1 files changed, 238 insertions, 88 deletions

diff --git a/drivers/net/wireless/mediatek/mt76/dma.c b/drivers/net/wireless/mediatek/mt76/dma.c
index 5f46d6daeaa7..f240016ed9f0 100644
--- a/drivers/net/wireless/mediatek/mt76/dma.c
+++ b/drivers/net/wireless/mediatek/mt76/dma.c
@@ -1,4 +1,4 @@
-// SPDX-License-Identifier: ISC
+// SPDX-License-Identifier: BSD-3-Clause-Clear
 /*
  * Copyright (C) 2016 Felix Fietkau <nbd@nbd.name>
  */
@@ -7,37 +7,6 @@
 #include "mt76.h"
 #include "dma.h"
 
-#if IS_ENABLED(CONFIG_NET_MEDIATEK_SOC_WED)
-
-#define Q_READ(_q, _field) ({						\
-	u32 _offset = offsetof(struct mt76_queue_regs, _field);		\
-	u32 _val;							\
-	if ((_q)->flags & MT_QFLAG_WED)					\
-		_val = mtk_wed_device_reg_read((_q)->wed,		\
-					       ((_q)->wed_regs +	\
-					        _offset));		\
-	else								\
-		_val = readl(&(_q)->regs->_field);			\
-	_val;								\
-})
-
-#define Q_WRITE(_q, _field, _val)	do {				\
-	u32 _offset = offsetof(struct mt76_queue_regs, _field);		\
-	if ((_q)->flags & MT_QFLAG_WED)					\
-		mtk_wed_device_reg_write((_q)->wed,			\
-					 ((_q)->wed_regs + _offset),	\
-					 _val);				\
-	else								\
-		writel(_val, &(_q)->regs->_field);			\
-} while (0)
-
-#else
-
-#define Q_READ(_q, _field)		readl(&(_q)->regs->_field)
-#define Q_WRITE(_q, _field, _val)	writel(_val, &(_q)->regs->_field)
-
-#endif
-
 static struct mt76_txwi_cache *
 mt76_alloc_txwi(struct mt76_dev *dev)
 {
@@ -186,24 +155,61 @@ mt76_free_pending_rxwi(struct mt76_dev *dev)
 EXPORT_SYMBOL_GPL(mt76_free_pending_rxwi);
 
 static void
+mt76_dma_queue_magic_cnt_init(struct mt76_dev *dev, struct mt76_queue *q)
+{
+	if (!mt76_queue_is_wed_rro(q))
+		return;
+
+	q->magic_cnt = 0;
+	if (mt76_queue_is_wed_rro_ind(q)) {
+		struct mt76_wed_rro_desc *rro_desc;
+		u32 data1 = FIELD_PREP(RRO_IND_DATA1_MAGIC_CNT_MASK,
+				       MT_DMA_WED_IND_CMD_CNT - 1);
+		int i;
+
+		rro_desc = (struct mt76_wed_rro_desc *)q->desc;
+		for (i = 0; i < q->ndesc; i++) {
+			struct mt76_wed_rro_ind *cmd;
+
+			cmd = (struct mt76_wed_rro_ind *)&rro_desc[i];
+			cmd->data1 = cpu_to_le32(data1);
+		}
+	} else if (mt76_queue_is_wed_rro_rxdmad_c(q)) {
+		struct mt76_rro_rxdmad_c *dmad = (void *)q->desc;
+		u32 data3 = FIELD_PREP(RRO_RXDMAD_DATA3_MAGIC_CNT_MASK,
+				       MT_DMA_MAGIC_CNT - 1);
+		int i;
+
+		for (i = 0; i < q->ndesc; i++)
+			dmad[i].data3 = cpu_to_le32(data3);
+	}
+}
+
+static void
 mt76_dma_sync_idx(struct mt76_dev *dev, struct mt76_queue *q)
 {
 	Q_WRITE(q, desc_base, q->desc_dma);
-	if (q->flags & MT_QFLAG_WED_RRO_EN)
+	if ((q->flags & MT_QFLAG_WED_RRO_EN) && !mt76_npu_device_active(dev))
 		Q_WRITE(q, ring_size, MT_DMA_RRO_EN | q->ndesc);
 	else
 		Q_WRITE(q, ring_size, q->ndesc);
+
+	if (mt76_queue_is_npu_tx(q)) {
+		writel(q->desc_dma, &q->regs->desc_base);
+		writel(q->ndesc, &q->regs->ring_size);
+	}
 	q->head = Q_READ(q, dma_idx);
 	q->tail = q->head;
 }
 
