
+
+	ret = ath11k_wmi_peer_rx_reorder_queue_setup(ar, vdev_id, peer_mac,
+						     paddr, tid, 1, ba_win_sz);
+	if (ret) {
+		ath11k_warn(ar->ab, "failed to setup rx reorder queue, tid :%d (%d)\n",
+			    tid, ret);
+		ath11k_dp_rx_tid_mem_free(ab, peer_mac, vdev_id, tid);
+	}
+
+	return ret;
+
+err_mem_free:
+	kfree(vaddr);
+
+	return ret;
+}
+
+int ath11k_dp_rx_ampdu_start(struct ath11k *ar,
+			     struct ieee80211_ampdu_params *params)
+{
+	struct ath11k_base *ab = ar->ab;
+	struct ath11k_sta *arsta = (void *)params->sta->drv_priv;
+	int vdev_id = arsta->arvif->vdev_id;
+	int ret;
+
+	ret = ath11k_peer_rx_tid_setup(ar, params->sta->addr, vdev_id,
+				       params->tid, params->buf_size,
+				       params->ssn);
+	if (ret)
+		ath11k_warn(ab, "failed to setup rx tid %d\n", ret);
+
+	return ret;
+}
+
+int ath11k_dp_rx_ampdu_stop(struct ath11k *ar,
+			    struct ieee80211_ampdu_params *params)
+{
+	struct ath11k_base *ab = ar->ab;
+	struct ath11k_peer *peer;
+	struct ath11k_sta *arsta = (void *)params->sta->drv_priv;
+	int vdev_id = arsta->arvif->vdev_id;
+	dma_addr_t paddr;
+	bool active;
+	int ret;
+
+	spin_lock_bh(&ab->base_lock);
+
+	peer = ath11k_peer_find(ab, vdev_id, params->sta->addr);
+	if (!peer) {
+		ath11k_warn(ab, "failed to find the peer to stop rx aggregation\n");
+		spin_unlock_bh(&ab->base_lock);
+		return -ENOENT;
+	}
+
+	paddr = peer->rx_tid[params->tid].paddr;
+	active = peer->rx_tid[params->tid].active;
+
+	ath11k_peer_rx_tid_delete(ar, peer, params->tid);
+
+	spin_unlock_bh(&ab->base_lock);
+
+	if (!active)
+		return 0;
+
+	ret = ath11k_wmi_peer_rx_reorder_queue_setup(ar, vdev_id,
+						     params->sta->addr, paddr,
+						     params->tid, 1, 1);
+	if (ret)
+		ath11k_warn(ab, "failed to send wmi to delete rx tid %d\n",
+			    ret);
+
+	return ret;
+}
+
+static int ath11k_get_ppdu_user_index(struct htt_ppdu_stats *ppdu_stats,
+				      u16 peer_id)
+{
+	int i;
+
+	for (i = 0; i < HTT_PPDU_STATS_MAX_USERS - 1; i++) {
+		if (ppdu_stats->user_stats[i].is_valid_peer_id) {
+			if (peer_id == ppdu_stats->user_stats[i].peer_id)
+				return i;
+		} else {
+			return i;
+		}
+	}
+
+	return -EINVAL;
+}
+
+static int ath11k_htt_tlv_ppdu_stats_parse(struct ath11k_base *ab,
+					   u16 tag, u16 len, const void *ptr,
+					   void *data)
+{
+	struct htt_ppdu_stats_info *ppdu_info;
+	struct htt_ppdu_user_stats *user_stats;
+	int cur_user;
+	u16 peer_id;
+
+	ppdu_info = (struct htt_ppdu_stats_info *)data;
+
+	switch (tag) {
+	case HTT_PPDU_STATS_TAG_COMMON:
+		if (len < sizeof(struct htt_ppdu_stats_common)) {
+			ath11k_warn(ab, "Invalid len %d for the tag 0x%x\n",
+				    len, tag);
+			return -EINVAL;
+		}
+		memcpy((void *)&ppdu_info->ppdu_stats.common, ptr,
+		       sizeof(struct htt_ppdu_stats_common));
+		break;
+	case HTT_PPDU_STATS_TAG_USR_RATE:
+		if (len < sizeof(struct htt_ppdu_stats_user_rate)) {
+			ath11k_warn(ab, "Invalid len %d for the tag 0x%x\n",
+				    len, tag);
+			return -EINVAL;
+		}
+
+		peer_id = ((struct htt_ppdu_stats_user_rate *)ptr)->sw_peer_id;
+		cur_user = ath11k_get_ppdu_user_index(&ppdu_info->ppdu_stats,
+						      peer_id);
+		if (cur_user < 0)
+			return -EINVAL;
+		user_stats = &ppdu_info->ppdu_stats.user_stats[cur_user];
+		user_stats->peer_id = peer_id;
+		user_stats->is_valid_peer_id = true;
+		memcpy((void *)&user_stats->rate, ptr,
+		       sizeof(struct htt_ppdu_stats_user_rate));
+		user_stats->tlv_flags |= BIT(tag);
+		break;
+	case HTT_PPDU_STATS_TAG_USR_COMPLTN_COMMON:
+		if (len < sizeof(struct htt_ppdu_stats_usr_cmpltn_cmn)) {
+			ath11k_warn(ab, "Invalid len %d for the tag 0x%x\n",
+				    len, tag);
+			return -EINVAL;
+		}
+
+		peer_id = ((struct htt_ppdu_stats_usr_cmpltn_cmn *)ptr)->sw_peer_id;
+		cur_user = ath11k_get_ppdu_user_index(&ppdu_info->ppdu_stats,
+						      peer_id);
+		if (cur_user < 0)
+			return -EINVAL;
+		user_stats = &ppdu_info->ppdu_stats.user_stats[cur_user];
+		user_stats->peer_id = peer_id;
+		user_stats->is_valid_peer_id = true;
+		memcpy((void *)&user_stats->cmpltn_cmn, ptr,
+		       sizeof(struct htt_ppdu_stats_usr_cmpltn_cmn));
+		user_stats->tlv_flags |= BIT(tag);
+		break;
+	case HTT_PPDU_STATS_TAG_USR_COMPLTN_ACK_BA_STATUS:
+		if (len <
+		    sizeof(struct htt_ppdu_stats_usr_cmpltn_ack_ba_status)) {
+			ath11k_warn(ab, "Invalid len %d for the tag 0x%x\n",
+				    len, tag);
+			return -EINVAL;
+		}
+
+		peer_id =
+		((struct htt_ppdu_stats_usr_cmpltn_ack_ba_status *)ptr)->sw_peer_id;
+		cur_user = ath11k_get_ppdu_user_index(&ppdu_info->ppdu_stats,
+						      peer_id);
+		if (cur_user < 0)
+			return -EINVAL;
+		user_stats = &ppdu_info->ppdu_stats.user_stats[cur_user];
+		user_stats->peer_id = peer_id;
+		user_stats->is_valid_peer_id = true;
+		memcpy((void *)&user_stats->ack_ba, ptr,
+		       sizeof(struct htt_ppdu_stats_usr_cmpltn_ack_ba_status));
+		user_stats->tlv_flags |= BIT(tag);
+		break;
+	}
+	return 0;
+}
+
+int ath11k_dp_htt_tlv_iter(struct ath11k_base *ab, const void *ptr, size_t len,
+			   int (*iter)(struct ath11k_base *ar, u16 tag, u16 len,
+				       const void *ptr, void *data),
+			   void *data)
+{
+	const struct htt_tlv *tlv;
+	const void *begin = ptr;
+	u16 tlv_tag, tlv_len;
+	int ret = -EINVAL;
+
+	while (len > 0) {
+		if (len < sizeof(*tlv)) {
+			ath11k_err(ab, "htt tlv parse failure at byte %zd (%zu bytes left, %zu expected)\n",
+				   ptr - begin, len, sizeof(*tlv));
+			return -EINVAL;
+		}
+		tlv = (struct htt_tlv *)ptr;
+		tlv_tag = FIELD_GET(HTT_TLV_TAG, tlv->header);
+		tlv_len = FIELD_GET(HTT_TLV_LEN, tlv->header);
+		ptr += sizeof(*tlv);
+		len -= sizeof(*tlv);
+
+		if (tlv_len > len) {
+			ath11k_err(ab, "htt tlv parse failure of tag %hhu at byte %zd (%zu bytes left, %hhu expected)\n",
+				   tlv_tag, ptr - begin, len, tlv_len);
+			return -EINVAL;
+		}
+		ret = iter(ab, tlv_tag, tlv_len, ptr, data);
+		if (ret == -ENOMEM)
+			return ret;
+
+		ptr += tlv_len;
+		len -= tlv_len;
+	}
+	return 0;
+}
+
+static u8 ath11k_bw_to_mac80211_bw(u8 bw)
+{
+	u8 ret = 0;
+
+	switch (bw) {
+	case ATH11K_BW_20:
+		ret = RATE_INFO_BW_20;
+		break;
+	case ATH11K_BW_40:
+		ret = RATE_INFO_BW_40;
+		break;
+	case ATH11K_BW_80:
+		ret = RATE_INFO_BW_80;
+		break;
+	case ATH11K_BW_160:
+		ret = RATE_INFO_BW_160;
+		break;
+	}
+
+	return ret;
+}
+
+static u32 ath11k_bw_to_mac80211_bwflags(u8 bw)
+{
+	u32 bwflags = 0;
+
+	switch (bw) {
+	case ATH11K_BW_40:
+		bwflags = IEEE80211_TX_RC_40_MHZ_WIDTH;
+		break;
+	case ATH11K_BW_80:
+		bwflags = IEEE80211_TX_RC_80_MHZ_WIDTH;
+		break;
+	case ATH11K_BW_160:
+		bwflags = IEEE80211_TX_RC_160_MHZ_WIDTH;
+		break;
+	}
+
+	return bwflags;
+}
+
+static void
+ath11k_update_per_peer_tx_stats(struct ath11k *ar,
+				struct htt_ppdu_stats *ppdu_stats, u8 user)
+{
+	struct ath11k_base *ab = ar->ab;
+	struct ath11k_peer *peer;
+	struct ieee80211_sta *sta;
+	struct ath11k_sta *arsta;
+	struct htt_ppdu_stats_user_rate *user_rate;
+	struct ieee80211_chanctx_conf *conf = NULL;
+	struct ath11k_per_peer_tx_stats *peer_stats = &ar->peer_tx_stats;
+	struct htt_ppdu_user_stats *usr_stats = &ppdu_stats->user_stats[user];
+	struct htt_ppdu_stats_common *common = &ppdu_stats->common;
+	int ret;
+	u8 flags, mcs, nss, bw, sgi, rate_idx = 0;
+	u32 succ_bytes = 0;
+	u16 rate = 0, succ_pkts = 0;
+	u32 tx_duration = 0;
+	bool is_ampdu = false;
+
+	if (!usr_stats)
+		return;
+
+	if (!(usr_stats->tlv_flags & BIT(HTT_PPDU_STATS_TAG_USR_RATE)))
+		return;
+
+	if (usr_stats->tlv_flags & BIT(HTT_PPDU_STATS_TAG_USR_COMPLTN_COMMON))
+		is_ampdu =
+			HTT_USR_CMPLTN_IS_AMPDU(usr_stats->cmpltn_cmn.flags);
+
+	if (usr_stats->tlv_flags &
+	    BIT(HTT_PPDU_STATS_TAG_USR_COMPLTN_ACK_BA_STATUS)) {
+		succ_bytes = usr_stats->ack_ba.success_bytes;
+		succ_pkts = FIELD_GET(HTT_PPDU_STATS_ACK_BA_INFO_NUM_MSDU_M,
+				      usr_stats->ack_ba.info);
+	}
+
+	if (common->fes_duration_us)
+		tx_duration = common->fes_duration_us;
+
+	user_rate = &usr_stats->rate;
+	flags = HTT_USR_RATE_PREAMBLE(user_rate->rate_flags);
+	bw = HTT_USR_RATE_BW(user_rate->rate_flags) - 2;
+	nss = HTT_USR_RATE_NSS(user_rate->rate_flags) + 1;
+	mcs = HTT_USR_RATE_MCS(user_rate->rate_flags);
+	sgi = HTT_USR_RATE_GI(user_rate->rate_flags);
+
+	/* Note: If host configured fixed rates and in some other special
+	 * cases, the broadcast/management frames are sent in different rates.
+	 * Firmware rate's control to be skipped for this?
+	 */
+
+	if (flags == WMI_RATE_PREAMBLE_VHT && mcs > 9) {
+		ath11k_warn(ab, "Invalid VHT mcs %hhd peer stats",  mcs);
+		return;
+	}
+
+	if (flags == WMI_RATE_PREAMBLE_HT && (mcs > 7 || nss < 1)) {
+		ath11k_warn(ab, "Invalid HT mcs %hhd nss %hhd peer stats",
+			    mcs, nss);
+		return;
+	}
+
+	if (flags == WMI_RATE_PREAMBLE_CCK || flags == WMI_RATE_PREAMBLE_OFDM) {
+		ret = ath11k_mac_hw_ratecode_to_legacy_rate(mcs,
+							    flags,
+							    &rate_idx,
+							    &rate);
+		if (ret < 0)
+			return;
+	}
+
+	rcu_read_lock();
+	spin_lock_bh(&ab->base_lock);
+	peer = ath11k_peer_find_by_id(ab, usr_stats->peer_id);
+
+	if (!peer || !peer->sta) {
+		spin_unlock_bh(&ab->base_lock);
+		rcu_read_unlock();
+		return;
+	}
+
+	sta = peer->sta;
+	arsta = (struct ath11k_sta *)sta->drv_priv;
+
+	memset(&arsta->txrate, 0, sizeof(arsta->txrate));
+	memset(&arsta->tx_info.status, 0, sizeof(arsta->tx_info.status));
+
+	switch (flags) {
+	case WMI_RATE_PREAMBLE_OFDM:
+		arsta->txrate.legacy = rate;
+		if (arsta->arvif && arsta->arvif->vif)
+			conf = rcu_dereference(arsta->arvif->vif->chanctx_conf);
+		if (conf && conf->def.chan->band == NL80211_BAND_5GHZ)
+			arsta->tx_info.status.rates[0].idx = rate_idx - 4;
+		break;
+	case WMI_RATE_PREAMBLE_CCK:
+		arsta->txrate.legacy = rate;
+		arsta->tx_info.status.rates[0].idx = rate_idx;
+		if (mcs > ATH11K_HW_RATE_CCK_LP_1M &&
+		    mcs <= ATH11K_HW_RATE_CCK_SP_2M)
+			arsta->tx_info.status.rates[0].flags |=
+					IEEE80211_TX_RC_USE_SHORT_PREAMBLE;
+		break;
+	case WMI_RATE_PREAMBLE_HT:
+		arsta->txrate.mcs = mcs + 8 * (nss - 1);
+		arsta->tx_info.status.rates[0].idx = arsta->txrate.mcs;
+		arsta->txrate.flags = RATE_INFO_FLAGS_MCS;
+		arsta->tx_info.status.rates[0].flags |= IEEE80211_TX_RC_MCS;
+		if (sgi) {
+			arsta->txrate.flags |= RATE_INFO_FLAGS_SHORT_GI;
+			arsta->tx_info.status.rates[0].flags |=
+					IEEE80211_TX_RC_SHORT_GI;
+		}
+		break;
+	case WMI_RATE_PREAMBLE_VHT:
+		arsta->txrate.mcs = mcs;
+		ieee80211_rate_set_vht(&arsta->tx_info.status.rates[0], mcs, nss);
+		arsta->txrate.flags = RATE_INFO_FLAGS_VHT_MCS;
+		arsta->tx_info.status.rates[0].flags |= IEEE80211_TX_RC_VHT_MCS;
+		if (sgi) {
+			arsta->txrate.flags |= RATE_INFO_FLAGS_SHORT_GI;
+			arsta->tx_info.status.rates[0].flags |=
+						IEEE80211_TX_RC_SHORT_GI;
+		}
+		break;
+	}
+
+	arsta->txrate.nss = nss;
+	arsta->txrate.bw = ath11k_bw_to_mac80211_bw(bw);
+	arsta->tx_info.status.rates[0].flags |= ath11k_bw_to_mac80211_bwflags(bw);
+
+	memcpy(&arsta->last_txrate, &arsta->txrate, sizeof(struct rate_info));
+
+	if (succ_pkts) {
+		arsta->tx_info.flags = IEEE80211_TX_STAT_ACK;
