[PATCH v3 1/4] wifi: rtw88: usb: Init RX burst length according to USB speed

[PATCH v3 1/4] wifi: rtw88: usb: Init RX burst length according to USB speed

[Bitterblue Smith]

This is needed in order to make USB RX aggregation work with RTL8811CU
(and presumably RTL8822BU and RTL8822CU also).

I don't know what BIT_DMA_BURST_CNT, BIT_DMA_MODE, and BIT_DROP_DATA_EN
are doing.

Tested with RTL8811CU and RTL8723DU.

The RX speed is unchanged in my tests.

Tested-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Bitterblue Smith <rtl8821cerfe2@gmail.com>
---
v3:
 - Add Tested-by: Sascha Hauer <s.hauer@pengutronix.de>

v2:
 - Move the code to rtw_usb_interface_cfg.
 - Let RTL8723DU and all other USB devices use it, not just RTL8822CU,
   RTL8822BU, and RTL8821CU.
 - Use the speed member of struct usb_device to determine the USB speed
   instead of reading hardware registers.
 - Update the subject line.
 - Add more information to the commit message.
 - Rebase on top of the latest rtw-next.
---
 drivers/net/wireless/realtek/rtw88/reg.h |  6 ++++++
 drivers/net/wireless/realtek/rtw88/usb.c | 23 ++++++++++++++++++++++-
 2 files changed, 28 insertions(+), 1 deletion(-)

diff --git a/drivers/net/wireless/realtek/rtw88/reg.h b/drivers/net/wireless/realtek/rtw88/reg.h
index e7b24465f549..4d9b8668e8b0 100644
--- a/drivers/net/wireless/realtek/rtw88/reg.h
+++ b/drivers/net/wireless/realtek/rtw88/reg.h
@@ -322,6 +322,12 @@
 #define REG_RXDMA_DPR		0x028C
 #define REG_RXDMA_MODE		0x0290
 #define BIT_DMA_MODE		BIT(1)
+#define BIT_DMA_BURST_CNT	GENMASK(3, 2)
+#define BIT_DMA_BURST_SIZE	GENMASK(5, 4)
+#define BIT_DMA_BURST_SIZE_64	2
+#define BIT_DMA_BURST_SIZE_512	1
+#define BIT_DMA_BURST_SIZE_1024	0
+
 #define REG_RXPKTNUM		0x02B0
 
 #define REG_INT_MIG		0x0304
diff --git a/drivers/net/wireless/realtek/rtw88/usb.c b/drivers/net/wireless/realtek/rtw88/usb.c
index 9145c11a063e..1c40d46a7eb4 100644
--- a/drivers/net/wireless/realtek/rtw88/usb.c
+++ b/drivers/net/wireless/realtek/rtw88/usb.c
@@ -720,9 +720,30 @@ static void rtw_usb_link_ps(struct rtw_dev *rtwdev, bool enter)
 	/* empty function for rtw_hci_ops */
 }
 
+static void rtw_usb_init_burst_pkt_len(struct rtw_dev *rtwdev)
+{
+	struct rtw_usb *rtwusb = rtw_get_usb_priv(rtwdev);
+	enum usb_device_speed speed = rtwusb->udev->speed;
+	u8 rxdma, burst_size;
+
+	rxdma = BIT_DMA_BURST_CNT | BIT_DMA_MODE;
+
+	if (speed == USB_SPEED_SUPER)
+		burst_size = BIT_DMA_BURST_SIZE_1024;
+	else if (speed == USB_SPEED_HIGH)
+		burst_size = BIT_DMA_BURST_SIZE_512;
+	else
+		burst_size = BIT_DMA_BURST_SIZE_64;
+
+	u8p_replace_bits(&rxdma, burst_size, BIT_DMA_BURST_SIZE);
+
+	rtw_write8(rtwdev, REG_RXDMA_MODE, rxdma);
+	rtw_write16_set(rtwdev, REG_TXDMA_OFFSET_CHK, BIT_DROP_DATA_EN);
+}
+
 static void rtw_usb_interface_cfg(struct rtw_dev *rtwdev)
 {
-	/* empty function for rtw_hci_ops */
+	rtw_usb_init_burst_pkt_len(rtwdev);
 }
 
 static struct rtw_hci_ops rtw_usb_ops = {
-- 
2.45.2

[PATCH v3 2/4] wifi: rtw88: usb: Update the RX stats after every frame

[Bitterblue Smith]

Update the number of received unicast data frames and bytes every time
a frame is received. This is what the PCI and SDIO drivers do.