-void __mt76_dma_queue_reset(struct mt76_dev *dev, struct mt76_queue *q,
-			    bool reset_idx)
+void mt76_dma_queue_reset(struct mt76_dev *dev, struct mt76_queue *q,
+			  bool reset_idx)
 {
 	if (!q || !q->ndesc)
 		return;
 
-	if (!mt76_queue_is_wed_rro_ind(q)) {
+	if (!mt76_queue_is_wed_rro_ind(q) &&
+	    !mt76_queue_is_wed_rro_rxdmad_c(q) && !mt76_queue_is_npu(q)) {
 		int i;
 
 		/* clear descriptors */
@@ -211,27 +217,26 @@ void __mt76_dma_queue_reset(struct mt76_dev *dev, struct mt76_queue *q,
 			q->desc[i].ctrl = cpu_to_le32(MT_DMA_CTL_DMA_DONE);
 	}
 
+	mt76_dma_queue_magic_cnt_init(dev, q);
 	if (reset_idx) {
-		Q_WRITE(q, cpu_idx, 0);
+		if (mt76_queue_is_emi(q))
+			*q->emi_cpu_idx = 0;
+		else
+			Q_WRITE(q, cpu_idx, 0);
 		Q_WRITE(q, dma_idx, 0);
 	}
 	mt76_dma_sync_idx(dev, q);
 }
 
-void mt76_dma_queue_reset(struct mt76_dev *dev, struct mt76_queue *q)
-{
-	__mt76_dma_queue_reset(dev, q, true);
-}
-
 static int
 mt76_dma_add_rx_buf(struct mt76_dev *dev, struct mt76_queue *q,
 		    struct mt76_queue_buf *buf, void *data)
 {
 	struct mt76_queue_entry *entry = &q->entry[q->head];
 	struct mt76_txwi_cache *txwi = NULL;
+	u32 buf1 = 0, ctrl, info = 0;
 	struct mt76_desc *desc;
 	int idx = q->head;
-	u32 buf1 = 0, ctrl;
 	int rx_token;
 
 	if (mt76_queue_is_wed_rro_ind(q)) {
@@ -240,6 +245,9 @@ mt76_dma_add_rx_buf(struct mt76_dev *dev, struct mt76_queue *q,
 		rro_desc = (struct mt76_wed_rro_desc *)q->desc;
 		data = &rro_desc[q->head];
 		goto done;
+	} else if (mt76_queue_is_wed_rro_rxdmad_c(q)) {
+		data = &q->desc[q->head];
+		goto done;
 	}
 
 	desc = &q->desc[q->head];
@@ -248,7 +256,7 @@ mt76_dma_add_rx_buf(struct mt76_dev *dev, struct mt76_queue *q,
 	buf1 = FIELD_PREP(MT_DMA_CTL_SDP0_H, buf->addr >> 32);
 #endif
 
-	if (mt76_queue_is_wed_rx(q)) {
+	if (mt76_queue_is_wed_rx(q) || mt76_queue_is_wed_rro_data(q)) {
 		txwi = mt76_get_rxwi(dev);
 		if (!txwi)
 			return -ENOMEM;
@@ -261,12 +269,26 @@ mt76_dma_add_rx_buf(struct mt76_dev *dev, struct mt76_queue *q,
 
 		buf1 |= FIELD_PREP(MT_DMA_CTL_TOKEN, rx_token);
 		ctrl |= MT_DMA_CTL_TO_HOST;
+
+		txwi->qid = q - dev->q_rx;
+	}
+
+	if (mt76_queue_is_wed_rro_msdu_pg(q) &&
+	    dev->drv->rx_rro_add_msdu_page) {
+		if (dev->drv->rx_rro_add_msdu_page(dev, q, buf->addr, data))
+			return -ENOMEM;
+	}
+
+	if (q->flags & MT_QFLAG_WED_RRO_EN) {
+		info |= FIELD_PREP(MT_DMA_MAGIC_MASK, q->magic_cnt);
+		if ((q->head + 1) == q->ndesc)
+			q->magic_cnt = (q->magic_cnt + 1) % MT_DMA_MAGIC_CNT;
 	}
 