+		arsta->tx_info.status.rates[0].count = 1;
+		ieee80211_tx_rate_update(ar->hw, sta, &arsta->tx_info);
+	}
+
+	memset(peer_stats, 0, sizeof(*peer_stats));
+
+	peer_stats->succ_pkts = succ_pkts;
+	peer_stats->succ_bytes = succ_bytes;
+	peer_stats->is_ampdu = is_ampdu;
+	peer_stats->duration = tx_duration;
+	peer_stats->ba_fails =
+		HTT_USR_CMPLTN_LONG_RETRY(usr_stats->cmpltn_cmn.flags) +
+		HTT_USR_CMPLTN_SHORT_RETRY(usr_stats->cmpltn_cmn.flags);
+
+	if (ath11k_debug_is_extd_tx_stats_enabled(ar))
+		ath11k_accumulate_per_peer_tx_stats(arsta,
+						    peer_stats, rate_idx);
+
+	spin_unlock_bh(&ab->base_lock);
+	rcu_read_unlock();
+}
+
+static void ath11k_htt_update_ppdu_stats(struct ath11k *ar,
+					 struct htt_ppdu_stats *ppdu_stats)
+{
+	u8 user;
+
+	for (user = 0; user < HTT_PPDU_STATS_MAX_USERS - 1; user++)
+		ath11k_update_per_peer_tx_stats(ar, ppdu_stats, user);
+}
+
+static
+struct htt_ppdu_stats_info *ath11k_dp_htt_get_ppdu_desc(struct ath11k *ar,
+							u32 ppdu_id)
+{
+	struct htt_ppdu_stats_info *ppdu_info = NULL;
+
+	spin_lock_bh(&ar->data_lock);
+	if (!list_empty(&ar->ppdu_stats_info)) {
+		list_for_each_entry(ppdu_info, &ar->ppdu_stats_info, list) {
+			if (ppdu_info && ppdu_info->ppdu_id == ppdu_id) {
+				spin_unlock_bh(&ar->data_lock);
+				return ppdu_info;
+			}
+		}
+
+		if (ar->ppdu_stat_list_depth > HTT_PPDU_DESC_MAX_DEPTH) {
+			ppdu_info = list_first_entry(&ar->ppdu_stats_info,
+						     typeof(*ppdu_info), list);
+			list_del(&ppdu_info->list);
+			ar->ppdu_stat_list_depth--;
+			ath11k_htt_update_ppdu_stats(ar, &ppdu_info->ppdu_stats);
+			kfree(ppdu_info);
+		}
+	}
+	spin_unlock_bh(&ar->data_lock);
+
+	ppdu_info = kzalloc(sizeof(*ppdu_info), GFP_KERNEL);
+	if (!ppdu_info)
+		return NULL;
+
+	spin_lock_bh(&ar->data_lock);
+	list_add_tail(&ppdu_info->list, &ar->ppdu_stats_info);
+	ar->ppdu_stat_list_depth++;
+	spin_unlock_bh(&ar->data_lock);
+
+	return ppdu_info;
+}
+
+static int ath11k_htt_pull_ppdu_stats(struct ath11k_base *ab,
+				      struct sk_buff *skb)
+{
+	struct ath11k_htt_ppdu_stats_msg *msg;
+	struct htt_ppdu_stats_info *ppdu_info;
+	struct ath11k *ar;
+	int ret;
+	u8 pdev_id;
+	u32 ppdu_id, len;
+
+	msg = (struct ath11k_htt_ppdu_stats_msg *)skb->data;
+	len = FIELD_GET(HTT_T2H_PPDU_STATS_INFO_PAYLOAD_SIZE, msg->info);
+	pdev_id = FIELD_GET(HTT_T2H_PPDU_STATS_INFO_PDEV_ID, msg->info);
+	ppdu_id = msg->ppdu_id;
+
+	rcu_read_lock();
+	ar = ath11k_mac_get_ar_by_pdev_id(ab, pdev_id);
+	if (!ar) {
+		ret = -EINVAL;
+		goto exit;
+	}
+
+	if (ath11k_debug_is_pktlog_lite_mode_enabled(ar))
+		trace_ath11k_htt_ppdu_stats(ar, skb->data, len);
+
+	ppdu_info = ath11k_dp_htt_get_ppdu_desc(ar, ppdu_id);
+	if (!ppdu_info) {
+		ret = -EINVAL;
+		goto exit;
+	}
+
+	ppdu_info->ppdu_id = ppdu_id;
+	ret = ath11k_dp_htt_tlv_iter(ab, msg->data, len,
+				     ath11k_htt_tlv_ppdu_stats_parse,
+				     (void *)ppdu_info);
+	if (ret) {
+		ath11k_warn(ab, "Failed to parse tlv %d\n", ret);
+		goto exit;
+	}
+
+exit:
+	rcu_read_unlock();
+
+	return ret;
+}
+
+static void ath11k_htt_pktlog(struct ath11k_base *ab, struct sk_buff *skb)
+{
+	struct htt_pktlog_msg *data = (struct htt_pktlog_msg *)skb->data;
+	struct ath11k *ar;
+	u32 len;
+	u8 pdev_id;
+
+	len = FIELD_GET(HTT_T2H_PPDU_STATS_INFO_PAYLOAD_SIZE, data->hdr);
+
+	if (len > ATH11K_HTT_PKTLOG_MAX_SIZE) {
+		ath11k_warn(ab, "htt pktlog buffer size %d, expected < %d\n",
+			    len,
+			    ATH11K_HTT_PKTLOG_MAX_SIZE);
+		return;
+	}
+
+	pdev_id = FIELD_GET(HTT_T2H_PPDU_STATS_INFO_PDEV_ID, data->hdr);
+	pdev_id = DP_HW2SW_MACID(pdev_id);
+	ar = ab->pdevs[pdev_id].ar;
+
+	trace_ath11k_htt_pktlog(ar, data->payload, len);
+}
+
+void ath11k_dp_htt_htc_t2h_msg_handler(struct ath11k_base *ab,
+				       struct sk_buff *skb)
+{
+	struct ath11k_dp *dp = &ab->dp;
+	struct htt_resp_msg *resp = (struct htt_resp_msg *)skb->data;
+	enum htt_t2h_msg_type type = FIELD_GET(HTT_T2H_MSG_TYPE, *(u32 *)resp);
+	u16 peer_id;
+	u8 vdev_id;
+	u8 mac_addr[ETH_ALEN];
+	u16 peer_mac_h16;
+	u16 ast_hash;
+
+	ath11k_dbg(ab, ATH11K_DBG_DP_HTT, "dp_htt rx msg type :0x%0x\n", type);
+
+	switch (type) {
+	case HTT_T2H_MSG_TYPE_VERSION_CONF:
+		dp->htt_tgt_ver_major = FIELD_GET(HTT_T2H_VERSION_CONF_MAJOR,
+						  resp->version_msg.version);
+		dp->htt_tgt_ver_minor = FIELD_GET(HTT_T2H_VERSION_CONF_MINOR,
+						  resp->version_msg.version);
+		complete(&dp->htt_tgt_version_received);
+		break;
+	case HTT_T2H_MSG_TYPE_PEER_MAP:
+		vdev_id = FIELD_GET(HTT_T2H_PEER_MAP_INFO_VDEV_ID,
+				    resp->peer_map_ev.info);
+		peer_id = FIELD_GET(HTT_T2H_PEER_MAP_INFO_PEER_ID,
+				    resp->peer_map_ev.info);
+		peer_mac_h16 = FIELD_GET(HTT_T2H_PEER_MAP_INFO1_MAC_ADDR_H16,
+					 resp->peer_map_ev.info1);
+		ath11k_dp_get_mac_addr(resp->peer_map_ev.mac_addr_l32,
+				       peer_mac_h16, mac_addr);
+		ast_hash = FIELD_GET(HTT_T2H_PEER_MAP_INFO2_AST_HASH_VAL,
+				     resp->peer_map_ev.info1);
+		ath11k_peer_map_event(ab, vdev_id, peer_id, mac_addr, ast_hash);
+		break;
+	case HTT_T2H_MSG_TYPE_PEER_UNMAP:
+		peer_id = FIELD_GET(HTT_T2H_PEER_UNMAP_INFO_PEER_ID,
+				    resp->peer_unmap_ev.info);
+		ath11k_peer_unmap_event(ab, peer_id);
+		break;
+	case HTT_T2H_MSG_TYPE_PPDU_STATS_IND:
+		ath11k_htt_pull_ppdu_stats(ab, skb);
+		break;
+	case HTT_T2H_MSG_TYPE_EXT_STATS_CONF:
+		ath11k_dbg_htt_ext_stats_handler(ab, skb);
+		break;
+	case HTT_T2H_MSG_TYPE_PKTLOG:
+		ath11k_htt_pktlog(ab, skb);
+		break;
+	default:
+		ath11k_warn(ab, "htt event %d not handled\n", type);
+		break;
+	}
+
+	dev_kfree_skb_any(skb);
+}
+
+static int ath11k_dp_rx_msdu_coalesce(struct ath11k *ar,
+				      struct sk_buff_head *msdu_list,
+				      struct sk_buff *first, struct sk_buff *last,
+				      u8 l3pad_bytes, int msdu_len)
+{
+	struct sk_buff *skb;
+	struct ath11k_skb_rxcb *rxcb = ATH11K_SKB_RXCB(first);
+	struct hal_rx_desc *ldesc;
+	int space_extra;
+	int rem_len;
+	int buf_len;
+
+	if (!rxcb->is_continuation) {
+		skb_put(first, HAL_RX_DESC_SIZE + l3pad_bytes + msdu_len);
+		skb_pull(first, HAL_RX_DESC_SIZE + l3pad_bytes);
+		return 0;
+	}
+
+	if (WARN_ON_ONCE(msdu_len <= (DP_RX_BUFFER_SIZE -
+			 (HAL_RX_DESC_SIZE + l3pad_bytes)))) {
+		skb_put(first, HAL_RX_DESC_SIZE + l3pad_bytes + msdu_len);
+		skb_pull(first, HAL_RX_DESC_SIZE + l3pad_bytes);
+		return 0;
+	}
+
+	ldesc = (struct hal_rx_desc *)last->data;
+	rxcb->is_first_msdu = ath11k_dp_rx_h_msdu_end_first_msdu(ldesc);
+	rxcb->is_last_msdu = ath11k_dp_rx_h_msdu_end_last_msdu(ldesc);
+
+	/* MSDU spans over multiple buffers because the length of the MSDU
+	 * exceeds DP_RX_BUFFER_SIZE - HAL_RX_DESC_SIZE. So assume the data
+	 * in the first buf is of length DP_RX_BUFFER_SIZE - HAL_RX_DESC_SIZE.
+	 */
+	skb_put(first, DP_RX_BUFFER_SIZE);
+	skb_pull(first, HAL_RX_DESC_SIZE + l3pad_bytes);
+
+	space_extra = msdu_len - (DP_RX_BUFFER_SIZE + skb_tailroom(first));
+	if (space_extra > 0 &&
+	    (pskb_expand_head(first, 0, space_extra, GFP_ATOMIC) < 0)) {
+		/* Free up all buffers of the MSDU */
+		while ((skb = __skb_dequeue(msdu_list)) != NULL) {
+			rxcb = ATH11K_SKB_RXCB(skb);
+			if (!rxcb->is_continuation) {
+				dev_kfree_skb_any(skb);
+				break;
+			}
+			dev_kfree_skb_any(skb);
+		}
+		return -ENOMEM;
+	}
+
+	/* When an MSDU spread over multiple buffers attention, MSDU_END and
+	 * MPDU_END tlvs are valid only in the last buffer. Copy those tlvs.
+	 */
+	ath11k_dp_rx_desc_end_tlv_copy(rxcb->rx_desc, ldesc);
+
+	rem_len = msdu_len -
+		  (DP_RX_BUFFER_SIZE - HAL_RX_DESC_SIZE - l3pad_bytes);
+	while ((skb = __skb_dequeue(msdu_list)) != NULL && rem_len > 0) {
+		rxcb = ATH11K_SKB_RXCB(skb);
+		if (rxcb->is_continuation)
+			buf_len = DP_RX_BUFFER_SIZE - HAL_RX_DESC_SIZE;
+		else
+			buf_len = rem_len;
+
+		if (buf_len > (DP_RX_BUFFER_SIZE - HAL_RX_DESC_SIZE)) {
+			WARN_ON_ONCE(1);
+			dev_kfree_skb_any(skb);
+			return -EINVAL;
+		}
+
+		skb_put(skb, buf_len + HAL_RX_DESC_SIZE);
+		skb_pull(skb, HAL_RX_DESC_SIZE);
+		skb_copy_from_linear_data(skb, skb_put(first, buf_len),
+					  buf_len);
+		dev_kfree_skb_any(skb);
+
+		rem_len -= buf_len;
+		if (!rxcb->is_continuation)
+			break;
+	}
+
+	return 0;
+}
+
+static struct sk_buff *ath11k_dp_rx_get_msdu_last_buf(struct sk_buff_head *msdu_list,
+						      struct sk_buff *first)
+{
+	struct sk_buff *skb;
+	struct ath11k_skb_rxcb *rxcb = ATH11K_SKB_RXCB(first);
+
+	if (!rxcb->is_continuation)
+		return first;
+
+	skb_queue_walk(msdu_list, skb) {
+		rxcb = ATH11K_SKB_RXCB(skb);
+		if (!rxcb->is_continuation)
+			return skb;
+	}
+
+	return NULL;
+}
+
+static int ath11k_dp_rx_retrieve_amsdu(struct ath11k *ar,
+				       struct sk_buff_head *msdu_list,
+				       struct sk_buff_head *amsdu_list)
+{
+	struct sk_buff *msdu = skb_peek(msdu_list);
+	struct sk_buff *last_buf;
+	struct ath11k_skb_rxcb *rxcb;
+	struct ieee80211_hdr *hdr;
+	struct hal_rx_desc *rx_desc, *lrx_desc;
+	u16 msdu_len;
+	u8 l3_pad_bytes;
+	u8 *hdr_status;
+	int ret;
+
+	if (!msdu)
+		return -ENOENT;
+
+	rx_desc = (struct hal_rx_desc *)msdu->data;
+	hdr_status = ath11k_dp_rx_h_80211_hdr(rx_desc);
+	hdr = (struct ieee80211_hdr *)hdr_status;
+	/* Process only data frames */
+	if (!ieee80211_is_data(hdr->frame_control)) {
+		__skb_unlink(msdu, msdu_list);
+		dev_kfree_skb_any(msdu);
+		return -EINVAL;
+	}
+
+	do {
+		__skb_unlink(msdu, msdu_list);
+		last_buf = ath11k_dp_rx_get_msdu_last_buf(msdu_list, msdu);
+		if (!last_buf) {
+			ath11k_warn(ar->ab,
+				    "No valid Rx buffer to access Atten/MSDU_END/MPDU_END tlvs\n");
+			ret = -EIO;
+			goto free_out;
+		}
+
+		rx_desc = (struct hal_rx_desc *)msdu->data;
+		lrx_desc = (struct hal_rx_desc *)last_buf->data;
+
+		if (!ath11k_dp_rx_h_attn_msdu_done(lrx_desc)) {
+			ath11k_warn(ar->ab, "msdu_done bit in attention is not set\n");
+			ret = -EIO;
+			goto free_out;
+		}
+
+		rxcb = ATH11K_SKB_RXCB(msdu);
+		rxcb->rx_desc = rx_desc;
+		msdu_len = ath11k_dp_rx_h_msdu_start_msdu_len(rx_desc);
+		l3_pad_bytes = ath11k_dp_rx_h_msdu_end_l3pad(lrx_desc);
+
+		if (!rxcb->is_continuation) {
+			skb_put(msdu, HAL_RX_DESC_SIZE + l3_pad_bytes + msdu_len);
+			skb_pull(msdu, HAL_RX_DESC_SIZE + l3_pad_bytes);
+		} else {
+			ret = ath11k_dp_rx_msdu_coalesce(ar, msdu_list,
+							 msdu, last_buf,
+							 l3_pad_bytes, msdu_len);
+			if (ret) {
+				ath11k_warn(ar->ab,
+					    "failed to coalesce msdu rx buffer%d\n", ret);
+				goto free_out;
+			}
+		}
+		__skb_queue_tail(amsdu_list, msdu);
+
+		/* Should we also consider msdu_cnt from mpdu_meta while
+		 * preparing amsdu list?
+		 */
+		if (rxcb->is_last_msdu)
+			break;
+	} while ((msdu = skb_peek(msdu_list)) != NULL);
+
+	return 0;
+
+free_out:
+	dev_kfree_skb_any(msdu);
+	__skb_queue_purge(amsdu_list);
+
+	return ret;
+}
+
+static void ath11k_dp_rx_h_csum_offload(struct sk_buff *msdu)
+{
+	struct ath11k_skb_rxcb *rxcb = ATH11K_SKB_RXCB(msdu);
+	bool ip_csum_fail, l4_csum_fail;
+
+	ip_csum_fail = ath11k_dp_rx_h_attn_ip_cksum_fail(rxcb->rx_desc);
+	l4_csum_fail = ath11k_dp_rx_h_attn_l4_cksum_fail(rxcb->rx_desc);
+
+	msdu->ip_summed = (ip_csum_fail || l4_csum_fail) ?