This has an influence on the power saving, bluetooth coexistence, and
(in a future patch) the use of RX aggregation.

Tested with RTL8811CU and RTL8723DU.

Tested-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Bitterblue Smith <rtl8821cerfe2@gmail.com>
---
v3:
 - Add Tested-by: Sascha Hauer <s.hauer@pengutronix.de>

v2:
 - Rebase on top of the latest rtw-next.
---
 drivers/net/wireless/realtek/rtw88/usb.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/wireless/realtek/rtw88/usb.c b/drivers/net/wireless/realtek/rtw88/usb.c
index 1c40d46a7eb4..10f1d724370e 100644
--- a/drivers/net/wireless/realtek/rtw88/usb.c
+++ b/drivers/net/wireless/realtek/rtw88/usb.c
@@ -581,6 +581,7 @@ static void rtw_usb_rx_handler(struct work_struct *work)
 		skb_reserve(skb, pkt_offset);
 
 		rtw_update_rx_freq_for_invalid(rtwdev, skb, &rx_status, &pkt_stat);
+		rtw_rx_stats(rtwdev, pkt_stat.vif, skb);
 		memcpy(skb->cb, &rx_status, sizeof(rx_status));
 		ieee80211_rx_irqsafe(rtwdev->hw, skb);
 	}
-- 
2.45.2

[PATCH v3 3/4] wifi: rtw88: usb: Support RX aggregation

[Bitterblue Smith]

The chips can be configured to aggregate several frames into a single
USB transfer. Modify rtw_usb_rx_handler() to support this case.

RX aggregation improves the RX speed on certain ARM systems, like the
NanoPi NEO Core2.

Currently none of the chips are configured to aggregate frames.

Tested with RTL8811CU and RTL8723DU.

Reviewed-by: Sascha Hauer <s.hauer@pengutronix.de>
Tested-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Bitterblue Smith <rtl8821cerfe2@gmail.com>
---
v3:
 - Don't initialise next_skb.
 - Add Tested-by and Reviewed-by: Sascha Hauer <s.hauer@pengutronix.de>

v2:
 - Simplify the code and make it more readable.
 - Rebase on top of latest rtw-next.
---
 drivers/net/wireless/realtek/rtw88/usb.c | 61 ++++++++++++++++--------
 1 file changed, 40 insertions(+), 21 deletions(-)

diff --git a/drivers/net/wireless/realtek/rtw88/usb.c b/drivers/net/wireless/realtek/rtw88/usb.c
index 10f1d724370e..4c7ba5c76a57 100644
--- a/drivers/net/wireless/realtek/rtw88/usb.c
+++ b/drivers/net/wireless/realtek/rtw88/usb.c
@@ -546,11 +546,12 @@ static void rtw_usb_rx_handler(struct work_struct *work)
 	struct rtw_usb *rtwusb = container_of(work, struct rtw_usb, rx_work);
 	struct rtw_dev *rtwdev = rtwusb->rtwdev;
 	const struct rtw_chip_info *chip = rtwdev->chip;
-	struct rtw_rx_pkt_stat pkt_stat;
+	u32 pkt_desc_sz = chip->rx_pkt_desc_sz;
 	struct ieee80211_rx_status rx_status;
+	u32 pkt_offset, next_pkt, urb_len;
+	struct rtw_rx_pkt_stat pkt_stat;
+	struct sk_buff *next_skb;
 	struct sk_buff *skb;
-	u32 pkt_desc_sz = chip->rx_pkt_desc_sz;
-	u32 pkt_offset;
 	u8 *rx_desc;
 	int limit;
 
@@ -559,31 +560,48 @@ static void rtw_usb_rx_handler(struct work_struct *work)
 		if (!skb)
 			break;
 