 	WRITE_ONCE(desc->buf0, cpu_to_le32(buf->addr));
 	WRITE_ONCE(desc->buf1, cpu_to_le32(buf1));
 	WRITE_ONCE(desc->ctrl, cpu_to_le32(ctrl));
-	WRITE_ONCE(desc->info, 0);
+	WRITE_ONCE(desc->info, cpu_to_le32(info));
 
 done:
 	entry->dma_addr[0] = buf->addr;
@@ -375,7 +397,10 @@ static void
 mt76_dma_kick_queue(struct mt76_dev *dev, struct mt76_queue *q)
 {
 	wmb();
-	Q_WRITE(q, cpu_idx, q->head);
+	if (mt76_queue_is_emi(q))
+		*q->emi_cpu_idx = cpu_to_le16(q->head);
+	else
+		Q_WRITE(q, cpu_idx, q->head);
 }
 
 static void
@@ -395,6 +420,7 @@ mt76_dma_tx_cleanup(struct mt76_dev *dev, struct mt76_queue *q, bool flush)
 
 	while (q->queued > 0 && q->tail != last) {
 		mt76_dma_tx_cleanup_idx(dev, q, q->tail, &entry);
+		mt76_npu_txdesc_cleanup(q, q->tail);
 		mt76_queue_tx_complete(dev, q, &entry);
 
 		if (entry.txwi) {
@@ -419,15 +445,61 @@ mt76_dma_tx_cleanup(struct mt76_dev *dev, struct mt76_queue *q, bool flush)
 }
 
 static void *
+mt76_dma_get_rxdmad_c_buf(struct mt76_dev *dev, struct mt76_queue *q,
+			  int idx, int *len, bool *more)
+{
+	struct mt76_queue_entry *e = &q->entry[idx];
+	struct mt76_rro_rxdmad_c *dmad = e->buf;
+	u32 data1 = le32_to_cpu(dmad->data1);
+	u32 data2 = le32_to_cpu(dmad->data2);
+	struct mt76_txwi_cache *t;
+	u16 rx_token_id;
+	u8 ind_reason;
+	void *buf;
+
+	rx_token_id = FIELD_GET(RRO_RXDMAD_DATA2_RX_TOKEN_ID_MASK, data2);
+	t = mt76_rx_token_release(dev, rx_token_id);
+	if (!t)
+		return ERR_PTR(-EAGAIN);
+
+	q = &dev->q_rx[t->qid];
+	dma_sync_single_for_cpu(dev->dma_dev, t->dma_addr,
+				SKB_WITH_OVERHEAD(q->buf_size),
+				page_pool_get_dma_dir(q->page_pool));
+
+	if (len)
+		*len = FIELD_GET(RRO_RXDMAD_DATA1_SDL0_MASK, data1);
+	if (more)
+		*more = !FIELD_GET(RRO_RXDMAD_DATA1_LS_MASK, data1);
+
+	buf = t->ptr;
+	ind_reason = FIELD_GET(RRO_RXDMAD_DATA2_IND_REASON_MASK, data2);
+	if (ind_reason == MT_DMA_WED_IND_REASON_REPEAT ||
+	    ind_reason == MT_DMA_WED_IND_REASON_OLDPKT) {
+		mt76_put_page_pool_buf(buf, false);
+		buf = ERR_PTR(-EAGAIN);
+	}
+	t->ptr = NULL;
+	t->dma_addr = 0;
+
+	mt76_put_rxwi(dev, t);
+
+	return buf;
+}
+
+static void *
 mt76_dma_get_buf(struct mt76_dev *dev, struct mt76_queue *q, int idx,
-		 int *len, u32 *info, bool *more, bool *drop)
+		 int *len, u32 *info, bool *more, bool *drop, bool flush)
 {
 	struct mt76_queue_entry *e = &q->entry[idx];
 	struct mt76_desc *desc = &q->desc[idx];
 	u32 ctrl, desc_info, buf1;
 	void *buf = e->buf;
 
-	if (mt76_queue_is_wed_rro_ind(q))
+	if (mt76_queue_is_wed_rro_rxdmad_c(q) && !flush)
+		buf = mt76_dma_get_rxdmad_c_buf(dev, q, idx, len, more);
+
+	if (mt76_queue_is_wed_rro(q))
 		goto done;
 
 	ctrl = le32_to_cpu(READ_ONCE(desc->ctrl));
@@ -482,20 +554,50 @@ mt76_dma_dequeue(struct mt76_dev *dev, struct mt76_queue *q, bool flush,
 	if (!q->queued)
 		return NULL;
 