+			  CHECKSUM_NONE : CHECKSUM_UNNECESSARY;
+}
+
+static int ath11k_dp_rx_crypto_mic_len(struct ath11k *ar,
+				       enum hal_encrypt_type enctype)
+{
+	switch (enctype) {
+	case HAL_ENCRYPT_TYPE_OPEN:
+	case HAL_ENCRYPT_TYPE_TKIP_NO_MIC:
+	case HAL_ENCRYPT_TYPE_TKIP_MIC:
+		return 0;
+	case HAL_ENCRYPT_TYPE_CCMP_128:
+		return IEEE80211_CCMP_MIC_LEN;
+	case HAL_ENCRYPT_TYPE_CCMP_256:
+		return IEEE80211_CCMP_256_MIC_LEN;
+	case HAL_ENCRYPT_TYPE_GCMP_128:
+	case HAL_ENCRYPT_TYPE_AES_GCMP_256:
+		return IEEE80211_GCMP_MIC_LEN;
+	case HAL_ENCRYPT_TYPE_WEP_40:
+	case HAL_ENCRYPT_TYPE_WEP_104:
+	case HAL_ENCRYPT_TYPE_WEP_128:
+	case HAL_ENCRYPT_TYPE_WAPI_GCM_SM4:
+	case HAL_ENCRYPT_TYPE_WAPI:
+		break;
+	}
+
+	ath11k_warn(ar->ab, "unsupported encryption type %d for mic len\n", enctype);
+	return 0;
+}
+
+static int ath11k_dp_rx_crypto_param_len(struct ath11k *ar,
+					 enum hal_encrypt_type enctype)
+{
+	switch (enctype) {
+	case HAL_ENCRYPT_TYPE_OPEN:
+		return 0;
+	case HAL_ENCRYPT_TYPE_TKIP_NO_MIC:
+	case HAL_ENCRYPT_TYPE_TKIP_MIC:
+		return IEEE80211_TKIP_IV_LEN;
+	case HAL_ENCRYPT_TYPE_CCMP_128:
+		return IEEE80211_CCMP_HDR_LEN;
+	case HAL_ENCRYPT_TYPE_CCMP_256:
+		return IEEE80211_CCMP_256_HDR_LEN;
+	case HAL_ENCRYPT_TYPE_GCMP_128:
+	case HAL_ENCRYPT_TYPE_AES_GCMP_256:
+		return IEEE80211_GCMP_HDR_LEN;
+	case HAL_ENCRYPT_TYPE_WEP_40:
+	case HAL_ENCRYPT_TYPE_WEP_104:
+	case HAL_ENCRYPT_TYPE_WEP_128:
+	case HAL_ENCRYPT_TYPE_WAPI_GCM_SM4:
+	case HAL_ENCRYPT_TYPE_WAPI:
+		break;
+	}
+
+	ath11k_warn(ar->ab, "unsupported encryption type %d\n", enctype);
+	return 0;
+}
+
+static int ath11k_dp_rx_crypto_icv_len(struct ath11k *ar,
+				       enum hal_encrypt_type enctype)
+{
+	switch (enctype) {
+	case HAL_ENCRYPT_TYPE_OPEN:
+	case HAL_ENCRYPT_TYPE_CCMP_128:
+	case HAL_ENCRYPT_TYPE_CCMP_256:
+	case HAL_ENCRYPT_TYPE_GCMP_128:
+	case HAL_ENCRYPT_TYPE_AES_GCMP_256:
+		return 0;
+	case HAL_ENCRYPT_TYPE_TKIP_NO_MIC:
+	case HAL_ENCRYPT_TYPE_TKIP_MIC:
+		return IEEE80211_TKIP_ICV_LEN;
+	case HAL_ENCRYPT_TYPE_WEP_40:
+	case HAL_ENCRYPT_TYPE_WEP_104:
+	case HAL_ENCRYPT_TYPE_WEP_128:
+	case HAL_ENCRYPT_TYPE_WAPI_GCM_SM4:
+	case HAL_ENCRYPT_TYPE_WAPI:
+		break;
+	}
+
+	ath11k_warn(ar->ab, "unsupported encryption type %d\n", enctype);
+	return 0;
+}
+
+static void ath11k_dp_rx_h_undecap_nwifi(struct ath11k *ar,
+					 struct sk_buff *msdu,
+					 u8 *first_hdr,
+					 enum hal_encrypt_type enctype,
+					 struct ieee80211_rx_status *status)
+{
+	struct ieee80211_hdr *hdr;
+	size_t hdr_len;
+	u8 da[ETH_ALEN];
+	u8 sa[ETH_ALEN];
+
+	/* pull decapped header and copy SA & DA */
+	hdr = (struct ieee80211_hdr *)msdu->data;
+	ether_addr_copy(da, ieee80211_get_DA(hdr));
+	ether_addr_copy(sa, ieee80211_get_SA(hdr));
+	skb_pull(msdu, ieee80211_hdrlen(hdr->frame_control));
+
+	/* push original 802.11 header */
+	hdr = (struct ieee80211_hdr *)first_hdr;
+	hdr_len = ieee80211_hdrlen(hdr->frame_control);
+
+	if (!(status->flag & RX_FLAG_IV_STRIPPED)) {
+		memcpy(skb_push(msdu,
+				ath11k_dp_rx_crypto_param_len(ar, enctype)),
+		       (void *)hdr + hdr_len,
+		       ath11k_dp_rx_crypto_param_len(ar, enctype));
+	}
+
+	memcpy(skb_push(msdu, hdr_len), hdr, hdr_len);
+
+	/* original 802.11 header has a different DA and in
+	 * case of 4addr it may also have different SA
+	 */
+	hdr = (struct ieee80211_hdr *)msdu->data;
+	ether_addr_copy(ieee80211_get_DA(hdr), da);
+	ether_addr_copy(ieee80211_get_SA(hdr), sa);
+}
+
+static void ath11k_dp_rx_h_undecap_raw(struct ath11k *ar, struct sk_buff *msdu,
+				       enum hal_encrypt_type enctype,
+				       struct ieee80211_rx_status *status,
+				       bool decrypted)
+{
+	struct ath11k_skb_rxcb *rxcb = ATH11K_SKB_RXCB(msdu);
+	struct ieee80211_hdr *hdr;
+	size_t hdr_len;
+	size_t crypto_len;
+
+	if (!rxcb->is_first_msdu ||
+	    !(rxcb->is_first_msdu && rxcb->is_last_msdu)) {
+		WARN_ON_ONCE(1);
+		return;
+	}
+
+	skb_trim(msdu, msdu->len - FCS_LEN);
+
+	if (!decrypted)
+		return;
+
+	hdr = (void *)msdu->data;
+
+	/* Tail */
+	if (status->flag & RX_FLAG_IV_STRIPPED) {
+		skb_trim(msdu, msdu->len -
+			 ath11k_dp_rx_crypto_mic_len(ar, enctype));
+
+		skb_trim(msdu, msdu->len -
+			 ath11k_dp_rx_crypto_icv_len(ar, enctype));
+	} else {
+		/* MIC */
+		if (status->flag & RX_FLAG_MIC_STRIPPED)
+			skb_trim(msdu, msdu->len -
+				 ath11k_dp_rx_crypto_mic_len(ar, enctype));
+
+		/* ICV */
+		if (status->flag & RX_FLAG_ICV_STRIPPED)
+			skb_trim(msdu, msdu->len -
+				 ath11k_dp_rx_crypto_icv_len(ar, enctype));
+	}
+
+	/* MMIC */
+	if ((status->flag & RX_FLAG_MMIC_STRIPPED) &&
+	    !ieee80211_has_morefrags(hdr->frame_control) &&
+	    enctype == HAL_ENCRYPT_TYPE_TKIP_MIC)
+		skb_trim(msdu, msdu->len - IEEE80211_CCMP_MIC_LEN);
+
+	/* Head */
+	if (status->flag & RX_FLAG_IV_STRIPPED) {
+		hdr_len = ieee80211_hdrlen(hdr->frame_control);
+		crypto_len = ath11k_dp_rx_crypto_param_len(ar, enctype);
+
+		memmove((void *)msdu->data + crypto_len,
+			(void *)msdu->data, hdr_len);
+		skb_pull(msdu, crypto_len);
+	}
+}
+
+static void *ath11k_dp_rx_h_find_rfc1042(struct ath11k *ar,
+					 struct sk_buff *msdu,
+					 enum hal_encrypt_type enctype)
+{
+	struct ath11k_skb_rxcb *rxcb = ATH11K_SKB_RXCB(msdu);
+	struct ieee80211_hdr *hdr;
+	size_t hdr_len, crypto_len;
+	void *rfc1042;
+	bool is_amsdu;
+
+	is_amsdu = !(rxcb->is_first_msdu && rxcb->is_last_msdu);
+	hdr = (struct ieee80211_hdr *)ath11k_dp_rx_h_80211_hdr(rxcb->rx_desc);
+	rfc1042 = hdr;
+
+	if (rxcb->is_first_msdu) {
+		hdr_len = ieee80211_hdrlen(hdr->frame_control);
+		crypto_len = ath11k_dp_rx_crypto_param_len(ar, enctype);
+
+		rfc1042 += hdr_len + crypto_len;
+	}
+
+	if (is_amsdu)
+		rfc1042 += sizeof(struct ath11k_dp_amsdu_subframe_hdr);
+
+	return rfc1042;
+}
+
+static void ath11k_dp_rx_h_undecap_eth(struct ath11k *ar,
+				       struct sk_buff *msdu,
+				       u8 *first_hdr,
+				       enum hal_encrypt_type enctype,
+				       struct ieee80211_rx_status *status)
+{
+	struct ieee80211_hdr *hdr;
+	struct ethhdr *eth;
+	size_t hdr_len;
+	u8 da[ETH_ALEN];
+	u8 sa[ETH_ALEN];
+	void *rfc1042;
+
+	rfc1042 = ath11k_dp_rx_h_find_rfc1042(ar, msdu, enctype);
+	if (WARN_ON_ONCE(!rfc1042))
+		return;
+
+	/* pull decapped header and copy SA & DA */
+	eth = (struct ethhdr *)msdu->data;
+	ether_addr_copy(da, eth->h_dest);
+	ether_addr_copy(sa, eth->h_source);
+	skb_pull(msdu, sizeof(struct ethhdr));
+
+	/* push rfc1042/llc/snap */
+	memcpy(skb_push(msdu, sizeof(struct ath11k_dp_rfc1042_hdr)), rfc1042,
+	       sizeof(struct ath11k_dp_rfc1042_hdr));
+
+	/* push original 802.11 header */
+	hdr = (struct ieee80211_hdr *)first_hdr;
+	hdr_len = ieee80211_hdrlen(hdr->frame_control);
+
+	if (!(status->flag & RX_FLAG_IV_STRIPPED)) {
+		memcpy(skb_push(msdu,
+				ath11k_dp_rx_crypto_param_len(ar, enctype)),
+		       (void *)hdr + hdr_len,
+		       ath11k_dp_rx_crypto_param_len(ar, enctype));
+	}
+
+	memcpy(skb_push(msdu, hdr_len), hdr, hdr_len);
+
+	/* original 802.11 header has a different DA and in
+	 * case of 4addr it may also have different SA
+	 */
+	hdr = (struct ieee80211_hdr *)msdu->data;
+	ether_addr_copy(ieee80211_get_DA(hdr), da);
+	ether_addr_copy(ieee80211_get_SA(hdr), sa);
+}
+
+static void ath11k_dp_rx_h_undecap(struct ath11k *ar, struct sk_buff *msdu,
+				   struct hal_rx_desc *rx_desc,
+				   enum hal_encrypt_type enctype,
+				   struct ieee80211_rx_status *status,
+				   bool decrypted)
+{
+	u8 *first_hdr;
+	u8 decap;
+
+	first_hdr = ath11k_dp_rx_h_80211_hdr(rx_desc);
+	decap = ath11k_dp_rx_h_mpdu_start_decap_type(rx_desc);
+
+	switch (decap) {
+	case DP_RX_DECAP_TYPE_NATIVE_WIFI:
+		ath11k_dp_rx_h_undecap_nwifi(ar, msdu, first_hdr,
+					     enctype, status);
+		break;
+	case DP_RX_DECAP_TYPE_RAW:
+		ath11k_dp_rx_h_undecap_raw(ar, msdu, enctype, status,
+					   decrypted);
+		break;
+	case DP_RX_DECAP_TYPE_ETHERNET2_DIX:
+		ath11k_dp_rx_h_undecap_eth(ar, msdu, first_hdr,
+					   enctype, status);
+		break;
+	case DP_RX_DECAP_TYPE_8023:
+		/* TODO: Handle undecap for these formats */
+		break;
+	}
+}
+
+static void ath11k_dp_rx_h_mpdu(struct ath11k *ar,
+				struct sk_buff_head *amsdu_list,
+				struct hal_rx_desc *rx_desc,
+				struct ieee80211_rx_status *rx_status)
+{
+	struct ieee80211_hdr *hdr;
+	enum hal_encrypt_type enctype;
+	struct sk_buff *last_msdu;
+	struct sk_buff *msdu;
+	struct ath11k_skb_rxcb *last_rxcb;
+	bool is_decrypted;
+	u32 err_bitmap;
+	u8 *qos;
+
+	if (skb_queue_empty(amsdu_list))
+		return;
+
+	hdr = (struct ieee80211_hdr *)ath11k_dp_rx_h_80211_hdr(rx_desc);
+
+	/* Each A-MSDU subframe will use the original header as the base and be
+	 * reported as a separate MSDU so strip the A-MSDU bit from QoS Ctl.
+	 */
+	if (ieee80211_is_data_qos(hdr->frame_control)) {
+		qos = ieee80211_get_qos_ctl(hdr);
+		qos[0] &= ~IEEE80211_QOS_CTL_A_MSDU_PRESENT;
+	}
+
+	is_decrypted = ath11k_dp_rx_h_attn_is_decrypted(rx_desc);
+	enctype = ath11k_dp_rx_h_mpdu_start_enctype(rx_desc);
+
+	/* Some attention flags are valid only in the last MSDU. */
+	last_msdu = skb_peek_tail(amsdu_list);
+	last_rxcb = ATH11K_SKB_RXCB(last_msdu);
+
+	err_bitmap = ath11k_dp_rx_h_attn_mpdu_err(last_rxcb->rx_desc);
+
+	/* Clear per-MPDU flags while leaving per-PPDU flags intact. */
+	rx_status->flag &= ~(RX_FLAG_FAILED_FCS_CRC |
+			     RX_FLAG_MMIC_ERROR |
+			     RX_FLAG_DECRYPTED |
+			     RX_FLAG_IV_STRIPPED |
+			     RX_FLAG_MMIC_STRIPPED);
+
+	if (err_bitmap & DP_RX_MPDU_ERR_FCS)
+		rx_status->flag |= RX_FLAG_FAILED_FCS_CRC;
+
+	if (err_bitmap & DP_RX_MPDU_ERR_TKIP_MIC)
+		rx_status->flag |= RX_FLAG_MMIC_ERROR;
+
+	if (is_decrypted)
+		rx_status->flag |= RX_FLAG_DECRYPTED | RX_FLAG_MMIC_STRIPPED |
+				   RX_FLAG_MIC_STRIPPED | RX_FLAG_ICV_STRIPPED;
+
+	skb_queue_walk(amsdu_list, msdu) {
+		ath11k_dp_rx_h_csum_offload(msdu);
+		ath11k_dp_rx_h_undecap(ar, msdu, rx_desc,
+				       enctype, rx_status, is_decrypted);
+	}
+}
+
+static void ath11k_dp_rx_h_rate(struct ath11k *ar, struct hal_rx_desc *rx_desc,
+				struct ieee80211_rx_status *rx_status)
+{
+	struct ieee80211_supported_band *sband;
+	enum rx_msdu_start_pkt_type pkt_type;
+	u8 bw;
+	u8 rate_mcs, nss;
+	u8 sgi;
+	bool is_cck;
+
+	pkt_type = ath11k_dp_rx_h_msdu_start_pkt_type(rx_desc);
+	bw = ath11k_dp_rx_h_msdu_start_rx_bw(rx_desc);
+	rate_mcs = ath11k_dp_rx_h_msdu_start_rate_mcs(rx_desc);
+	nss = ath11k_dp_rx_h_msdu_start_nss(rx_desc);
+	sgi = ath11k_dp_rx_h_msdu_start_sgi(rx_desc);
+
+	switch (pkt_type) {
+	case RX_MSDU_START_PKT_TYPE_11A:
+	case RX_MSDU_START_PKT_TYPE_11B:
+		is_cck = (pkt_type == RX_MSDU_START_PKT_TYPE_11B);
+		sband = &ar->mac.sbands[rx_status->band];
+		rx_status->rate_idx = ath11k_mac_hw_rate_to_idx(sband, rate_mcs,
+								is_cck);
+		break;
+	case RX_MSDU_START_PKT_TYPE_11N:
+		rx_status->encoding = RX_ENC_HT;
+		if (rate_mcs > ATH11K_HT_MCS_MAX) {
+			ath11k_warn(ar->ab,
+				    "Received with invalid mcs in HT mode %d\n",
+				     rate_mcs);
+			break;
+		}
+		rx_status->rate_idx = rate_mcs + (8 * (nss - 1));
+		if (sgi)
+			rx_status->enc_flags |= RX_ENC_FLAG_SHORT_GI;
+		rx_status->bw = ath11k_bw_to_mac80211_bw(bw);
+		break;
+	case RX_MSDU_START_PKT_TYPE_11AC:
+		rx_status->encoding = RX_ENC_VHT;
+		rx_status->rate_idx = rate_mcs;
+		if (rate_mcs > ATH11K_VHT_MCS_MAX) {
+			ath11k_warn(ar->ab,
+				    "Received with invalid mcs in VHT mode %d\n",
+				     rate_mcs);
+			break;
+		}
+		rx_status->nss = nss;
+		if (sgi)
+			rx_status->enc_flags |= RX_ENC_FLAG_SHORT_GI;
+		rx_status->bw = ath11k_bw_to_mac80211_bw(bw);
+		break;
+	case RX_MSDU_START_PKT_TYPE_11AX:
+		rx_status->rate_idx = rate_mcs;
+		if (rate_mcs > ATH11K_HE_MCS_MAX) {
+			ath11k_warn(ar->ab,
+				    "Received with invalid mcs in HE mode %d\n",
+				    rate_mcs);
+			break;
+		}
+		rx_status->encoding = RX_ENC_HE;
+		rx_status->nss = nss;
+		rx_status->bw = ath11k_bw_to_mac80211_bw(bw);
+		break;
+	}
+}
+
+static void ath11k_dp_rx_h_ppdu(struct ath11k *ar, struct hal_rx_desc *rx_desc,
+				struct ieee80211_rx_status *rx_status)
+{
+	u8 channel_num;
+
+	rx_status->freq = 0;
+	rx_status->rate_idx = 0;
+	rx_status->nss = 0;
+	rx_status->encoding = RX_ENC_LEGACY;
+	rx_status->bw = RATE_INFO_BW_20;
+
+	rx_status->flag |= RX_FLAG_NO_SIGNAL_VAL;
+
+	channel_num = ath11k_dp_rx_h_msdu_start_freq(rx_desc);
+
+	if (channel_num >= 1 && channel_num <= 14) {
+		rx_status->band = NL80211_BAND_2GHZ;
+	} else if (channel_num >= 36 && channel_num <= 173) {
+		rx_status->band = NL80211_BAND_5GHZ;
+	} else {
+		ath11k_warn(ar->ab, "Unsupported Channel info received %d\n",
+			    channel_num);
+		return;
+	}
+
+	rx_status->freq = ieee80211_channel_to_frequency(channel_num,
+							 rx_status->band);
+
+	ath11k_dp_rx_h_rate(ar, rx_desc, rx_status);
+}
+
+static void ath11k_dp_rx_process_amsdu(struct ath11k *ar,
+				       struct sk_buff_head *amsdu_list,
+				       struct ieee80211_rx_status *rx_status)
+{
+	struct sk_buff *first;
+	struct ath11k_skb_rxcb *rxcb;
+	struct hal_rx_desc *rx_desc;
+	bool first_mpdu;
+
+	if (skb_queue_empty(amsdu_list))
+		return;
+
+	first = skb_peek(amsdu_list);
+	rxcb = ATH11K_SKB_RXCB(first);
+	rx_desc = rxcb->rx_desc;
+
+	first_mpdu = ath11k_dp_rx_h_attn_first_mpdu(rx_desc);
+	if (first_mpdu)
+		ath11k_dp_rx_h_ppdu(ar, rx_desc, rx_status);
+
+	ath11k_dp_rx_h_mpdu(ar, amsdu_list, rx_desc, rx_status);
+}
+
+static char *ath11k_print_get_tid(struct ieee80211_hdr *hdr, char *out,
+				  size_t size)
+{
+	u8 *qc;
+	int tid;
+
+	if (!ieee80211_is_data_qos(hdr->frame_control))
+		return "";
+
+	qc = ieee80211_get_qos_ctl(hdr);
+	tid = *qc & IEEE80211_QOS_CTL_TID_MASK;
+	snprintf(out, size, "tid %d", tid);
+
+	return out;
+}
+
+static void ath11k_dp_rx_deliver_msdu(struct ath11k *ar, struct napi_struct *napi,
+				      struct sk_buff *msdu)
+{
+	struct ieee80211_rx_status *status;
+	struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)msdu->data;
+	char tid[32];
+
+	status = IEEE80211_SKB_RXCB(msdu);
+
+	ath11k_dbg(ar->ab, ATH11K_DBG_DATA,
+		   "rx skb %pK len %u peer %pM %s %s sn %u %s%s%s%s%s%s%s %srate_idx %u vht_nss %u freq %u band %u flag 0x%x fcs-err %i mic-err %i amsdu-more %i\n",
+		   msdu,
+		   msdu->len,
+		   ieee80211_get_SA(hdr),
+		   ath11k_print_get_tid(hdr, tid, sizeof(tid)),
+		   is_multicast_ether_addr(ieee80211_get_DA(hdr)) ?