-		rx_desc = skb->data;
-		chip->ops->query_rx_desc(rtwdev, rx_desc, &pkt_stat,
-					 &rx_status);
-		pkt_offset = pkt_desc_sz + pkt_stat.drv_info_sz +
-			     pkt_stat.shift;
-
-		if (pkt_stat.is_c2h) {
-			skb_put(skb, pkt_stat.pkt_len + pkt_offset);
-			rtw_fw_c2h_cmd_rx_irqsafe(rtwdev, pkt_offset, skb);
-			continue;
-		}
-
 		if (skb_queue_len(&rtwusb->rx_queue) >= RTW_USB_MAX_RXQ_LEN) {
 			dev_dbg_ratelimited(rtwdev->dev, "failed to get rx_queue, overflow\n");
 			dev_kfree_skb_any(skb);
 			continue;
 		}
 
-		skb_put(skb, pkt_stat.pkt_len);
-		skb_reserve(skb, pkt_offset);
+		urb_len = skb->len;
+
+		do {
+			rx_desc = skb->data;
+			chip->ops->query_rx_desc(rtwdev, rx_desc, &pkt_stat,
+						 &rx_status);
+			pkt_offset = pkt_desc_sz + pkt_stat.drv_info_sz +
+				     pkt_stat.shift;
+
+			next_pkt = round_up(pkt_stat.pkt_len + pkt_offset, 8);
+
+			if (urb_len >= next_pkt + pkt_desc_sz)
+				next_skb = skb_clone(skb, GFP_KERNEL);
+			else
+				next_skb = NULL;
+
+			if (pkt_stat.is_c2h) {
+				skb_trim(skb, pkt_stat.pkt_len + pkt_offset);
+				rtw_fw_c2h_cmd_rx_irqsafe(rtwdev, pkt_offset, skb);
+			} else {
+				skb_pull(skb, pkt_offset);
+				skb_trim(skb, pkt_stat.pkt_len);
+				rtw_update_rx_freq_for_invalid(rtwdev, skb,
+							       &rx_status,
+							       &pkt_stat);
+				rtw_rx_stats(rtwdev, pkt_stat.vif, skb);
+				memcpy(skb->cb, &rx_status, sizeof(rx_status));
+				ieee80211_rx_irqsafe(rtwdev->hw, skb);
+			}
+
+			skb = next_skb;
+			if (skb)
+				skb_pull(skb, next_pkt);
 
-		rtw_update_rx_freq_for_invalid(rtwdev, skb, &rx_status, &pkt_stat);
-		rtw_rx_stats(rtwdev, pkt_stat.vif, skb);
-		memcpy(skb->cb, &rx_status, sizeof(rx_status));
-		ieee80211_rx_irqsafe(rtwdev->hw, skb);
+			urb_len -= next_pkt;
+		} while (skb);
 	}
 }
 
@@ -627,6 +645,7 @@ static void rtw_usb_read_port_complete(struct urb *urb)
 			if (skb)
 				dev_kfree_skb_any(skb);
 		} else {
+			skb_put(skb, urb->actual_length);
 			skb_queue_tail(&rtwusb->rx_queue, skb);
 			queue_work(rtwusb->rxwq, &rtwusb->rx_work);
 		}
-- 
2.45.2

[PATCH v3 4/4] wifi: rtw88: Enable USB RX aggregation for 8822c/8822b/8821c

[Bitterblue Smith]

Enable USB RX aggregation when there is at least 1 Mbps RX or TX
traffic, otherwise disable it.

USB RX aggregation improves the RX speed on certain ARM systems, like
the NanoPi NEO Core2. With RTL8821CU, before: 28 Mbps, after: 231 Mbps.

The official drivers for these chips use the same logic for SDIO, but
for some reason the SDIO driver in rtw88 always enables RX aggregation,
so this patch only toggles aggregation for USB devices.

RTL8703B is likely not found in USB devices, and RTL8723DU doesn't like
aggregation.

Tested-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Bitterblue Smith <rtl8821cerfe2@gmail.com>
---
v3:
 - Add Tested-by: Sascha Hauer <s.hauer@pengutronix.de>

v2:
 - Rename {tx,rx}_unicast_shift to {tx,rx}_unicast_mbps.
 - Move the RX aggregation code from rtw8822c.c, rtw8822b.c, rtw8821c.c
   to usb.c.
 - Delete the rx_aggregation member from struct rtw_chip_ops and add
   dynamic_rx_agg member to struct rtw_hci_ops.
 - Rebase on top of the latest rtw-next.
---
 drivers/net/wireless/realtek/rtw88/hci.h  |  7 ++++
 drivers/net/wireless/realtek/rtw88/main.c | 13 +++++---
 drivers/net/wireless/realtek/rtw88/pci.c  |  1 +
 drivers/net/wireless/realtek/rtw88/sdio.c |  1 +
 drivers/net/wireless/realtek/rtw88/usb.c  | 40 +++++++++++++++++++++++
 5 files changed, 58 insertions(+), 4 deletions(-)