-	if (mt76_queue_is_wed_rro_data(q))
-		return NULL;
+	if (mt76_queue_is_wed_rro_data(q) || mt76_queue_is_wed_rro_msdu_pg(q))
+		goto done;
+
+	if (mt76_queue_is_wed_rro_ind(q)) {
+		struct mt76_wed_rro_ind *cmd;
+		u8 magic_cnt;
 
-	if (!mt76_queue_is_wed_rro_ind(q)) {
+		if (flush)
+			goto done;
+
+		cmd = q->entry[idx].buf;
+		magic_cnt = FIELD_GET(RRO_IND_DATA1_MAGIC_CNT_MASK,
+				      le32_to_cpu(cmd->data1));
+		if (magic_cnt != q->magic_cnt)
+			return NULL;
+
+		if (q->tail == q->ndesc - 1)
+			q->magic_cnt = (q->magic_cnt + 1) % MT_DMA_WED_IND_CMD_CNT;
+	} else if (mt76_queue_is_wed_rro_rxdmad_c(q)) {
+		struct mt76_rro_rxdmad_c *dmad;
+		u16 magic_cnt;
+
+		if (flush)
+			goto done;
+
+		dmad = q->entry[idx].buf;
+		magic_cnt = FIELD_GET(RRO_RXDMAD_DATA3_MAGIC_CNT_MASK,
+				      le32_to_cpu(dmad->data3));
+		if (magic_cnt != q->magic_cnt)
+			return NULL;
+
+		if (q->tail == q->ndesc - 1)
+			q->magic_cnt = (q->magic_cnt + 1) % MT_DMA_MAGIC_CNT;
+	} else {
 		if (flush)
 			q->desc[idx].ctrl |= cpu_to_le32(MT_DMA_CTL_DMA_DONE);
 		else if (!(q->desc[idx].ctrl & cpu_to_le32(MT_DMA_CTL_DMA_DONE)))
 			return NULL;
 	}
-
+done:
 	q->tail = (q->tail + 1) % q->ndesc;
 	q->queued--;
 
-	return mt76_dma_get_buf(dev, q, idx, len, info, more, drop);
+	return mt76_dma_get_buf(dev, q, idx, len, info, more, drop, flush);
 }
 
 static int
@@ -553,6 +655,10 @@ mt76_dma_tx_queue_skb(struct mt76_phy *phy, struct mt76_queue *q,
 	if (test_bit(MT76_RESET, &phy->state))
 		goto free_skb;
 
+	/* TODO: Take into account unlinear skbs */
+	if (mt76_npu_device_active(dev) && skb_linearize(skb))
+		goto free_skb;
+
 	t = mt76_get_txwi(dev);
 	if (!t)
 		goto free_skb;
@@ -600,6 +706,9 @@ mt76_dma_tx_queue_skb(struct mt76_phy *phy, struct mt76_queue *q,
 	if (ret < 0)
 		goto unmap;
 
+	if (mt76_npu_device_active(dev))
+		return mt76_npu_dma_add_buf(phy, q, skb, &tx_info.buf[1], txwi);
+
 	return mt76_dma_add_buf(dev, q, tx_info.buf, tx_info.nbuf,
 				tx_info.info, tx_info.skb, t);
 
@@ -631,7 +740,8 @@ free_skb:
 	return ret;
 }
 
-int mt76_dma_rx_fill(struct mt76_dev *dev, struct mt76_queue *q,
+static int
+mt76_dma_rx_fill_buf(struct mt76_dev *dev, struct mt76_queue *q,
 		     bool allow_direct)
 {
 	int len = SKB_WITH_OVERHEAD(q->buf_size);
@@ -640,27 +750,21 @@ int mt76_dma_rx_fill(struct mt76_dev *dev, struct mt76_queue *q,
 	if (!q->ndesc)
 		return 0;
 
-	spin_lock_bh(&q->lock);
-
 	while (q->queued < q->ndesc - 1) {
 		struct mt76_queue_buf qbuf = {};
-		enum dma_data_direction dir;
-		dma_addr_t addr;
-		int offset;
 		void *buf = NULL;
+		int offset;
 
-		if (mt76_queue_is_wed_rro_ind(q))
+		if (mt76_queue_is_wed_rro_ind(q) ||
+		    mt76_queue_is_wed_rro_rxdmad_c(q))
 			goto done;
 
 		buf = mt76_get_page_pool_buf(q, &offset, q->buf_size);
 		if (!buf)
 			break;
 