+							"mcast" : "ucast",
+		   (__le16_to_cpu(hdr->seq_ctrl) & IEEE80211_SCTL_SEQ) >> 4,
+		   (status->encoding == RX_ENC_LEGACY) ? "legacy" : "",
+		   (status->encoding == RX_ENC_HT) ? "ht" : "",
+		   (status->encoding == RX_ENC_VHT) ? "vht" : "",
+		   (status->encoding == RX_ENC_HE) ? "he" : "",
+		   (status->bw == RATE_INFO_BW_40) ? "40" : "",
+		   (status->bw == RATE_INFO_BW_80) ? "80" : "",
+		   (status->bw == RATE_INFO_BW_160) ? "160" : "",
+		   status->enc_flags & RX_ENC_FLAG_SHORT_GI ? "sgi " : "",
+		   status->rate_idx,
+		   status->nss,
+		   status->freq,
+		   status->band, status->flag,
+		   !!(status->flag & RX_FLAG_FAILED_FCS_CRC),
+		   !!(status->flag & RX_FLAG_MMIC_ERROR),
+		   !!(status->flag & RX_FLAG_AMSDU_MORE));
+
+	/* TODO: trace rx packet */
+
+	ieee80211_rx_napi(ar->hw, NULL, msdu, napi);
+}
+
+static void ath11k_dp_rx_pre_deliver_amsdu(struct ath11k *ar,
+					   struct sk_buff_head *amsdu_list,
+					   struct ieee80211_rx_status *rxs)
+{
+	struct sk_buff *msdu;
+	struct sk_buff *first_subframe;
+	struct ieee80211_rx_status *status;
+
+	first_subframe = skb_peek(amsdu_list);
+
+	skb_queue_walk(amsdu_list, msdu) {
+		/* Setup per-MSDU flags */
+		if (skb_queue_empty(amsdu_list))
+			rxs->flag &= ~RX_FLAG_AMSDU_MORE;
+		else
+			rxs->flag |= RX_FLAG_AMSDU_MORE;
+
+		if (msdu == first_subframe) {
+			first_subframe = NULL;
+			rxs->flag &= ~RX_FLAG_ALLOW_SAME_PN;
+		} else {
+			rxs->flag |= RX_FLAG_ALLOW_SAME_PN;
+		}
+		rxs->flag |= RX_FLAG_SKIP_MONITOR;
+
+		status = IEEE80211_SKB_RXCB(msdu);
+		*status = *rxs;
+	}
+}
+
+static void ath11k_dp_rx_process_pending_packets(struct ath11k_base *ab,
+						 struct napi_struct *napi,
+						 struct sk_buff_head *pending_q,
+						 int *quota, u8 mac_id)
+{
+	struct ath11k *ar;
+	struct sk_buff *msdu;
+	struct ath11k_pdev *pdev;
+
+	if (skb_queue_empty(pending_q))
+		return;
+
+	ar = ab->pdevs[mac_id].ar;
+
+	rcu_read_lock();
+	pdev = rcu_dereference(ab->pdevs_active[mac_id]);
+
+	while (*quota && (msdu = __skb_dequeue(pending_q))) {
+		if (!pdev) {
+			dev_kfree_skb_any(msdu);
+			continue;
+		}
+
+		ath11k_dp_rx_deliver_msdu(ar, napi, msdu);
+		(*quota)--;
+	}
+	rcu_read_unlock();
+}
+
+int ath11k_dp_process_rx(struct ath11k_base *ab, int mac_id,
+			 struct napi_struct *napi, struct sk_buff_head *pending_q,
+			 int budget)
+{
+	struct ath11k *ar = ab->pdevs[mac_id].ar;
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct ieee80211_rx_status *rx_status = &dp->rx_status;
+	struct dp_rxdma_ring *rx_ring = &dp->rx_refill_buf_ring;
+	struct hal_srng *srng;
+	struct hal_rx_meta_info meta_info;
+	struct sk_buff *msdu;
+	struct sk_buff_head msdu_list;
+	struct sk_buff_head amsdu_list;
+	struct ath11k_skb_rxcb *rxcb;
+	u32 *rx_desc;
+	int buf_id;
+	int num_buffs_reaped = 0;
+	int quota = budget;
+	int ret;
+	bool done = false;
+
+	/* Process any pending packets from the previous napi poll.
+	 * Note: All msdu's in this pending_q corresponds to the same mac id
+	 * due to pdev based reo dest mapping and also since each irq group id
+	 * maps to specific reo dest ring.
+	 */
+	ath11k_dp_rx_process_pending_packets(ab, napi, pending_q, &quota,
+					     mac_id);
+
+	/* If all quota is exhausted by processing the pending_q,
+	 * Wait for the next napi poll to reap the new info
+	 */
+	if (!quota)
+		goto exit;
+
+	__skb_queue_head_init(&msdu_list);
+
+	srng = &ab->hal.srng_list[dp->reo_dst_ring.ring_id];
+
+	spin_lock_bh(&srng->lock);
+
+	ath11k_hal_srng_access_begin(ab, srng);
+
+try_again:
+	while ((rx_desc = ath11k_hal_srng_dst_get_next_entry(ab, srng))) {
+		memset(&meta_info, 0, sizeof(meta_info));
+		ath11k_hal_rx_parse_dst_ring_desc(ab, rx_desc, &meta_info);
+
+		buf_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_BUF_ID,
+				   meta_info.msdu_meta.cookie);
+		spin_lock_bh(&rx_ring->idr_lock);
+		msdu = idr_find(&rx_ring->bufs_idr, buf_id);
+		if (!msdu) {
+			ath11k_warn(ab, "frame rx with invalid buf_id %d\n",
+				    buf_id);
+			spin_unlock_bh(&rx_ring->idr_lock);
+			continue;
+		}
+
+		idr_remove(&rx_ring->bufs_idr, buf_id);
+		spin_unlock_bh(&rx_ring->idr_lock);
+
+		rxcb = ATH11K_SKB_RXCB(msdu);
+		dma_unmap_single(ab->dev, rxcb->paddr,
+				 msdu->len + skb_tailroom(msdu),
+				 DMA_FROM_DEVICE);
+
+		num_buffs_reaped++;
+
+		if (meta_info.push_reason !=
+		    HAL_REO_DEST_RING_PUSH_REASON_ROUTING_INSTRUCTION) {
+			/* TODO: Check if the msdu can be sent up for processing */
+			dev_kfree_skb_any(msdu);
+			ab->soc_stats.hal_reo_error[dp->reo_dst_ring.ring_id]++;
+			continue;
+		}
+
+		rxcb->is_first_msdu = meta_info.msdu_meta.first;
+		rxcb->is_last_msdu = meta_info.msdu_meta.last;
+		rxcb->is_continuation = meta_info.msdu_meta.continuation;
+		rxcb->mac_id = mac_id;
+		__skb_queue_tail(&msdu_list, msdu);
+
+		/* Stop reaping from the ring once quota is exhausted
+		 * and we've received all msdu's in the the AMSDU. The
+		 * additional msdu's reaped in excess of quota here would
+		 * be pushed into the pending queue to be processed during
+		 * the next napi poll.
+		 * Note: More profiling can be done to see the impact on
+		 * pending_q and throughput during various traffic & density
+		 * and how use of budget instead of remaining quota affects it.
+		 */
+		if (num_buffs_reaped >= quota && rxcb->is_last_msdu &&
+		    !rxcb->is_continuation) {
+			done = true;
+			break;
+		}
+	}
+
+	/* Hw might have updated the head pointer after we cached it.
+	 * In this case, even though there are entries in the ring we'll
+	 * get rx_desc NULL. Give the read another try with updated cached
+	 * head pointer so that we can reap complete MPDU in the current
+	 * rx processing.
+	 */
+	if (!done && ath11k_hal_srng_dst_num_free(ab, srng, true)) {
+		ath11k_hal_srng_access_end(ab, srng);
+		goto try_again;
+	}
+
+	ath11k_hal_srng_access_end(ab, srng);
+
+	spin_unlock_bh(&srng->lock);
+
+	if (!num_buffs_reaped)
+		goto exit;
+
+	/* Should we reschedule it later if we are not able to replenish all
+	 * the buffers?
+	 */
+	ath11k_dp_rxbufs_replenish(ab, mac_id, rx_ring, num_buffs_reaped,
+				   HAL_RX_BUF_RBM_SW3_BM, GFP_ATOMIC);
+
+	rcu_read_lock();
+	if (!rcu_dereference(ab->pdevs_active[mac_id])) {
+		__skb_queue_purge(&msdu_list);
+		goto rcu_unlock;
+	}
+
+	if (test_bit(ATH11K_CAC_RUNNING, &ar->dev_flags)) {
+		__skb_queue_purge(&msdu_list);
+		goto rcu_unlock;
+	}
+
+	while (!skb_queue_empty(&msdu_list)) {
+		__skb_queue_head_init(&amsdu_list);
+		ret = ath11k_dp_rx_retrieve_amsdu(ar, &msdu_list, &amsdu_list);
+		if (ret) {
+			if (ret == -EIO) {
+				ath11k_err(ab, "rx ring got corrupted %d\n", ret);
+				__skb_queue_purge(&msdu_list);
+				/* Should stop processing any more rx in
+				 * future from this ring?
+				 */
+				goto rcu_unlock;
+			}
+
+			/* A-MSDU retrieval got failed due to non-fatal condition,
+			 * continue processing with the next msdu.
+			 */
+			continue;
+		}
+
+		ath11k_dp_rx_process_amsdu(ar, &amsdu_list, rx_status);
+
+		ath11k_dp_rx_pre_deliver_amsdu(ar, &amsdu_list, rx_status);
+		skb_queue_splice_tail(&amsdu_list, pending_q);
+	}
+
+	while (quota && (msdu = __skb_dequeue(pending_q))) {
+		ath11k_dp_rx_deliver_msdu(ar, napi, msdu);
+		quota--;
+	}
+
+rcu_unlock:
+	rcu_read_unlock();
+exit:
+	return budget - quota;
+}
+
+static void ath11k_dp_rx_update_peer_stats(struct ath11k_sta *arsta,
+					   struct hal_rx_mon_ppdu_info *ppdu_info)
+{
+	struct ath11k_rx_peer_stats *rx_stats = arsta->rx_stats;
+	u32 num_msdu;
+
+	if (!rx_stats)
+		return;
+
+	num_msdu = ppdu_info->tcp_msdu_count + ppdu_info->tcp_ack_msdu_count +
+		   ppdu_info->udp_msdu_count + ppdu_info->other_msdu_count;
+
+	rx_stats->num_msdu += num_msdu;
+	rx_stats->tcp_msdu_count += ppdu_info->tcp_msdu_count +
+				    ppdu_info->tcp_ack_msdu_count;
+	rx_stats->udp_msdu_count += ppdu_info->udp_msdu_count;
+	rx_stats->other_msdu_count += ppdu_info->other_msdu_count;
+
+	if (ppdu_info->preamble_type == HAL_RX_PREAMBLE_11A ||
+	    ppdu_info->preamble_type == HAL_RX_PREAMBLE_11B) {
+		ppdu_info->nss = 1;
+		ppdu_info->mcs = HAL_RX_MAX_MCS;
+		ppdu_info->tid = IEEE80211_NUM_TIDS;
+	}
+
+	if (ppdu_info->nss > 0 && ppdu_info->nss <= HAL_RX_MAX_NSS)
+		rx_stats->nss_count[ppdu_info->nss - 1] += num_msdu;
+
+	if (ppdu_info->mcs <= HAL_RX_MAX_MCS)
+		rx_stats->mcs_count[ppdu_info->mcs] += num_msdu;
+
+	if (ppdu_info->gi < HAL_RX_GI_MAX)
+		rx_stats->gi_count[ppdu_info->gi] += num_msdu;
+
+	if (ppdu_info->bw < HAL_RX_BW_MAX)
+		rx_stats->bw_count[ppdu_info->bw] += num_msdu;
+
+	if (ppdu_info->ldpc < HAL_RX_SU_MU_CODING_MAX)
+		rx_stats->coding_count[ppdu_info->ldpc] += num_msdu;
+
+	if (ppdu_info->tid <= IEEE80211_NUM_TIDS)
+		rx_stats->tid_count[ppdu_info->tid] += num_msdu;
+
+	if (ppdu_info->preamble_type < HAL_RX_PREAMBLE_MAX)
+		rx_stats->pream_cnt[ppdu_info->preamble_type] += num_msdu;
+
+	if (ppdu_info->reception_type < HAL_RX_RECEPTION_TYPE_MAX)
+		rx_stats->reception_type[ppdu_info->reception_type] += num_msdu;
+
+	if (ppdu_info->is_stbc)
+		rx_stats->stbc_count += num_msdu;
+
+	if (ppdu_info->beamformed)
+		rx_stats->beamformed_count += num_msdu;
+
+	if (ppdu_info->num_mpdu_fcs_ok > 1)
+		rx_stats->ampdu_msdu_count += num_msdu;
+	else
+		rx_stats->non_ampdu_msdu_count += num_msdu;
+
+	rx_stats->num_mpdu_fcs_ok += ppdu_info->num_mpdu_fcs_ok;
+	rx_stats->num_mpdu_fcs_err += ppdu_info->num_mpdu_fcs_err;
+
+	arsta->rssi_comb = ppdu_info->rssi_comb;
+	rx_stats->rx_duration += ppdu_info->rx_duration;
+	arsta->rx_duration = rx_stats->rx_duration;
+}
+
+static struct sk_buff *ath11k_dp_rx_alloc_mon_status_buf(struct ath11k_base *ab,
+							 struct dp_rxdma_ring *rx_ring,
+							 int *buf_id, gfp_t gfp)
+{
+	struct sk_buff *skb;
+	dma_addr_t paddr;
+
+	skb = dev_alloc_skb(DP_RX_BUFFER_SIZE +
+			    DP_RX_BUFFER_ALIGN_SIZE);
+
+	if (!skb)
+		goto fail_alloc_skb;
