* [PATCH v3 3/4] wifi: rtw88: usb: Support RX aggregation
2024-08-06 17:39 [PATCH v3 1/4] wifi: rtw88: usb: Init RX burst length according to USB speed Bitterblue Smith
2024-08-06 17:41 ` [PATCH v3 2/4] wifi: rtw88: usb: Update the RX stats after every frame Bitterblue Smith
@ 2024-08-06 17:45 ` Bitterblue Smith
2024-08-06 17:47 ` [PATCH v3 4/4] wifi: rtw88: Enable USB RX aggregation for 8822c/8822b/8821c Bitterblue Smith
2 siblings, 0 replies; 6+ messages in thread
From: Bitterblue Smith @ 2024-08-06 17:45 UTC (permalink / raw)
To: linux-wireless@vger.kernel.org; +Cc: Ping-Ke Shih, Sascha Hauer
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
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH v3 4/4] wifi: rtw88: Enable USB RX aggregation for 8822c/8822b/8821c
2024-08-06 17:39 [PATCH v3 1/4] wifi: rtw88: usb: Init RX burst length according to USB speed Bitterblue Smith
2024-08-06 17:41 ` [PATCH v3 2/4] wifi: rtw88: usb: Update the RX stats after every frame Bitterblue Smith
2024-08-06 17:45 ` [PATCH v3 3/4] wifi: rtw88: usb: Support RX aggregation Bitterblue Smith
@ 2024-08-06 17:47 ` Bitterblue Smith
2024-08-07 0:37 ` Ping-Ke Shih
2 siblings, 1 reply; 6+ messages in thread
From: Bitterblue Smith @ 2024-08-06 17:47 UTC (permalink / raw)
To: linux-wireless@vger.kernel.org; +Cc: Ping-Ke Shih, Sascha Hauer
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
^ permalink raw reply related [flat|nested] 6+ messages in thread