-		addr = page_pool_get_dma_addr(virt_to_head_page(buf)) + offset;
-		dir = page_pool_get_dma_dir(q->page_pool);
-		dma_sync_single_for_device(dev->dma_dev, addr, len, dir);
-
-		qbuf.addr = addr + q->buf_offset;
+		qbuf.addr = page_pool_get_dma_addr(virt_to_head_page(buf)) +
+			    offset + q->buf_offset;
 done:
 		qbuf.len = len - q->buf_offset;
 		qbuf.skip_unmap = false;
@@ -674,6 +778,16 @@ done:
 	if (frames || mt76_queue_is_wed_rx(q))
 		mt76_dma_kick_queue(dev, q);
 
+	return frames;
+}
+
+int mt76_dma_rx_fill(struct mt76_dev *dev, struct mt76_queue *q,
+		     bool allow_direct)
+{
+	int frames;
+
+	spin_lock_bh(&q->lock);
+	frames = mt76_dma_rx_fill_buf(dev, q, allow_direct);
 	spin_unlock_bh(&q->lock);
 
 	return frames;
@@ -693,27 +807,23 @@ mt76_dma_alloc_queue(struct mt76_dev *dev, struct mt76_queue *q,
 	q->ndesc = n_desc;
 	q->buf_size = bufsize;
 	q->hw_idx = idx;
+	q->dev = dev;
+
+	if (mt76_queue_is_wed_rro_ind(q))
+		size = sizeof(struct mt76_wed_rro_desc);
+	else if (mt76_queue_is_npu_tx(q))
+		size = sizeof(struct airoha_npu_tx_dma_desc);
+	else if (mt76_queue_is_npu_rx(q))
+		size = sizeof(struct airoha_npu_rx_dma_desc);
+	else
+		size = sizeof(struct mt76_desc);
 
-	size = mt76_queue_is_wed_rro_ind(q) ? sizeof(struct mt76_wed_rro_desc)
-					    : sizeof(struct mt76_desc);
 	q->desc = dmam_alloc_coherent(dev->dma_dev, q->ndesc * size,
 				      &q->desc_dma, GFP_KERNEL);
 	if (!q->desc)
 		return -ENOMEM;
 
-	if (mt76_queue_is_wed_rro_ind(q)) {
-		struct mt76_wed_rro_desc *rro_desc;
-		int i;
-
-		rro_desc = (struct mt76_wed_rro_desc *)q->desc;
-		for (i = 0; i < q->ndesc; i++) {
-			struct mt76_wed_rro_ind *cmd;
-
-			cmd = (struct mt76_wed_rro_ind *)&rro_desc[i];
-			cmd->magic_cnt = MT_DMA_WED_IND_CMD_CNT - 1;
-		}
-	}
-
+	mt76_dma_queue_magic_cnt_init(dev, q);
 	size = q->ndesc * sizeof(*q->entry);
 	q->entry = devm_kzalloc(dev->dev, size, GFP_KERNEL);
 	if (!q->entry)
@@ -723,6 +833,7 @@ mt76_dma_alloc_queue(struct mt76_dev *dev, struct mt76_queue *q,
 	if (ret)
 		return ret;
 
+	mt76_npu_queue_setup(dev, q);
 	ret = mt76_wed_dma_setup(dev, q, false);
 	if (ret)
 		return ret;
@@ -733,7 +844,10 @@ mt76_dma_alloc_queue(struct mt76_dev *dev, struct mt76_queue *q,
 			return 0;
 	}
 
-	mt76_dma_queue_reset(dev, q);
+	/* HW specific driver is supposed to reset brand-new EMI queues since
+	 * it needs to set cpu index pointer.
+	 */
+	mt76_dma_queue_reset(dev, q, !mt76_queue_is_emi(q));
 
 	return 0;
 }
@@ -747,6 +861,11 @@ mt76_dma_rx_cleanup(struct mt76_dev *dev, struct mt76_queue *q)
 	if (!q->ndesc)
 		return;
 
+	if (mt76_queue_is_npu(q)) {
+		mt76_npu_queue_cleanup(dev, q);
+		return;
+	}
+
 	do {
 		spin_lock_bh(&q->lock);
 		buf = mt76_dma_dequeue(dev, q, true, NULL, NULL, &more, NULL);
@@ -776,7 +895,8 @@ mt76_dma_rx_reset(struct mt76_dev *dev, enum mt76_rxq_id qid)
 	if (!q->ndesc)
 		return;
 