+
+	if (!IS_ALIGNED((unsigned long)skb->data,
+			DP_RX_BUFFER_ALIGN_SIZE)) {
+		skb_pull(skb, PTR_ALIGN(skb->data, DP_RX_BUFFER_ALIGN_SIZE) -
+			 skb->data);
+	}
+
+	paddr = dma_map_single(ab->dev, skb->data,
+			       skb->len + skb_tailroom(skb),
+			       DMA_BIDIRECTIONAL);
+	if (unlikely(dma_mapping_error(ab->dev, paddr)))
+		goto fail_free_skb;
+
+	spin_lock_bh(&rx_ring->idr_lock);
+	*buf_id = idr_alloc(&rx_ring->bufs_idr, skb, 0,
+			    rx_ring->bufs_max, gfp);
+	spin_unlock_bh(&rx_ring->idr_lock);
+	if (*buf_id < 0)
+		goto fail_dma_unmap;
+
+	ATH11K_SKB_RXCB(skb)->paddr = paddr;
+	return skb;
+
+fail_dma_unmap:
+	dma_unmap_single(ab->dev, paddr, skb->len + skb_tailroom(skb),
+			 DMA_BIDIRECTIONAL);
+fail_free_skb:
+	dev_kfree_skb_any(skb);
+fail_alloc_skb:
+	return NULL;
+}
+
+int ath11k_dp_rx_mon_status_bufs_replenish(struct ath11k_base *ab, int mac_id,
+					   struct dp_rxdma_ring *rx_ring,
+					   int req_entries,
+					   enum hal_rx_buf_return_buf_manager mgr,
+					   gfp_t gfp)
+{
+	struct hal_srng *srng;
+	u32 *desc;
+	struct sk_buff *skb;
+	int num_free;
+	int num_remain;
+	int buf_id;
+	u32 cookie;
+	dma_addr_t paddr;
+
+	req_entries = min(req_entries, rx_ring->bufs_max);
+
+	srng = &ab->hal.srng_list[rx_ring->refill_buf_ring.ring_id];
+
+	spin_lock_bh(&srng->lock);
+
+	ath11k_hal_srng_access_begin(ab, srng);
+
+	num_free = ath11k_hal_srng_src_num_free(ab, srng, true);
+
+	req_entries = min(num_free, req_entries);
+	num_remain = req_entries;
+
+	while (num_remain > 0) {
+		skb = ath11k_dp_rx_alloc_mon_status_buf(ab, rx_ring,
+							&buf_id, gfp);
+		if (!skb)
+			break;
+		paddr = ATH11K_SKB_RXCB(skb)->paddr;
+
+		desc = ath11k_hal_srng_src_get_next_entry(ab, srng);
+		if (!desc)
+			goto fail_desc_get;
+
+		cookie = FIELD_PREP(DP_RXDMA_BUF_COOKIE_PDEV_ID, mac_id) |
+			 FIELD_PREP(DP_RXDMA_BUF_COOKIE_BUF_ID, buf_id);
+
+		num_remain--;
+
+		ath11k_hal_rx_buf_addr_info_set(desc, paddr, cookie, mgr);
+	}
+
+	ath11k_hal_srng_access_end(ab, srng);
+
+	spin_unlock_bh(&srng->lock);
+
+	return req_entries - num_remain;
+
+fail_desc_get:
+	spin_lock_bh(&rx_ring->idr_lock);
+	idr_remove(&rx_ring->bufs_idr, buf_id);
+	spin_unlock_bh(&rx_ring->idr_lock);
+	dma_unmap_single(ab->dev, paddr, skb->len + skb_tailroom(skb),
+			 DMA_BIDIRECTIONAL);
+	dev_kfree_skb_any(skb);
+	ath11k_hal_srng_access_end(ab, srng);
+	spin_unlock_bh(&srng->lock);
+
+	return req_entries - num_remain;
+}
+
+static int ath11k_dp_rx_reap_mon_status_ring(struct ath11k_base *ab, int mac_id,
+					     int *budget, struct sk_buff_head *skb_list)
+{
+	struct ath11k *ar = ab->pdevs[mac_id].ar;
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct dp_rxdma_ring *rx_ring = &dp->rx_mon_status_refill_ring;
+	struct hal_srng *srng;
+	void *rx_mon_status_desc;
+	struct sk_buff *skb;
+	struct ath11k_skb_rxcb *rxcb;
+	struct hal_tlv_hdr *tlv;
+	u32 cookie;
+	int buf_id;
+	dma_addr_t paddr;
+	u8 rbm;
+	int num_buffs_reaped = 0;
+
+	srng = &ab->hal.srng_list[rx_ring->refill_buf_ring.ring_id];
+
+	spin_lock_bh(&srng->lock);
+
+	ath11k_hal_srng_access_begin(ab, srng);
+	while (*budget) {
+		*budget -= 1;
+		rx_mon_status_desc =
+			ath11k_hal_srng_src_peek(ab, srng);
+		if (!rx_mon_status_desc)
+			break;
+
+		ath11k_hal_rx_buf_addr_info_get(rx_mon_status_desc, &paddr,
+						&cookie, &rbm);
+		if (paddr) {
+			buf_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_BUF_ID, cookie);
+
+			spin_lock_bh(&rx_ring->idr_lock);
+			skb = idr_find(&rx_ring->bufs_idr, buf_id);
+			if (!skb) {
+				ath11k_warn(ab, "rx monitor status with invalid buf_id %d\n",
+					    buf_id);
+				spin_unlock_bh(&rx_ring->idr_lock);
+				continue;
+			}
+
+			idr_remove(&rx_ring->bufs_idr, buf_id);
+			spin_unlock_bh(&rx_ring->idr_lock);
+
+			rxcb = ATH11K_SKB_RXCB(skb);
+
+			dma_sync_single_for_cpu(ab->dev, rxcb->paddr,
+						skb->len + skb_tailroom(skb),
+						DMA_FROM_DEVICE);
+
+			dma_unmap_single(ab->dev, rxcb->paddr,
+					 skb->len + skb_tailroom(skb),
+					 DMA_BIDIRECTIONAL);
+
+			tlv = (struct hal_tlv_hdr *)skb->data;
+			if (FIELD_GET(HAL_TLV_HDR_TAG, tlv->tl) !=
+					HAL_RX_STATUS_BUFFER_DONE) {
+				ath11k_hal_srng_src_get_next_entry(ab, srng);
+				continue;
+			}
+
+			__skb_queue_tail(skb_list, skb);
+		}
+
+		skb = ath11k_dp_rx_alloc_mon_status_buf(ab, rx_ring,
+							&buf_id, GFP_ATOMIC);
+
+		if (!skb) {
+			ath11k_hal_rx_buf_addr_info_set(rx_mon_status_desc, 0, 0,
+							HAL_RX_BUF_RBM_SW3_BM);
+			num_buffs_reaped++;
+			break;
+		}
+		rxcb = ATH11K_SKB_RXCB(skb);
+
+		cookie = FIELD_PREP(DP_RXDMA_BUF_COOKIE_PDEV_ID, mac_id) |
+			 FIELD_PREP(DP_RXDMA_BUF_COOKIE_BUF_ID, buf_id);
+
+		ath11k_hal_rx_buf_addr_info_set(rx_mon_status_desc, rxcb->paddr,
+						cookie, HAL_RX_BUF_RBM_SW3_BM);
+		ath11k_hal_srng_src_get_next_entry(ab, srng);
+		num_buffs_reaped++;
+	}
+	ath11k_hal_srng_access_end(ab, srng);
+	spin_unlock_bh(&srng->lock);
+
+	return num_buffs_reaped;
+}
+
+int ath11k_dp_rx_process_mon_status(struct ath11k_base *ab, int mac_id,
+				    struct napi_struct *napi, int budget)
+{
+	struct ath11k *ar = ab->pdevs[mac_id].ar;
+	enum hal_rx_mon_status hal_status;
+	struct sk_buff *skb;
+	struct sk_buff_head skb_list;
+	struct hal_rx_mon_ppdu_info ppdu_info;
+	struct ath11k_peer *peer;
+	struct ath11k_sta *arsta;
+	int num_buffs_reaped = 0;
+
+	__skb_queue_head_init(&skb_list);
+
+	num_buffs_reaped = ath11k_dp_rx_reap_mon_status_ring(ab, mac_id, &budget,
+							     &skb_list);
+	if (!num_buffs_reaped)
+		goto exit;
+
+	while ((skb = __skb_dequeue(&skb_list))) {
+		memset(&ppdu_info, 0, sizeof(ppdu_info));
+		ppdu_info.peer_id = HAL_INVALID_PEERID;
+
+		if (ath11k_debug_is_pktlog_rx_stats_enabled(ar))
+			trace_ath11k_htt_rxdesc(ar, skb->data, DP_RX_BUFFER_SIZE);
+
+		hal_status = ath11k_hal_rx_parse_mon_status(ab, &ppdu_info, skb);
+
+		if (ppdu_info.peer_id == HAL_INVALID_PEERID ||
+		    hal_status != HAL_RX_MON_STATUS_PPDU_DONE) {
+			dev_kfree_skb_any(skb);
+			continue;
+		}
+
+		rcu_read_lock();
+		spin_lock_bh(&ab->base_lock);
+		peer = ath11k_peer_find_by_id(ab, ppdu_info.peer_id);
+
+		if (!peer || !peer->sta) {
+			ath11k_warn(ab, "failed to find the peer with peer_id %d\n",
+				    ppdu_info.peer_id);
+			spin_unlock_bh(&ab->base_lock);
+			rcu_read_unlock();
+			dev_kfree_skb_any(skb);
+			continue;
+		}
+
+		arsta = (struct ath11k_sta *)peer->sta->drv_priv;
+		ath11k_dp_rx_update_peer_stats(arsta, &ppdu_info);
+
+		if (ath11k_debug_is_pktlog_peer_valid(ar, peer->addr))
+			trace_ath11k_htt_rxdesc(ar, skb->data, DP_RX_BUFFER_SIZE);
+
+		spin_unlock_bh(&ab->base_lock);
+		rcu_read_unlock();
+
+		dev_kfree_skb_any(skb);
+	}
+exit:
+	return num_buffs_reaped;
+}
+
+static int ath11k_dp_rx_link_desc_return(struct ath11k_base *ab,
+					 u32 *link_desc,
+					 enum hal_wbm_rel_bm_act action)
+{
+	struct ath11k_dp *dp = &ab->dp;
+	struct hal_srng *srng;
+	u32 *desc;
+	int ret = 0;
+
+	srng = &ab->hal.srng_list[dp->wbm_desc_rel_ring.ring_id];
+
+	spin_lock_bh(&srng->lock);
+
+	ath11k_hal_srng_access_begin(ab, srng);
+
+	desc = ath11k_hal_srng_src_get_next_entry(ab, srng);
+	if (!desc) {
+		ret = -ENOBUFS;
+		goto exit;
+	}
+
+	ath11k_hal_rx_msdu_link_desc_set(ab, (void *)desc, (void *)link_desc,
+					 action);
+
+exit:
+	ath11k_hal_srng_access_end(ab, srng);
+
+	spin_unlock_bh(&srng->lock);
+
+	return ret;
+}
+
+static void ath11k_dp_rx_frag_h_mpdu(struct ath11k *ar,
+				     struct sk_buff *msdu,
+				     struct hal_rx_desc *rx_desc,
+				     struct ieee80211_rx_status *rx_status)
+{
+	u8 rx_channel;
+	enum hal_encrypt_type enctype;
+	bool is_decrypted;
+	u32 err_bitmap;
+
+	is_decrypted = ath11k_dp_rx_h_attn_is_decrypted(rx_desc);
+	enctype = ath11k_dp_rx_h_mpdu_start_enctype(rx_desc);
+	err_bitmap = ath11k_dp_rx_h_attn_mpdu_err(rx_desc);
+
+	if (err_bitmap & DP_RX_MPDU_ERR_FCS)
+		rx_status->flag |= RX_FLAG_FAILED_FCS_CRC;
+
+	if (err_bitmap & DP_RX_MPDU_ERR_TKIP_MIC)
+		rx_status->flag |= RX_FLAG_MMIC_ERROR;
+
+	rx_status->encoding = RX_ENC_LEGACY;
+	rx_status->bw = RATE_INFO_BW_20;
+
+	rx_status->flag |= RX_FLAG_NO_SIGNAL_VAL;
+
+	rx_channel = ath11k_dp_rx_h_msdu_start_freq(rx_desc);
+
+	if (rx_channel >= 1 && rx_channel <= 14) {
+		rx_status->band = NL80211_BAND_2GHZ;
+	} else if (rx_channel >= 36 && rx_channel <= 173) {
+		rx_status->band = NL80211_BAND_5GHZ;
+	} else {
+		ath11k_warn(ar->ab, "Unsupported Channel info received %d\n",
+			    rx_channel);
+		return;
+	}
+
+	rx_status->freq = ieee80211_channel_to_frequency(rx_channel,
+							 rx_status->band);
+	ath11k_dp_rx_h_rate(ar, rx_desc, rx_status);
+
+	/* Rx fragments are received in raw mode */
+	skb_trim(msdu, msdu->len - FCS_LEN);
+
+	if (is_decrypted) {
+		rx_status->flag |= RX_FLAG_DECRYPTED | RX_FLAG_MIC_STRIPPED;
+		skb_trim(msdu, msdu->len -
+			 ath11k_dp_rx_crypto_mic_len(ar, enctype));
+	}
+}
+
+static int
+ath11k_dp_process_rx_err_buf(struct ath11k *ar, struct napi_struct *napi,
+			     int buf_id, bool frag)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct dp_rxdma_ring *rx_ring = &dp->rx_refill_buf_ring;
+	struct ieee80211_rx_status rx_status = {0};
+	struct sk_buff *msdu;
+	struct ath11k_skb_rxcb *rxcb;
+	struct ieee80211_rx_status *status;
+	struct hal_rx_desc *rx_desc;
+	u16 msdu_len;
+
+	spin_lock_bh(&rx_ring->idr_lock);
+	msdu = idr_find(&rx_ring->bufs_idr, buf_id);
+	if (!msdu) {
+		ath11k_warn(ar->ab, "rx err buf with invalid buf_id %d\n",
+			    buf_id);
+		spin_unlock_bh(&rx_ring->idr_lock);
+		return -EINVAL;
+	}
+
+	idr_remove(&rx_ring->bufs_idr, buf_id);
+	spin_unlock_bh(&rx_ring->idr_lock);
+
+	rxcb = ATH11K_SKB_RXCB(msdu);
+	dma_unmap_single(ar->ab->dev, rxcb->paddr,
+			 msdu->len + skb_tailroom(msdu),
+			 DMA_FROM_DEVICE);
+
+	if (!frag) {
+		/* Process only rx fragments below, and drop
+		 * msdu's indicated due to error reasons.