diff --git a/drivers/net/wireless/realtek/rtw88/hci.h b/drivers/net/wireless/realtek/rtw88/hci.h
index 830d7532f2a3..96aeda26014e 100644
--- a/drivers/net/wireless/realtek/rtw88/hci.h
+++ b/drivers/net/wireless/realtek/rtw88/hci.h
@@ -18,6 +18,7 @@ struct rtw_hci_ops {
 	void (*deep_ps)(struct rtw_dev *rtwdev, bool enter);
 	void (*link_ps)(struct rtw_dev *rtwdev, bool enter);
 	void (*interface_cfg)(struct rtw_dev *rtwdev);
+	void (*dynamic_rx_agg)(struct rtw_dev *rtwdev, bool enable);
 
 	int (*write_data_rsvd_page)(struct rtw_dev *rtwdev, u8 *buf, u32 size);
 	int (*write_data_h2c)(struct rtw_dev *rtwdev, u8 *buf, u32 size);
@@ -72,6 +73,12 @@ static inline void rtw_hci_interface_cfg(struct rtw_dev *rtwdev)
 	rtwdev->hci.ops->interface_cfg(rtwdev);
 }
 
+static inline void rtw_hci_dynamic_rx_agg(struct rtw_dev *rtwdev, bool enable)
+{
+	if (rtwdev->hci.ops->dynamic_rx_agg)
+		rtwdev->hci.ops->dynamic_rx_agg(rtwdev, enable);
+}
+
 static inline int
 rtw_hci_write_data_rsvd_page(struct rtw_dev *rtwdev, u8 *buf, u32 size)
 {
diff --git a/drivers/net/wireless/realtek/rtw88/main.c b/drivers/net/wireless/realtek/rtw88/main.c
index 9c58b7a41b95..fd944248e6e7 100644
--- a/drivers/net/wireless/realtek/rtw88/main.c
+++ b/drivers/net/wireless/realtek/rtw88/main.c
@@ -212,6 +212,7 @@ static void rtw_watch_dog_work(struct work_struct *work)
 	struct rtw_traffic_stats *stats = &rtwdev->stats;
 	struct rtw_watch_dog_iter_data data = {};
 	bool busy_traffic = test_bit(RTW_FLAG_BUSY_TRAFFIC, rtwdev->flags);
+	u32 tx_unicast_mbps, rx_unicast_mbps;
 	bool ps_active;
 
 	mutex_lock(&rtwdev->mutex);
@@ -236,13 +237,17 @@ static void rtw_watch_dog_work(struct work_struct *work)
 	else
 		ps_active = false;
 
-	ewma_tp_add(&stats->tx_ewma_tp,
-		    (u32)(stats->tx_unicast >> RTW_TP_SHIFT));
-	ewma_tp_add(&stats->rx_ewma_tp,
-		    (u32)(stats->rx_unicast >> RTW_TP_SHIFT));
+	tx_unicast_mbps = stats->tx_unicast >> RTW_TP_SHIFT;
+	rx_unicast_mbps = stats->rx_unicast >> RTW_TP_SHIFT;
+
+	ewma_tp_add(&stats->tx_ewma_tp, tx_unicast_mbps);
+	ewma_tp_add(&stats->rx_ewma_tp, rx_unicast_mbps);
 	stats->tx_throughput = ewma_tp_read(&stats->tx_ewma_tp);
 	stats->rx_throughput = ewma_tp_read(&stats->rx_ewma_tp);
 