-	if (!mt76_queue_is_wed_rro_ind(q)) {
+	if (!mt76_queue_is_wed_rro_ind(q) &&
+	    !mt76_queue_is_wed_rro_rxdmad_c(q) && !mt76_queue_is_npu(q)) {
 		int i;
 
 		for (i = 0; i < q->ndesc; i++)
@@ -796,7 +916,10 @@ mt76_dma_rx_reset(struct mt76_dev *dev, enum mt76_rxq_id qid)
 		return;
 
 	mt76_dma_sync_idx(dev, q);
-	mt76_dma_rx_fill(dev, q, false);
+	if (mt76_queue_is_npu(q))
+		mt76_npu_fill_rx_queue(dev, q);
+	else
+		mt76_dma_rx_fill(dev, q, false);
 }
 
 static void
@@ -836,8 +959,9 @@ mt76_dma_rx_process(struct mt76_dev *dev, struct mt76_queue *q, int budget)
 	bool allow_direct = !mt76_queue_is_wed_rx(q);
 	bool more;
 
-	if (IS_ENABLED(CONFIG_NET_MEDIATEK_SOC_WED) &&
-	    mt76_queue_is_wed_tx_free(q)) {
+	if ((q->flags & MT_QFLAG_WED_RRO_EN) ||
+	    (IS_ENABLED(CONFIG_NET_MEDIATEK_SOC_WED) &&
+	     mt76_queue_is_wed_tx_free(q))) {
 		dma_idx = Q_READ(q, dma_idx);
 		check_ddone = true;
 	}
@@ -859,6 +983,20 @@ mt76_dma_rx_process(struct mt76_dev *dev, struct mt76_queue *q, int budget)
 		if (!data)
 			break;
 
+		if (PTR_ERR(data) == -EAGAIN) {
+			done++;
+			continue;
+		}
+
+		if (mt76_queue_is_wed_rro_ind(q) && dev->drv->rx_rro_ind_process)
+			dev->drv->rx_rro_ind_process(dev, data);
+
+		if (mt76_queue_is_wed_rro(q) &&
+		    !mt76_queue_is_wed_rro_rxdmad_c(q)) {
+			done++;
+			continue;
+		}
+
 		if (drop)
 			goto free_frag;
 
@@ -936,6 +1074,15 @@ int mt76_dma_rx_poll(struct napi_struct *napi, int budget)
 }
 EXPORT_SYMBOL_GPL(mt76_dma_rx_poll);
 
+static void
+mt76_dma_rx_queue_init(struct mt76_dev *dev, enum mt76_rxq_id qid,
+		       int (*poll)(struct napi_struct *napi, int budget))
+{
+	netif_napi_add(dev->napi_dev, &dev->napi[qid], poll);
+	mt76_dma_rx_fill_buf(dev, &dev->q_rx[qid], false);
+	napi_enable(&dev->napi[qid]);
+}
+
 static int
 mt76_dma_init(struct mt76_dev *dev,
 	      int (*poll)(struct napi_struct *napi, int budget))
@@ -968,9 +1115,10 @@ mt76_dma_init(struct mt76_dev *dev,
 	init_completion(&dev->mmio.wed_reset_complete);
 
 	mt76_for_each_q_rx(dev, i) {
-		netif_napi_add(dev->napi_dev, &dev->napi[i], poll);
-		mt76_dma_rx_fill(dev, &dev->q_rx[i], false);
-		napi_enable(&dev->napi[i]);
+		if (mt76_queue_is_wed_rro(&dev->q_rx[i]))
+			continue;
+
+		mt76_dma_rx_queue_init(dev, i, poll);
 	}
 
 	return 0;
@@ -983,6 +1131,7 @@ static const struct mt76_queue_ops mt76_dma_ops = {
 	.tx_queue_skb_raw = mt76_dma_tx_queue_skb_raw,
 	.tx_queue_skb = mt76_dma_tx_queue_skb,
 	.tx_cleanup = mt76_dma_tx_cleanup,
+	.rx_queue_init = mt76_dma_rx_queue_init,
 	.rx_cleanup = mt76_dma_rx_cleanup,
 	.rx_reset = mt76_dma_rx_reset,
 	.kick = mt76_dma_kick_queue,
@@ -999,6 +1148,7 @@ void mt76_dma_cleanup(struct mt76_dev *dev)
 	int i;
 
 	mt76_worker_disable(&dev->tx_worker);
+	napi_disable(&dev->tx_napi);
 	netif_napi_del(&dev->tx_napi);
 
 	for (i = 0; i < ARRAY_SIZE(dev->phys); i++) {