+		 */
+		dev_kfree_skb_any(msdu);
+		return 0;
+	}
+
+	rcu_read_lock();
+	if (!rcu_dereference(ar->ab->pdevs_active[ar->pdev_idx])) {
+		dev_kfree_skb_any(msdu);
+		goto exit;
+	}
+
+	if (test_bit(ATH11K_CAC_RUNNING, &ar->dev_flags)) {
+		dev_kfree_skb_any(msdu);
+		goto exit;
+	}
+
+	rx_desc = (struct hal_rx_desc *)msdu->data;
+	msdu_len = ath11k_dp_rx_h_msdu_start_msdu_len(rx_desc);
+	skb_put(msdu, HAL_RX_DESC_SIZE + msdu_len);
+	skb_pull(msdu, HAL_RX_DESC_SIZE);
+
+	ath11k_dp_rx_frag_h_mpdu(ar, msdu, rx_desc, &rx_status);
+
+	status = IEEE80211_SKB_RXCB(msdu);
+
+	*status = rx_status;
+
+	ath11k_dp_rx_deliver_msdu(ar, napi, msdu);
+
+exit:
+	rcu_read_unlock();
+	return 0;
+}
+
+int ath11k_dp_process_rx_err(struct ath11k_base *ab, struct napi_struct *napi,
+			     int budget)
+{
+	struct hal_rx_msdu_meta meta[HAL_NUM_RX_MSDUS_PER_LINK_DESC];
+	struct dp_link_desc_bank *link_desc_banks;
+	enum hal_rx_buf_return_buf_manager rbm;
+	int tot_n_bufs_reaped, quota, ret, i;
+	int n_bufs_reaped[MAX_RADIOS] = {0};
+	struct hal_rx_meta_info meta_info;
+	struct dp_rxdma_ring *rx_ring;
+	struct dp_srng *reo_except;
+	u32 desc_bank, num_msdus;
+	struct hal_srng *srng;
+	struct ath11k_dp *dp;
+	void *link_desc_va;
+	int buf_id, mac_id;
+	struct ath11k *ar;
+	dma_addr_t paddr;
+	u32 *desc;
+	bool is_frag;
+
+	tot_n_bufs_reaped = 0;
+	quota = budget;
+
+	dp = &ab->dp;
+	reo_except = &dp->reo_except_ring;
+	link_desc_banks = dp->link_desc_banks;
+
+	srng = &ab->hal.srng_list[reo_except->ring_id];
+
+	spin_lock_bh(&srng->lock);
+
+	ath11k_hal_srng_access_begin(ab, srng);
+
+	while (budget &&
+	       (desc = ath11k_hal_srng_dst_get_next_entry(ab, srng))) {
+		ab->soc_stats.err_ring_pkts++;
+		ret = ath11k_hal_desc_reo_parse_err(ab, desc, &paddr,
+						    &desc_bank);
+		if (ret) {
+			ath11k_warn(ab, "failed to parse error reo desc %d\n",
+				    ret);
+			continue;
+		}
+		link_desc_va = link_desc_banks[desc_bank].vaddr +
+			       (paddr - link_desc_banks[desc_bank].paddr);
+		ath11k_hal_rx_msdu_link_info_get(link_desc_va, &num_msdus, meta,
+						 &rbm);
+		if (rbm != HAL_RX_BUF_RBM_WBM_IDLE_DESC_LIST &&
+		    rbm != HAL_RX_BUF_RBM_SW3_BM) {
+			ab->soc_stats.invalid_rbm++;
+			ath11k_warn(ab, "invalid return buffer manager %d\n", rbm);
+			ath11k_dp_rx_link_desc_return(ab, desc,
+						      HAL_WBM_REL_BM_ACT_REL_MSDU);
+			continue;
+		}
+
+		memset(&meta_info, 0, sizeof(meta_info));
+		ath11k_hal_rx_parse_dst_ring_desc(ab, desc, &meta_info);
+
+		is_frag = meta_info.mpdu_meta.frag;
+
+		/* Return the link desc back to wbm idle list */
+		ath11k_dp_rx_link_desc_return(ab, desc,
+					      HAL_WBM_REL_BM_ACT_PUT_IN_IDLE);
+
+		for (i = 0; i < num_msdus; i++) {
+			buf_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_BUF_ID,
+					   meta[i].cookie);
+
+			mac_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_PDEV_ID,
+					   meta[i].cookie);
+
+			ar = ab->pdevs[mac_id].ar;
+
+			if (!ath11k_dp_process_rx_err_buf(ar, napi, buf_id,
+							  is_frag)) {
+				n_bufs_reaped[mac_id]++;
+				tot_n_bufs_reaped++;
+			}
+		}
+
+		if (tot_n_bufs_reaped >= quota) {
+			tot_n_bufs_reaped = quota;
+			goto exit;
+		}
+
+		budget = quota - tot_n_bufs_reaped;
+	}
+
+exit:
+	ath11k_hal_srng_access_end(ab, srng);
+
+	spin_unlock_bh(&srng->lock);
+
+	for (i = 0; i <  ab->num_radios; i++) {
+		if (!n_bufs_reaped[i])
+			continue;
+
+		ar = ab->pdevs[i].ar;
+		rx_ring = &ar->dp.rx_refill_buf_ring;
+
+		ath11k_dp_rxbufs_replenish(ab, i, rx_ring, n_bufs_reaped[i],
+					   HAL_RX_BUF_RBM_SW3_BM, GFP_ATOMIC);
+	}
+
+	return tot_n_bufs_reaped;
+}
+
+static void ath11k_dp_rx_null_q_desc_sg_drop(struct ath11k *ar,
+					     int msdu_len,
+					     struct sk_buff_head *msdu_list)
+{
+	struct sk_buff *skb, *tmp;
+	struct ath11k_skb_rxcb *rxcb;
+	int n_buffs;
+
+	n_buffs = DIV_ROUND_UP(msdu_len,
+			       (DP_RX_BUFFER_SIZE - HAL_RX_DESC_SIZE));
+
+	skb_queue_walk_safe(msdu_list, skb, tmp) {
+		rxcb = ATH11K_SKB_RXCB(skb);
+		if (rxcb->err_rel_src == HAL_WBM_REL_SRC_MODULE_REO &&
+		    rxcb->err_code == HAL_REO_DEST_RING_ERROR_CODE_DESC_ADDR_ZERO) {
+			if (!n_buffs)
+				break;
+			__skb_unlink(skb, msdu_list);
+			dev_kfree_skb_any(skb);
+			n_buffs--;
+		}
+	}
+}
+
+static int ath11k_dp_rx_h_null_q_desc(struct ath11k *ar, struct sk_buff *msdu,
+				      struct ieee80211_rx_status *status,
+				      struct sk_buff_head *msdu_list)
+{
+	struct sk_buff_head amsdu_list;
+	u16 msdu_len;
+	struct hal_rx_desc *desc = (struct hal_rx_desc *)msdu->data;
+	u8 l3pad_bytes;
+	struct ath11k_skb_rxcb *rxcb = ATH11K_SKB_RXCB(msdu);
+
+	msdu_len = ath11k_dp_rx_h_msdu_start_msdu_len(desc);
+
+	if ((msdu_len + HAL_RX_DESC_SIZE) > DP_RX_BUFFER_SIZE) {
+		/* First buffer will be freed by the caller, so deduct it's length */
+		msdu_len = msdu_len - (DP_RX_BUFFER_SIZE - HAL_RX_DESC_SIZE);
+		ath11k_dp_rx_null_q_desc_sg_drop(ar, msdu_len, msdu_list);
+		return -EINVAL;
+	}
+
+	if (!ath11k_dp_rx_h_attn_msdu_done(desc)) {
+		ath11k_warn(ar->ab,
+			    "msdu_done bit not set in null_q_des processing\n");
+		__skb_queue_purge(msdu_list);
+		return -EIO;
+	}
+
+	/* Handle NULL queue descriptor violations arising out a missing
+	 * REO queue for a given peer or a given TID. This typically
+	 * may happen if a packet is received on a QOS enabled TID before the
+	 * ADDBA negotiation for that TID, when the TID queue is setup. Or
+	 * it may also happen for MC/BC frames if they are not routed to the
+	 * non-QOS TID queue, in the absence of any other default TID queue.
+	 * This error can show up both in a REO destination or WBM release ring.
+	 */
+
+	__skb_queue_head_init(&amsdu_list);
+
+	rxcb->is_first_msdu = ath11k_dp_rx_h_msdu_end_first_msdu(desc);
+	rxcb->is_last_msdu = ath11k_dp_rx_h_msdu_end_last_msdu(desc);
+
+	l3pad_bytes = ath11k_dp_rx_h_msdu_end_l3pad(desc);
+
+	if ((HAL_RX_DESC_SIZE + l3pad_bytes + msdu_len) > DP_RX_BUFFER_SIZE)
+		return -EINVAL;
+
+	skb_put(msdu, HAL_RX_DESC_SIZE + l3pad_bytes + msdu_len);
+	skb_pull(msdu, HAL_RX_DESC_SIZE + l3pad_bytes);
+
+	ath11k_dp_rx_h_ppdu(ar, desc, status);
+
+	__skb_queue_tail(&amsdu_list, msdu);
+
+	ath11k_dp_rx_h_mpdu(ar, &amsdu_list, desc, status);
+
+	/* Please note that caller will having the access to msdu and completing
+	 * rx with mac80211. Need not worry about cleaning up amsdu_list.
+	 */
+
+	return 0;
+}
+
+static bool ath11k_dp_rx_h_reo_err(struct ath11k *ar, struct sk_buff *msdu,
+				   struct ieee80211_rx_status *status,
+				   struct sk_buff_head *msdu_list)
+{
+	struct ath11k_skb_rxcb *rxcb = ATH11K_SKB_RXCB(msdu);
+	bool drop = false;
+
+	ar->ab->soc_stats.reo_error[rxcb->err_code]++;
+
+	switch (rxcb->err_code) {
+	case HAL_REO_DEST_RING_ERROR_CODE_DESC_ADDR_ZERO:
+		if (ath11k_dp_rx_h_null_q_desc(ar, msdu, status, msdu_list))
+			drop = true;
+		break;
+	default:
+		/* TODO: Review other errors and process them to mac80211
+		 * as appropriate.
+		 */
+		drop = true;
+		break;
+	}
+
+	return drop;
+}
+
+static void ath11k_dp_rx_h_tkip_mic_err(struct ath11k *ar, struct sk_buff *msdu,
+					struct ieee80211_rx_status *status)
+{
+	u16 msdu_len;
+	struct hal_rx_desc *desc = (struct hal_rx_desc *)msdu->data;
+	u8 l3pad_bytes;
+	struct ath11k_skb_rxcb *rxcb = ATH11K_SKB_RXCB(msdu);
+
+	rxcb->is_first_msdu = ath11k_dp_rx_h_msdu_end_first_msdu(desc);
+	rxcb->is_last_msdu = ath11k_dp_rx_h_msdu_end_last_msdu(desc);
+
+	l3pad_bytes = ath11k_dp_rx_h_msdu_end_l3pad(desc);
+	msdu_len = ath11k_dp_rx_h_msdu_start_msdu_len(desc);
+	skb_put(msdu, HAL_RX_DESC_SIZE + l3pad_bytes + msdu_len);
+	skb_pull(msdu, HAL_RX_DESC_SIZE + l3pad_bytes);
+
+	ath11k_dp_rx_h_ppdu(ar, desc, status);
+
+	status->flag |= (RX_FLAG_MMIC_STRIPPED | RX_FLAG_MMIC_ERROR |
+			 RX_FLAG_DECRYPTED);
+
+	ath11k_dp_rx_h_undecap(ar, msdu, desc,
+			       HAL_ENCRYPT_TYPE_TKIP_MIC, status, false);
+}
+
+static bool ath11k_dp_rx_h_rxdma_err(struct ath11k *ar,  struct sk_buff *msdu,
+				     struct ieee80211_rx_status *status)
+{
+	struct ath11k_skb_rxcb *rxcb = ATH11K_SKB_RXCB(msdu);
+	bool drop = false;
+
+	ar->ab->soc_stats.rxdma_error[rxcb->err_code]++;
+
+	switch (rxcb->err_code) {
+	case HAL_REO_ENTR_RING_RXDMA_ECODE_TKIP_MIC_ERR:
+		ath11k_dp_rx_h_tkip_mic_err(ar, msdu, status);
+		break;
+	default:
+		/* TODO: Review other rxdma error code to check if anything is
+		 * worth reporting to mac80211
+		 */
+		drop = true;
+		break;
+	}
+
+	return drop;
+}
+
+static void ath11k_dp_rx_wbm_err(struct ath11k *ar,
+				 struct napi_struct *napi,
+				 struct sk_buff *msdu,
+				 struct sk_buff_head *msdu_list)
+{
+	struct ath11k_skb_rxcb *rxcb = ATH11K_SKB_RXCB(msdu);
+	struct ieee80211_rx_status rxs = {0};
+	struct ieee80211_rx_status *status;
+	bool drop = true;
+
+	switch (rxcb->err_rel_src) {
+	case HAL_WBM_REL_SRC_MODULE_REO:
+		drop = ath11k_dp_rx_h_reo_err(ar, msdu, &rxs, msdu_list);
+		break;
+	case HAL_WBM_REL_SRC_MODULE_RXDMA:
+		drop = ath11k_dp_rx_h_rxdma_err(ar, msdu, &rxs);
+		break;
+	default:
+		/* msdu will get freed */
+		break;
+	}
+
+	if (drop) {
+		dev_kfree_skb_any(msdu);
+		return;
+	}
+
+	status = IEEE80211_SKB_RXCB(msdu);
+	*status = rxs;
+
+	ath11k_dp_rx_deliver_msdu(ar, napi, msdu);
+}
+
+int ath11k_dp_rx_process_wbm_err(struct ath11k_base *ab,
+				 struct napi_struct *napi, int budget)
+{
+	struct ath11k *ar;
+	struct ath11k_dp *dp = &ab->dp;
+	struct dp_rxdma_ring *rx_ring;
+	struct hal_rx_wbm_rel_info err_info;
+	struct hal_srng *srng;
+	struct sk_buff *msdu;
+	struct sk_buff_head msdu_list[MAX_RADIOS];
+	struct ath11k_skb_rxcb *rxcb;
+	u32 *rx_desc;
+	int buf_id, mac_id;
+	int num_buffs_reaped[MAX_RADIOS] = {0};
+	int total_num_buffs_reaped = 0;
+	int ret, i;
+
+	for (i = 0; i < MAX_RADIOS; i++)
+		__skb_queue_head_init(&msdu_list[i]);
+
+	srng = &ab->hal.srng_list[dp->rx_rel_ring.ring_id];
+
+	spin_lock_bh(&srng->lock);
+
+	ath11k_hal_srng_access_begin(ab, srng);
+
+	while (budget) {
+		rx_desc = ath11k_hal_srng_dst_get_next_entry(ab, srng);
+		if (!rx_desc)
+			break;
+
+		ret = ath11k_hal_wbm_desc_parse_err(ab, rx_desc, &err_info);
+		if (ret) {
+			ath11k_warn(ab,
+				    "failed to parse rx error in wbm_rel ring desc %d\n",
+				    ret);
+			continue;
+		}
+
+		buf_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_BUF_ID, err_info.cookie);
+		mac_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_PDEV_ID, err_info.cookie);
+
+		ar = ab->pdevs[mac_id].ar;
+		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) {
+			ath11k_warn(ab, "frame rx with invalid buf_id %d pdev %d\n",
+				    buf_id, mac_id);
+			spin_unlock_bh(&rx_ring->idr_lock);
+			continue;
+		}
+
+		idr_remove(&rx_ring->bufs_idr, buf_id);
+		spin_unlock_bh(&rx_ring->idr_lock);
+
+		rxcb = ATH11K_SKB_RXCB(msdu);
+		dma_unmap_single(ab->dev, rxcb->paddr,
+				 msdu->len + skb_tailroom(msdu),
+				 DMA_FROM_DEVICE);
+
+		num_buffs_reaped[mac_id]++;
+		total_num_buffs_reaped++;
+		budget--;
+
+		if (err_info.push_reason !=
+		    HAL_REO_DEST_RING_PUSH_REASON_ERR_DETECTED) {
+			dev_kfree_skb_any(msdu);
+			continue;
+		}
+
+		rxcb->err_rel_src = err_info.err_rel_src;
+		rxcb->err_code = err_info.err_code;
+		rxcb->rx_desc = (struct hal_rx_desc *)msdu->data;
+		__skb_queue_tail(&msdu_list[mac_id], msdu);
+	}
+
+	ath11k_hal_srng_access_end(ab, srng);
+
+	spin_unlock_bh(&srng->lock);
+
+	if (!total_num_buffs_reaped)
+		goto done;
+
+	for (i = 0; i <  ab->num_radios; i++) {
+		if (!num_buffs_reaped[i])
+			continue;
+
+		ar = ab->pdevs[i].ar;
+		rx_ring = &ar->dp.rx_refill_buf_ring;
+
+		ath11k_dp_rxbufs_replenish(ab, i, rx_ring, num_buffs_reaped[i],
+					   HAL_RX_BUF_RBM_SW3_BM, GFP_ATOMIC);
+	}
+
+	rcu_read_lock();
+	for (i = 0; i <  ab->num_radios; i++) {
+		if (!rcu_dereference(ab->pdevs_active[i])) {
+			__skb_queue_purge(&msdu_list[i]);
+			continue;
+		}
+
+		ar = ab->pdevs[i].ar;
+
+		if (test_bit(ATH11K_CAC_RUNNING, &ar->dev_flags)) {
+			__skb_queue_purge(&msdu_list[i]);
+			continue;
+		}
+
+		while ((msdu = __skb_dequeue(&msdu_list[i])) != NULL)
+			ath11k_dp_rx_wbm_err(ar, napi, msdu, &msdu_list[i]);
+	}
+	rcu_read_unlock();
+done:
+	return total_num_buffs_reaped;
+}
+
+int ath11k_dp_process_rxdma_err(struct ath11k_base *ab, int mac_id, int budget)
+{
+	struct ath11k *ar = ab->pdevs[mac_id].ar;
+	struct dp_srng *err_ring = &ar->dp.rxdma_err_dst_ring;
+	struct dp_rxdma_ring *rx_ring = &ar->dp.rx_refill_buf_ring;
+	struct dp_link_desc_bank *link_desc_banks = ab->dp.link_desc_banks;
+	struct hal_srng *srng;
+	struct hal_rx_msdu_meta meta[HAL_NUM_RX_MSDUS_PER_LINK_DESC];
+	enum hal_rx_buf_return_buf_manager rbm;
+	enum hal_reo_entr_rxdma_ecode rxdma_err_code;
+	struct ath11k_skb_rxcb *rxcb;
+	struct sk_buff *skb;
+	struct hal_reo_entrance_ring *entr_ring;
+	void *desc;