+	rtw_hci_dynamic_rx_agg(rtwdev,
+			       tx_unicast_mbps >= 1 || rx_unicast_mbps >= 1);
+
 	/* reset tx/rx statictics */
 	stats->tx_unicast = 0;
 	stats->rx_unicast = 0;
diff --git a/drivers/net/wireless/realtek/rtw88/pci.c b/drivers/net/wireless/realtek/rtw88/pci.c
index 5d0580da13fb..0b9b8807af2c 100644
--- a/drivers/net/wireless/realtek/rtw88/pci.c
+++ b/drivers/net/wireless/realtek/rtw88/pci.c
@@ -1601,6 +1601,7 @@ static struct rtw_hci_ops rtw_pci_ops = {
 	.deep_ps = rtw_pci_deep_ps,
 	.link_ps = rtw_pci_link_ps,
 	.interface_cfg = rtw_pci_interface_cfg,
+	.dynamic_rx_agg = NULL,
 
 	.read8 = rtw_pci_read8,
 	.read16 = rtw_pci_read16,
diff --git a/drivers/net/wireless/realtek/rtw88/sdio.c b/drivers/net/wireless/realtek/rtw88/sdio.c
index 763aa8212a4b..21d0754dd7f6 100644
--- a/drivers/net/wireless/realtek/rtw88/sdio.c
+++ b/drivers/net/wireless/realtek/rtw88/sdio.c
@@ -1157,6 +1157,7 @@ static struct rtw_hci_ops rtw_sdio_ops = {
 	.deep_ps = rtw_sdio_deep_ps,
 	.link_ps = rtw_sdio_link_ps,
 	.interface_cfg = rtw_sdio_interface_cfg,
+	.dynamic_rx_agg = NULL,
 
 	.read8 = rtw_sdio_read8,
 	.read16 = rtw_sdio_read16,
diff --git a/drivers/net/wireless/realtek/rtw88/usb.c b/drivers/net/wireless/realtek/rtw88/usb.c
index 4c7ba5c76a57..cbb5e17e65d1 100644
--- a/drivers/net/wireless/realtek/rtw88/usb.c
+++ b/drivers/net/wireless/realtek/rtw88/usb.c
@@ -766,6 +766,45 @@ static void rtw_usb_interface_cfg(struct rtw_dev *rtwdev)
 	rtw_usb_init_burst_pkt_len(rtwdev);
 }
 
+static void rtw_usb_dynamic_rx_agg_v1(struct rtw_dev *rtwdev, bool enable)
+{
+	u8 size, timeout;
+	u16 val16;
+
+	rtw_write32_set(rtwdev, REG_RXDMA_AGG_PG_TH, BIT_EN_PRE_CALC);
+	rtw_write8_set(rtwdev, REG_TXDMA_PQ_MAP, BIT_RXDMA_AGG_EN);
+	rtw_write8_clr(rtwdev, REG_RXDMA_AGG_PG_TH + 3, BIT(7));
+
+	if (enable) {
+		size = 0x5;
+		timeout = 0x20;
+	} else {
+		size = 0x0;
+		timeout = 0x1;
+	}
+	val16 = u16_encode_bits(size, BIT_RXDMA_AGG_PG_TH) |
+		u16_encode_bits(timeout, BIT_DMA_AGG_TO_V1);
+
+	rtw_write16(rtwdev, REG_RXDMA_AGG_PG_TH, val16);
+}
+
+static void rtw_usb_dynamic_rx_agg(struct rtw_dev *rtwdev, bool enable)
+{
+	switch (rtwdev->chip->id) {
+	case RTW_CHIP_TYPE_8822C:
+	case RTW_CHIP_TYPE_8822B:
+	case RTW_CHIP_TYPE_8821C:
+		rtw_usb_dynamic_rx_agg_v1(rtwdev, enable);
+		break;
+	case RTW_CHIP_TYPE_8723D:
+		/* Doesn't like aggregation. */
+		break;
+	case RTW_CHIP_TYPE_8703B:
+		/* Likely not found in USB devices. */
+		break;
+	}
+}
+
 static struct rtw_hci_ops rtw_usb_ops = {
 	.tx_write = rtw_usb_tx_write,
 	.tx_kick_off = rtw_usb_tx_kick_off,
@@ -775,6 +814,7 @@ static struct rtw_hci_ops rtw_usb_ops = {
 	.deep_ps = rtw_usb_deep_ps,
 	.link_ps = rtw_usb_link_ps,
 	.interface_cfg = rtw_usb_interface_cfg,
+	.dynamic_rx_agg = rtw_usb_dynamic_rx_agg,
 