+	int num_buf_freed = 0;
+	int quota = budget;
+	dma_addr_t paddr;
+	u32 desc_bank;
+	void *link_desc_va;
+	int num_msdus;
+	int i;
+	int buf_id;
+
+	srng = &ab->hal.srng_list[err_ring->ring_id];
+
+	spin_lock_bh(&srng->lock);
+
+	ath11k_hal_srng_access_begin(ab, srng);
+
+	while (quota-- &&
+	       (desc = ath11k_hal_srng_dst_get_next_entry(ab, srng))) {
+		ath11k_hal_rx_reo_ent_paddr_get(ab, desc, &paddr, &desc_bank);
+
+		entr_ring = (struct hal_reo_entrance_ring *)desc;
+		rxdma_err_code =
+			FIELD_GET(HAL_REO_ENTR_RING_INFO1_RXDMA_ERROR_CODE,
+				  entr_ring->info1);
+		ab->soc_stats.rxdma_error[rxdma_err_code]++;
+
+		link_desc_va = link_desc_banks[desc_bank].vaddr +
+			       (paddr - link_desc_banks[desc_bank].paddr);
+		ath11k_hal_rx_msdu_link_info_get(link_desc_va, &num_msdus, meta,
+						 &rbm);
+
+		for (i = 0; i < num_msdus; i++) {
+			buf_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_BUF_ID,
+					   meta[i].cookie);
+
+			spin_lock_bh(&rx_ring->idr_lock);
+			skb = idr_find(&rx_ring->bufs_idr, buf_id);
+			if (!skb) {
+				ath11k_warn(ab, "rxdma error with invalid buf_id %d\n",
+					    buf_id);
+				spin_unlock_bh(&rx_ring->idr_lock);
+				continue;
+			}
+
+			idr_remove(&rx_ring->bufs_idr, buf_id);
+			spin_unlock_bh(&rx_ring->idr_lock);
+
+			rxcb = ATH11K_SKB_RXCB(skb);
+			dma_unmap_single(ab->dev, rxcb->paddr,
+					 skb->len + skb_tailroom(skb),
+					 DMA_FROM_DEVICE);
+			dev_kfree_skb_any(skb);
+
+			num_buf_freed++;
+		}
+
+		ath11k_dp_rx_link_desc_return(ab, desc,
+					      HAL_WBM_REL_BM_ACT_PUT_IN_IDLE);
+	}
+
+	ath11k_hal_srng_access_end(ab, srng);
+
+	spin_unlock_bh(&srng->lock);
+
+	if (num_buf_freed)
+		ath11k_dp_rxbufs_replenish(ab, mac_id, rx_ring, num_buf_freed,
+					   HAL_RX_BUF_RBM_SW3_BM, GFP_ATOMIC);
+
+	return budget - quota;
+}
+
+void ath11k_dp_process_reo_status(struct ath11k_base *ab)
+{
+	struct ath11k_dp *dp = &ab->dp;
+	struct hal_srng *srng;
+	struct dp_reo_cmd *cmd, *tmp;
+	bool found = false;
+	u32 *reo_desc;
+	u16 tag;
+	struct hal_reo_status reo_status;
+
+	srng = &ab->hal.srng_list[dp->reo_status_ring.ring_id];
+
+	memset(&reo_status, 0, sizeof(reo_status));
+
+	spin_lock_bh(&srng->lock);
+
+	ath11k_hal_srng_access_begin(ab, srng);
+
+	while ((reo_desc = ath11k_hal_srng_dst_get_next_entry(ab, srng))) {
+		tag = FIELD_GET(HAL_SRNG_TLV_HDR_TAG, *reo_desc);
+
+		switch (tag) {
+		case HAL_REO_GET_QUEUE_STATS_STATUS:
+			ath11k_hal_reo_status_queue_stats(ab, reo_desc,
+							  &reo_status);
+			break;
+		case HAL_REO_FLUSH_QUEUE_STATUS:
+			ath11k_hal_reo_flush_queue_status(ab, reo_desc,
+							  &reo_status);
+			break;
+		case HAL_REO_FLUSH_CACHE_STATUS:
+			ath11k_hal_reo_flush_cache_status(ab, reo_desc,
+							  &reo_status);
+			break;
+		case HAL_REO_UNBLOCK_CACHE_STATUS:
+			ath11k_hal_reo_unblk_cache_status(ab, reo_desc,
+							  &reo_status);
+			break;
+		case HAL_REO_FLUSH_TIMEOUT_LIST_STATUS:
+			ath11k_hal_reo_flush_timeout_list_status(ab, reo_desc,
+								 &reo_status);
+			break;
+		case HAL_REO_DESCRIPTOR_THRESHOLD_REACHED_STATUS:
+			ath11k_hal_reo_desc_thresh_reached_status(ab, reo_desc,
+								  &reo_status);
+			break;
+		case HAL_REO_UPDATE_RX_REO_QUEUE_STATUS:
+			ath11k_hal_reo_update_rx_reo_queue_status(ab, reo_desc,
+								  &reo_status);
+			break;
+		default:
+			ath11k_warn(ab, "Unknown reo status type %d\n", tag);
+			continue;
+		}
+
+		spin_lock_bh(&dp->reo_cmd_lock);
+		list_for_each_entry_safe(cmd, tmp, &dp->reo_cmd_list, list) {
+			if (reo_status.uniform_hdr.cmd_num == cmd->cmd_num) {
+				found = true;
+				list_del(&cmd->list);
+				break;
+			}
+		}
+		spin_unlock_bh(&dp->reo_cmd_lock);
+
+		if (found) {
+			cmd->handler(dp, (void *)&cmd->data,
+				     reo_status.uniform_hdr.cmd_status);
+			kfree(cmd);
+		}
+
+		found = false;
+	}
+
+	ath11k_hal_srng_access_end(ab, srng);
+
+	spin_unlock_bh(&srng->lock);
+}
+
+void ath11k_dp_rx_pdev_free(struct ath11k_base *ab, int mac_id)
+{
+	struct ath11k *ar = ab->pdevs[mac_id].ar;
+
+	ath11k_dp_rx_pdev_srng_free(ar);
+	ath11k_dp_rxdma_pdev_buf_free(ar);
+}
+
+int ath11k_dp_rx_pdev_alloc(struct ath11k_base *ab, int mac_id)
+{
+	struct ath11k *ar = ab->pdevs[mac_id].ar;
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	u32 ring_id;
+	int ret;
+
+	ret = ath11k_dp_rx_pdev_srng_alloc(ar);
+	if (ret) {
+		ath11k_warn(ab, "failed to setup rx srngs\n");
+		return ret;
+	}
+
+	ret = ath11k_dp_rxdma_pdev_buf_setup(ar);
+	if (ret) {
+		ath11k_warn(ab, "failed to setup rxdma ring\n");
+		return ret;
+	}
+
+	ring_id = dp->rx_refill_buf_ring.refill_buf_ring.ring_id;
+	ret = ath11k_dp_tx_htt_srng_setup(ab, ring_id, mac_id, HAL_RXDMA_BUF);
+	if (ret) {
+		ath11k_warn(ab, "failed to configure rx_refill_buf_ring %d\n",
+			    ret);
+		return ret;
+	}
+
+	ring_id = dp->rxdma_err_dst_ring.ring_id;
+	ret = ath11k_dp_tx_htt_srng_setup(ab, ring_id, mac_id, HAL_RXDMA_DST);
+	if (ret) {
+		ath11k_warn(ab, "failed to configure rxdma_err_dest_ring %d\n",
+			    ret);
+		return ret;
+	}
+
+	ring_id = dp->rxdma_mon_buf_ring.refill_buf_ring.ring_id;
+	ret = ath11k_dp_tx_htt_srng_setup(ab, ring_id,
+					  mac_id, HAL_RXDMA_MONITOR_BUF);
+	if (ret) {
+		ath11k_warn(ab, "failed to configure rxdma_mon_buf_ring %d\n",
+			    ret);
+		return ret;
+	}
+	ret = ath11k_dp_tx_htt_srng_setup(ab,
+					  dp->rxdma_mon_dst_ring.ring_id,
+					  mac_id, HAL_RXDMA_MONITOR_DST);
+	if (ret) {
+		ath11k_warn(ab, "failed to configure rxdma_mon_dst_ring %d\n",
+			    ret);
+		return ret;
+	}
+	ret = ath11k_dp_tx_htt_srng_setup(ab,
+					  dp->rxdma_mon_desc_ring.ring_id,
+					  mac_id, HAL_RXDMA_MONITOR_DESC);
+	if (ret) {
+		ath11k_warn(ab, "failed to configure rxdma_mon_dst_ring %d\n",
+			    ret);
+		return ret;
+	}
+	ring_id = dp->rx_mon_status_refill_ring.refill_buf_ring.ring_id;
+	ret = ath11k_dp_tx_htt_srng_setup(ab, ring_id, mac_id,
+					  HAL_RXDMA_MONITOR_STATUS);
+	if (ret) {
+		ath11k_warn(ab,
+			    "failed to configure mon_status_refill_ring %d\n",
+			    ret);
+		return ret;
+	}
+	return 0;
+}
+
+static void ath11k_dp_mon_set_frag_len(u32 *total_len, u32 *frag_len)
+{
+	if (*total_len >= (DP_RX_BUFFER_SIZE - sizeof(struct hal_rx_desc))) {
+		*frag_len = DP_RX_BUFFER_SIZE - sizeof(struct hal_rx_desc);
+		*total_len -= *frag_len;
+	} else {
+		*frag_len = *total_len;
+		*total_len = 0;
+	}
+}
+
+static
+int ath11k_dp_rx_monitor_link_desc_return(struct ath11k *ar,
+					  void *p_last_buf_addr_info,
+					  u8 mac_id)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct dp_srng *dp_srng;
+	void *hal_srng;
+	void *src_srng_desc;
+	int ret = 0;
+
+	dp_srng = &dp->rxdma_mon_desc_ring;
+	hal_srng = &ar->ab->hal.srng_list[dp_srng->ring_id];
+
+	ath11k_hal_srng_access_begin(ar->ab, hal_srng);
+
+	src_srng_desc = ath11k_hal_srng_src_get_next_entry(ar->ab, hal_srng);
+
+	if (src_srng_desc) {
+		struct ath11k_buffer_addr *src_desc =
+				(struct ath11k_buffer_addr *)src_srng_desc;
+
+		*src_desc = *((struct ath11k_buffer_addr *)p_last_buf_addr_info);
+	} else {
+		ath11k_dbg(ar->ab, ATH11K_DBG_DATA,
+			   "Monitor Link Desc Ring %d Full", mac_id);
+		ret = -ENOMEM;
+	}
+
+	ath11k_hal_srng_access_end(ar->ab, hal_srng);
+	return ret;
+}
+
+static
+void ath11k_dp_rx_mon_next_link_desc_get(void *rx_msdu_link_desc,
+					 dma_addr_t *paddr, u32 *sw_cookie,
+					 void **pp_buf_addr_info)
+{
+	struct hal_rx_msdu_link *msdu_link =
+			(struct hal_rx_msdu_link *)rx_msdu_link_desc;
+	struct ath11k_buffer_addr *buf_addr_info;
+	u8 rbm = 0;
+
+	buf_addr_info = (struct ath11k_buffer_addr *)&msdu_link->buf_addr_info;
+
+	ath11k_hal_rx_buf_addr_info_get(buf_addr_info, paddr, sw_cookie, &rbm);
+
+	*pp_buf_addr_info = (void *)buf_addr_info;
+}
+
+static int ath11k_dp_pkt_set_pktlen(struct sk_buff *skb, u32 len)
+{
+	if (skb->len > len) {
+		skb_trim(skb, len);
+	} else {
+		if (skb_tailroom(skb) < len - skb->len) {
+			if ((pskb_expand_head(skb, 0,
+					      len - skb->len - skb_tailroom(skb),
+					      GFP_ATOMIC))) {
+				dev_kfree_skb_any(skb);
+				return -ENOMEM;
+			}
+		}
+		skb_put(skb, (len - skb->len));
+	}
+	return 0;
+}
+
+static void ath11k_hal_rx_msdu_list_get(struct ath11k *ar,
+					void *msdu_link_desc,
+					struct hal_rx_msdu_list *msdu_list,
+					u16 *num_msdus)
+{
+	struct hal_rx_msdu_details *msdu_details = NULL;
+	struct rx_msdu_desc *msdu_desc_info = NULL;
+	struct hal_rx_msdu_link *msdu_link = NULL;
+	int i;
+	u32 last = FIELD_PREP(RX_MSDU_DESC_INFO0_LAST_MSDU_IN_MPDU, 1);
+	u32 first = FIELD_PREP(RX_MSDU_DESC_INFO0_FIRST_MSDU_IN_MPDU, 1);
+	u8  tmp  = 0;
+
+	msdu_link = (struct hal_rx_msdu_link *)msdu_link_desc;
+	msdu_details = &msdu_link->msdu_link[0];
+
+	for (i = 0; i < HAL_RX_NUM_MSDU_DESC; i++) {
+		if (FIELD_GET(BUFFER_ADDR_INFO0_ADDR,
+			      msdu_details[i].buf_addr_info.info0) == 0) {
+			msdu_desc_info = &msdu_details[i - 1].rx_msdu_info;
+			msdu_desc_info->info0 |= last;
+			;
+			break;
+		}
+		msdu_desc_info = &msdu_details[i].rx_msdu_info;
+
+		if (!i)
+			msdu_desc_info->info0 |= first;
+		else if (i == (HAL_RX_NUM_MSDU_DESC - 1))
+			msdu_desc_info->info0 |= last;
+		msdu_list->msdu_info[i].msdu_flags = msdu_desc_info->info0;
+		msdu_list->msdu_info[i].msdu_len =
+			 HAL_RX_MSDU_PKT_LENGTH_GET(msdu_desc_info->info0);
+		msdu_list->sw_cookie[i] =
+			FIELD_GET(BUFFER_ADDR_INFO1_SW_COOKIE,
+				  msdu_details[i].buf_addr_info.info1);
+		tmp = FIELD_GET(BUFFER_ADDR_INFO1_RET_BUF_MGR,
+				msdu_details[i].buf_addr_info.info1);
+		msdu_list->rbm[i] = tmp;
+	}
+	*num_msdus = i;
+}
+
+static u32 ath11k_dp_rx_mon_comp_ppduid(u32 msdu_ppdu_id, u32 *ppdu_id,
+					u32 *rx_bufs_used)
+{
+	u32 ret = 0;
+
+	if ((*ppdu_id < msdu_ppdu_id) &&
+	    ((msdu_ppdu_id - *ppdu_id) < DP_NOT_PPDU_ID_WRAP_AROUND)) {
+		*ppdu_id = msdu_ppdu_id;
+		ret = msdu_ppdu_id;
+	} else if ((*ppdu_id > msdu_ppdu_id) &&
+		((*ppdu_id - msdu_ppdu_id) > DP_NOT_PPDU_ID_WRAP_AROUND)) {
+		/* mon_dst is behind than mon_status
+		 * skip dst_ring and free it
+		 */
+		*rx_bufs_used += 1;
+		*ppdu_id = msdu_ppdu_id;
+		ret = msdu_ppdu_id;
+	}
+	return ret;
+}
+
+static void ath11k_dp_mon_get_buf_len(struct hal_rx_msdu_desc_info *info,
+				      bool *is_frag, u32 *total_len,
+				      u32 *frag_len, u32 *msdu_cnt)
+{
+	if (info->msdu_flags & RX_MSDU_DESC_INFO0_MSDU_CONTINUATION) {
+		if (!*is_frag) {
+			*total_len = info->msdu_len;
+			*is_frag = true;
+		}
+		ath11k_dp_mon_set_frag_len(total_len,
+					   frag_len);
+	} else {
+		if (*is_frag) {
+			ath11k_dp_mon_set_frag_len(total_len,
+						   frag_len);
+		} else {
+			*frag_len = info->msdu_len;
+		}
+		*is_frag = false;
+		*msdu_cnt -= 1;
+	}
+}
+
+static u32
+ath11k_dp_rx_mon_mpdu_pop(struct ath11k *ar,
+			  void *ring_entry, struct sk_buff **head_msdu,
+			  struct sk_buff **tail_msdu, u32 *npackets,
+			  u32 *ppdu_id)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct ath11k_mon_data *pmon = (struct ath11k_mon_data *)&dp->mon_data;
+	struct dp_rxdma_ring *rx_ring = &dp->rxdma_mon_buf_ring;
+	struct sk_buff *msdu = NULL, *last = NULL;
+	struct hal_rx_msdu_list msdu_list;
+	void *p_buf_addr_info, *p_last_buf_addr_info;
+	struct hal_rx_desc *rx_desc;
+	void *rx_msdu_link_desc;
+	dma_addr_t paddr;
+	u16 num_msdus = 0;
+	u32 rx_buf_size, rx_pkt_offset, sw_cookie;
+	u32 rx_bufs_used = 0, i = 0;
+	u32 msdu_ppdu_id = 0, msdu_cnt = 0;
+	u32 total_len = 0, frag_len = 0;
+	bool is_frag, is_first_msdu;
+	bool drop_mpdu = false;
+	struct ath11k_skb_rxcb *rxcb;
+	struct hal_reo_entrance_ring *ent_desc =
+			(struct hal_reo_entrance_ring *)ring_entry;
+	int buf_id;
+
+	ath11k_hal_rx_reo_ent_buf_paddr_get(ring_entry, &paddr,
+					    &sw_cookie, &p_last_buf_addr_info,
+					    &msdu_cnt);
+
+	if (FIELD_GET(HAL_REO_ENTR_RING_INFO1_RXDMA_PUSH_REASON,
+		      ent_desc->info1) ==
+		      HAL_REO_DEST_RING_PUSH_REASON_ERR_DETECTED) {
+		u8 rxdma_err =
+			FIELD_GET(HAL_REO_ENTR_RING_INFO1_RXDMA_ERROR_CODE,
+				  ent_desc->info1);
+		if (rxdma_err == HAL_REO_ENTR_RING_RXDMA_ECODE_FLUSH_REQUEST_ERR ||
+		    rxdma_err == HAL_REO_ENTR_RING_RXDMA_ECODE_MPDU_LEN_ERR ||
+		    rxdma_err == HAL_REO_ENTR_RING_RXDMA_ECODE_OVERFLOW_ERR) {
+			drop_mpdu = true;
+			pmon->rx_mon_stats.dest_mpdu_drop++;
+		}
+	}
+
+	is_frag = false;
+	is_first_msdu = true;
+
+	do {
+		if (pmon->mon_last_linkdesc_paddr == paddr) {
+			pmon->rx_mon_stats.dup_mon_linkdesc_cnt++;
+			return rx_bufs_used;
+		}
+
+		rx_msdu_link_desc =
+			(void *)pmon->link_desc_banks[sw_cookie].vaddr +
+			(paddr - pmon->link_desc_banks[sw_cookie].paddr);
+
+		ath11k_hal_rx_msdu_list_get(ar, rx_msdu_link_desc, &msdu_list,
+					    &num_msdus);
+
+		for (i = 0; i < num_msdus; i++) {
+			u32 l2_hdr_offset;
+
+			if (pmon->mon_last_buf_cookie == msdu_list.sw_cookie[i]) {