 	.write8  = rtw_usb_write8,
 	.write16 = rtw_usb_write16,
-- 
2.45.2

RE: [PATCH v3 4/4] wifi: rtw88: Enable USB RX aggregation for 8822c/8822b/8821c

[Ping-Ke Shih]

Bitterblue Smith <rtl8821cerfe2@gmail.com> wrote:
> @@ -236,13 +237,17 @@ static void rtw_watch_dog_work(struct work_struct *work)
>         else
>                 ps_active = false;
> 
> -       ewma_tp_add(&stats->tx_ewma_tp,
> -                   (u32)(stats->tx_unicast >> RTW_TP_SHIFT));
> -       ewma_tp_add(&stats->rx_ewma_tp,
> -                   (u32)(stats->rx_unicast >> RTW_TP_SHIFT));
> +       tx_unicast_mbps = stats->tx_unicast >> RTW_TP_SHIFT;
> +       rx_unicast_mbps = stats->rx_unicast >> RTW_TP_SHIFT;
> +
> +       ewma_tp_add(&stats->tx_ewma_tp, tx_unicast_mbps);
> +       ewma_tp_add(&stats->rx_ewma_tp, rx_unicast_mbps);
>         stats->tx_throughput = ewma_tp_read(&stats->tx_ewma_tp);
>         stats->rx_throughput = ewma_tp_read(&stats->rx_ewma_tp);
> 
> +       rtw_hci_dynamic_rx_agg(rtwdev,
> +                              tx_unicast_mbps >= 1 || rx_unicast_mbps >= 1);
> +

Not sure if you have tried RTL8822CU with this dynamic_rx_agg? 
I suspect RTL8822CU can't access IO before rtw_leave_lps(), at least RTL8822CE can't.
Let's move rtw_hci_dynamic_rx_agg() right after rtw_phy_dynamic_mechanism() below.

Sorry to forget this point on v2 review. 

>         /* reset tx/rx statictics */
>         stats->tx_unicast = 0;
>         stats->rx_unicast = 0;

Re: [PATCH v3 4/4] wifi: rtw88: Enable USB RX aggregation for 8822c/8822b/8821c

[Bitterblue Smith]

On 07/08/2024 03:37, Ping-Ke Shih wrote:
> Bitterblue Smith <rtl8821cerfe2@gmail.com> wrote:
>> @@ -236,13 +237,17 @@ static void rtw_watch_dog_work(struct work_struct *work)
>>         else
>>                 ps_active = false;
>>
>> -       ewma_tp_add(&stats->tx_ewma_tp,
>> -                   (u32)(stats->tx_unicast >> RTW_TP_SHIFT));
>> -       ewma_tp_add(&stats->rx_ewma_tp,
>> -                   (u32)(stats->rx_unicast >> RTW_TP_SHIFT));
>> +       tx_unicast_mbps = stats->tx_unicast >> RTW_TP_SHIFT;
>> +       rx_unicast_mbps = stats->rx_unicast >> RTW_TP_SHIFT;
>> +
>> +       ewma_tp_add(&stats->tx_ewma_tp, tx_unicast_mbps);
>> +       ewma_tp_add(&stats->rx_ewma_tp, rx_unicast_mbps);
>>         stats->tx_throughput = ewma_tp_read(&stats->tx_ewma_tp);
>>         stats->rx_throughput = ewma_tp_read(&stats->rx_ewma_tp);
>>
>> +       rtw_hci_dynamic_rx_agg(rtwdev,
>> +                              tx_unicast_mbps >= 1 || rx_unicast_mbps >= 1);
>> +
> 
> Not sure if you have tried RTL8822CU with this dynamic_rx_agg? 
> I suspect RTL8822CU can't access IO before rtw_leave_lps(), at least RTL8822CE can't.
> Let's move rtw_hci_dynamic_rx_agg() right after rtw_phy_dynamic_mechanism() below.
> 
> Sorry to forget this point on v2 review. 
> 
>>         /* reset tx/rx statictics */
>>         stats->tx_unicast = 0;
>>         stats->rx_unicast = 0;
> 
> 
> 

I received RTL8822CU (LM842) two days ago and tested these
patches with it, but I was too lazy to update the commit
messages. I didn't notice any problems. The RX speed measured
with iperf3 before was ~200 Mbps, after it's ~300 Mbps on my
x86_64 laptop.

I will move the function call.