+				ath11k_dbg(ar->ab, ATH11K_DBG_DATA,
+					   "i %d last_cookie %d is same\n",
+					   i, pmon->mon_last_buf_cookie);
+				drop_mpdu = true;
+				pmon->rx_mon_stats.dup_mon_buf_cnt++;
+				continue;
+			}
+			buf_id = FIELD_GET(DP_RXDMA_BUF_COOKIE_BUF_ID,
+					   msdu_list.sw_cookie[i]);
+
+			spin_lock_bh(&rx_ring->idr_lock);
+			msdu = idr_find(&rx_ring->bufs_idr, buf_id);
+			spin_unlock_bh(&rx_ring->idr_lock);
+			if (!msdu) {
+				ath11k_dbg(ar->ab, ATH11K_DBG_DATA,
+					   "msdu_pop: invalid buf_id %d\n", buf_id);
+				break;
+			}
+			rxcb = ATH11K_SKB_RXCB(msdu);
+			if (!rxcb->unmapped) {
+				dma_unmap_single(ar->ab->dev, rxcb->paddr,
+						 msdu->len +
+						 skb_tailroom(msdu),
+						 DMA_FROM_DEVICE);
+				rxcb->unmapped = 1;
+			}
+			if (drop_mpdu) {
+				ath11k_dbg(ar->ab, ATH11K_DBG_DATA,
+					   "i %d drop msdu %p *ppdu_id %x\n",
+					   i, msdu, *ppdu_id);
+				dev_kfree_skb_any(msdu);
+				msdu = NULL;
+				goto next_msdu;
+			}
+
+			rx_desc = (struct hal_rx_desc *)msdu->data;
+
+			rx_pkt_offset = sizeof(struct hal_rx_desc);
+			l2_hdr_offset = ath11k_dp_rx_h_msdu_end_l3pad(rx_desc);
+
+			if (is_first_msdu) {
+				if (!ath11k_dp_rxdesc_mpdu_valid(rx_desc)) {
+					drop_mpdu = true;
+					dev_kfree_skb_any(msdu);
+					msdu = NULL;
+					pmon->mon_last_linkdesc_paddr = paddr;
+					goto next_msdu;
+				}
+
+				msdu_ppdu_id =
+					ath11k_dp_rxdesc_get_ppduid(rx_desc);
+
+				if (ath11k_dp_rx_mon_comp_ppduid(msdu_ppdu_id,
+								 ppdu_id,
+								 &rx_bufs_used))
+					return rx_bufs_used;
+				pmon->mon_last_linkdesc_paddr = paddr;
+				is_first_msdu = false;
+			}
+			ath11k_dp_mon_get_buf_len(&msdu_list.msdu_info[i],
+						  &is_frag, &total_len,
+						  &frag_len, &msdu_cnt);
+			rx_buf_size = rx_pkt_offset + l2_hdr_offset + frag_len;
+
+			ath11k_dp_pkt_set_pktlen(msdu, rx_buf_size);
+
+			if (!(*head_msdu))
+				*head_msdu = msdu;
+			else if (last)
+				last->next = msdu;
+
+			last = msdu;
+next_msdu:
+			pmon->mon_last_buf_cookie = msdu_list.sw_cookie[i];
+			rx_bufs_used++;
+			spin_lock_bh(&rx_ring->idr_lock);
+			idr_remove(&rx_ring->bufs_idr, buf_id);
+			spin_unlock_bh(&rx_ring->idr_lock);
+		}
+
+		ath11k_dp_rx_mon_next_link_desc_get(rx_msdu_link_desc, &paddr,
+						    &sw_cookie,
+						    &p_buf_addr_info);
+
+		if (ath11k_dp_rx_monitor_link_desc_return(ar,
+							  p_last_buf_addr_info,
+							  dp->mac_id))
+			ath11k_dbg(ar->ab, ATH11K_DBG_DATA,
+				   "dp_rx_monitor_link_desc_return failed");
+
+		p_last_buf_addr_info = p_buf_addr_info;
+
+	} while (paddr && msdu_cnt);
+
+	if (last)
+		last->next = NULL;
+
+	*tail_msdu = msdu;
+
+	if (msdu_cnt == 0)
+		*npackets = 1;
+
+	return rx_bufs_used;
+}
+
+static void ath11k_dp_rx_msdus_set_payload(struct sk_buff *msdu)
+{
+	u32 rx_pkt_offset, l2_hdr_offset;
+
+	rx_pkt_offset = sizeof(struct hal_rx_desc);
+	l2_hdr_offset = ath11k_dp_rx_h_msdu_end_l3pad((struct hal_rx_desc *)msdu->data);
+	skb_pull(msdu, rx_pkt_offset + l2_hdr_offset);
+}
+
+static struct sk_buff *
+ath11k_dp_rx_mon_merg_msdus(struct ath11k *ar,
+			    u32 mac_id, struct sk_buff *head_msdu,
+			    struct sk_buff *last_msdu,
+			    struct ieee80211_rx_status *rxs)
+{
+	struct sk_buff *msdu, *mpdu_buf, *prev_buf;
+	u32 decap_format, wifi_hdr_len;
+	struct hal_rx_desc *rx_desc;
+	char *hdr_desc;
+	u8 *dest;
+	struct ieee80211_hdr_3addr *wh;
+
+	mpdu_buf = NULL;
+
+	if (!head_msdu)
+		goto err_merge_fail;
+
+	rx_desc = (struct hal_rx_desc *)head_msdu->data;
+
+	if (ath11k_dp_rxdesc_get_mpdulen_err(rx_desc))
+		return NULL;
+
+	decap_format = ath11k_dp_rxdesc_get_decap_format(rx_desc);
+
+	ath11k_dp_rx_h_ppdu(ar, rx_desc, rxs);
+
+	if (decap_format == DP_RX_DECAP_TYPE_RAW) {
+		ath11k_dp_rx_msdus_set_payload(head_msdu);
+
+		prev_buf = head_msdu;
+		msdu = head_msdu->next;
+
+		while (msdu) {
+			ath11k_dp_rx_msdus_set_payload(msdu);
+
+			prev_buf = msdu;
+			msdu = msdu->next;
+		}
+
+		prev_buf->next = NULL;
+
+		skb_trim(prev_buf, prev_buf->len - HAL_RX_FCS_LEN);
+	} else if (decap_format == DP_RX_DECAP_TYPE_NATIVE_WIFI) {
+		__le16 qos_field;
+		u8 qos_pkt = 0;
+
+		rx_desc = (struct hal_rx_desc *)head_msdu->data;
+		hdr_desc = ath11k_dp_rxdesc_get_80211hdr(rx_desc);
+
+		/* Base size */
+		wifi_hdr_len = sizeof(struct ieee80211_hdr_3addr);
+		wh = (struct ieee80211_hdr_3addr *)hdr_desc;
+
+		if (ieee80211_is_data_qos(wh->frame_control)) {
+			struct ieee80211_qos_hdr *qwh =
+					(struct ieee80211_qos_hdr *)hdr_desc;
+
+			qos_field = qwh->qos_ctrl;
+			qos_pkt = 1;
+		}
+		msdu = head_msdu;
+
+		while (msdu) {
+			rx_desc = (struct hal_rx_desc *)msdu->data;
+			hdr_desc = ath11k_dp_rxdesc_get_80211hdr(rx_desc);
+
+			if (qos_pkt) {
+				dest = skb_push(msdu, sizeof(__le16));
+				if (!dest)
+					goto err_merge_fail;
+				memcpy(dest, hdr_desc, wifi_hdr_len);
+				memcpy(dest + wifi_hdr_len,
+				       (u8 *)&qos_field, sizeof(__le16));
+			}
+			ath11k_dp_rx_msdus_set_payload(msdu);
+			prev_buf = msdu;
+			msdu = msdu->next;
+		}
+		dest = skb_put(prev_buf, HAL_RX_FCS_LEN);
+		if (!dest)
+			goto err_merge_fail;
+
+		ath11k_dbg(ar->ab, ATH11K_DBG_DATA,
+			   "mpdu_buf %pK mpdu_buf->len %u",
+			   prev_buf, prev_buf->len);
+	} else {
+		ath11k_dbg(ar->ab, ATH11K_DBG_DATA,
+			   "decap format %d is not supported!\n",
+			   decap_format);
+		goto err_merge_fail;
+	}
+
+	return head_msdu;
+
+err_merge_fail:
+	if (mpdu_buf && decap_format != DP_RX_DECAP_TYPE_RAW) {
+		ath11k_dbg(ar->ab, ATH11K_DBG_DATA,
+			   "err_merge_fail mpdu_buf %pK", mpdu_buf);
+		/* Free the head buffer */
+		dev_kfree_skb_any(mpdu_buf);
+	}
+	return NULL;
+}
+
+static int ath11k_dp_rx_mon_deliver(struct ath11k *ar, u32 mac_id,
+				    struct sk_buff *head_msdu,
+				    struct sk_buff *tail_msdu,
+				    struct napi_struct *napi)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct sk_buff *mon_skb, *skb_next, *header;
+	struct ieee80211_rx_status *rxs = &dp->rx_status, *status;
+
+	mon_skb = ath11k_dp_rx_mon_merg_msdus(ar, mac_id, head_msdu,
+					      tail_msdu, rxs);
+
+	if (!mon_skb)
+		goto mon_deliver_fail;
+
+	header = mon_skb;
+
+	rxs->flag = 0;
+	do {
+		skb_next = mon_skb->next;
+		if (!skb_next)
+			rxs->flag &= ~RX_FLAG_AMSDU_MORE;
+		else
+			rxs->flag |= RX_FLAG_AMSDU_MORE;
+
+		if (mon_skb == header) {
+			header = NULL;
+			rxs->flag &= ~RX_FLAG_ALLOW_SAME_PN;
+		} else {
+			rxs->flag |= RX_FLAG_ALLOW_SAME_PN;
+		}
+		rxs->flag |= RX_FLAG_ONLY_MONITOR;
+
+		status = IEEE80211_SKB_RXCB(mon_skb);
+		*status = *rxs;
+
+		ath11k_dp_rx_deliver_msdu(ar, napi, mon_skb);
+		mon_skb = skb_next;
+	} while (mon_skb && (mon_skb != tail_msdu));
+	rxs->flag = 0;
+
+	return 0;
+
+mon_deliver_fail:
+	mon_skb = head_msdu;
+	while (mon_skb) {
+		skb_next = mon_skb->next;
+		dev_kfree_skb_any(mon_skb);
+		mon_skb = skb_next;
+	}
+	return -EINVAL;
+}
+
+static void ath11k_dp_rx_mon_dest_process(struct ath11k *ar, u32 quota,
+					  struct napi_struct *napi)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct ath11k_mon_data *pmon = (struct ath11k_mon_data *)&dp->mon_data;
+	void *ring_entry;
+	void *mon_dst_srng;
+	u32 ppdu_id;
+	u32 rx_bufs_used;
+	struct ath11k_pdev_mon_stats *rx_mon_stats;
+	u32	 npackets = 0;
+
+	mon_dst_srng = &ar->ab->hal.srng_list[dp->rxdma_mon_dst_ring.ring_id];
+
+	if (!mon_dst_srng) {
+		ath11k_warn(ar->ab,
+			    "HAL Monitor Destination Ring Init Failed -- %pK",
+			    mon_dst_srng);
+		return;
+	}
+
+	spin_lock_bh(&pmon->mon_lock);
+
+	ath11k_hal_srng_access_begin(ar->ab, mon_dst_srng);
+
+	ppdu_id = pmon->mon_ppdu_info.ppdu_id;
+	rx_bufs_used = 0;
+	rx_mon_stats = &pmon->rx_mon_stats;
+
+	while ((ring_entry = ath11k_hal_srng_dst_peek(ar->ab, mon_dst_srng))) {
+		struct sk_buff *head_msdu, *tail_msdu;
+
+		head_msdu = NULL;
+		tail_msdu = NULL;
+
+		rx_bufs_used += ath11k_dp_rx_mon_mpdu_pop(ar, ring_entry,
+							  &head_msdu,
+							  &tail_msdu,
+							  &npackets, &ppdu_id);
+
+		if (ppdu_id != pmon->mon_ppdu_info.ppdu_id) {
+			pmon->mon_ppdu_status = DP_PPDU_STATUS_START;
+			ath11k_dbg(ar->ab, ATH11K_DBG_DATA,
+				   "dest_rx: new ppdu_id %x != status ppdu_id %x",
+				   ppdu_id, pmon->mon_ppdu_info.ppdu_id);
+			break;
+		}
+		if (head_msdu && tail_msdu) {
+			ath11k_dp_rx_mon_deliver(ar, dp->mac_id, head_msdu,
+						 tail_msdu, napi);
+			rx_mon_stats->dest_mpdu_done++;
+		}
+
+		ring_entry = ath11k_hal_srng_dst_get_next_entry(ar->ab,
+								mon_dst_srng);
+	}
+	ath11k_hal_srng_access_end(ar->ab, mon_dst_srng);
+
+	spin_unlock_bh(&pmon->mon_lock);
+
+	if (rx_bufs_used) {
+		rx_mon_stats->dest_ppdu_done++;
+		ath11k_dp_rxbufs_replenish(ar->ab, dp->mac_id,
+					   &dp->rxdma_mon_buf_ring,
+					   rx_bufs_used,
+					   HAL_RX_BUF_RBM_SW3_BM, GFP_ATOMIC);
+	}
+}
+
+static void ath11k_dp_rx_mon_status_process_tlv(struct ath11k *ar,
+						u32 quota,
+						struct napi_struct *napi)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct ath11k_mon_data *pmon = (struct ath11k_mon_data *)&dp->mon_data;
+	struct hal_rx_mon_ppdu_info *ppdu_info;
+	struct sk_buff *status_skb;
+	u32 tlv_status = HAL_TLV_STATUS_BUF_DONE;
+	struct ath11k_pdev_mon_stats *rx_mon_stats;
+
+	ppdu_info = &pmon->mon_ppdu_info;
+	rx_mon_stats = &pmon->rx_mon_stats;
+
+	if (pmon->mon_ppdu_status != DP_PPDU_STATUS_START)
+		return;
+
+	while (!skb_queue_empty(&pmon->rx_status_q)) {
+		status_skb = skb_dequeue(&pmon->rx_status_q);
+
+		tlv_status = ath11k_hal_rx_parse_mon_status(ar->ab, ppdu_info,
+							    status_skb);
+		if (tlv_status == HAL_TLV_STATUS_PPDU_DONE) {
+			rx_mon_stats->status_ppdu_done++;
+			pmon->mon_ppdu_status = DP_PPDU_STATUS_DONE;
+			ath11k_dp_rx_mon_dest_process(ar, quota, napi);
+			pmon->mon_ppdu_status = DP_PPDU_STATUS_START;
+		}
+		dev_kfree_skb_any(status_skb);
+	}
+}
+
+static int ath11k_dp_mon_process_rx(struct ath11k_base *ab, int mac_id,
+				    struct napi_struct *napi, int budget)
+{
+	struct ath11k *ar = ab->pdevs[mac_id].ar;
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct ath11k_mon_data *pmon = (struct ath11k_mon_data *)&dp->mon_data;
+	int num_buffs_reaped = 0;
+
+	num_buffs_reaped = ath11k_dp_rx_reap_mon_status_ring(ar->ab, dp->mac_id, &budget,
+							     &pmon->rx_status_q);
+	if (num_buffs_reaped)
+		ath11k_dp_rx_mon_status_process_tlv(ar, budget, napi);
+
+	return num_buffs_reaped;
+}
+
+int ath11k_dp_rx_process_mon_rings(struct ath11k_base *ab, int mac_id,
+				   struct napi_struct *napi, int budget)
+{
+	struct ath11k *ar = ab->pdevs[mac_id].ar;
+	int ret = 0;
+
+	if (test_bit(ATH11K_FLAG_MONITOR_ENABLED, &ar->monitor_flags))
+		ret = ath11k_dp_mon_process_rx(ab, mac_id, napi, budget);
+	else
+		ret = ath11k_dp_rx_process_mon_status(ab, mac_id, napi, budget);
+	return ret;
+}
+
+static int ath11k_dp_rx_pdev_mon_status_attach(struct ath11k *ar)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct ath11k_mon_data *pmon = (struct ath11k_mon_data *)&dp->mon_data;
+
+	skb_queue_head_init(&pmon->rx_status_q);
+
+	pmon->mon_ppdu_status = DP_PPDU_STATUS_START;
+
+	memset(&pmon->rx_mon_stats, 0,
+	       sizeof(pmon->rx_mon_stats));
+	return 0;
+}
+
+int ath11k_dp_rx_pdev_mon_attach(struct ath11k *ar)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct ath11k_mon_data *pmon = &dp->mon_data;
+	struct hal_srng *mon_desc_srng = NULL;
+	struct dp_srng *dp_srng;
+	int ret = 0;
+	u32 n_link_desc = 0;
+
+	ret = ath11k_dp_rx_pdev_mon_status_attach(ar);
+	if (ret) {
+		ath11k_warn(ar->ab, "pdev_mon_status_attach() failed");
+		return ret;
+	}
+
+	dp_srng = &dp->rxdma_mon_desc_ring;
+	n_link_desc = dp_srng->size /
+		ath11k_hal_srng_get_entrysize(HAL_RXDMA_MONITOR_DESC);
+	mon_desc_srng =
+		&ar->ab->hal.srng_list[dp->rxdma_mon_desc_ring.ring_id];
+
+	ret = ath11k_dp_link_desc_setup(ar->ab, pmon->link_desc_banks,
+					HAL_RXDMA_MONITOR_DESC, mon_desc_srng,
+					n_link_desc);
+	if (ret) {
+		ath11k_warn(ar->ab, "mon_link_desc_pool_setup() failed");
+		return ret;
+	}
+	pmon->mon_last_linkdesc_paddr = 0;
+	pmon->mon_last_buf_cookie = DP_RX_DESC_COOKIE_MAX + 1;
+	spin_lock_init(&pmon->mon_lock);
+	return 0;
+}
+
+static int ath11k_dp_mon_link_free(struct ath11k *ar)
+{
+	struct ath11k_pdev_dp *dp = &ar->dp;
+	struct ath11k_mon_data *pmon = &dp->mon_data;
+
+	ath11k_dp_link_desc_cleanup(ar->ab, pmon->link_desc_banks,
+				    HAL_RXDMA_MONITOR_DESC,
+				    &dp->rxdma_mon_desc_ring);
+	return 0;
+}
+
+int ath11k_dp_rx_pdev_mon_detach(struct ath11k *ar)
+{
+	ath11k_dp_mon_link_free(ar);
+	return 0;
+}





[Index of Archives]     [Device Tree Compilter]     [Device Tree Spec]     [Linux Driver Backports]     [Video for Linux]     [Linux USB Devel]     [Linux PCI Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Yosemite Backpacking]


  Powered by Linux