nanopi-r76s: fix slow WiFi by enabling SDIO SDR104 (#9929)

SuperKali · web-flow · commit 2858333a7f07 · 2026-06-04T07:37:13.000+02:00
* rockchip64: mmc: skip SDIO 1.8V switch when host already at 1.8V

A UHS SDIO card can report S18A=1 even on a fixed-1.8V rail, so the core
ran CMD11 and stalled the dw_mmc voltage switch, capping the bus at HS
50MHz. Guard the switch on signal_voltage, mirroring mmc_sd_init_card.
Added for current, edge and bleedingedge.

Signed-off-by: SuperKali &lt;hello@superkali.me&gt;

* nanopi-r76s: add cap-sd-highspeed to SDIO WiFi node

Gives the RTL8822CS a clean 50MHz high-speed fallback if SDR104 tuning
fails, instead of dropping to 25MHz legacy. With the mmc CMD11 fix the
WiFi now links at SDR104 200MHz, ~210 Mbit/s down versus 51 before.

Signed-off-by: SuperKali &lt;hello@superkali.me&gt;

---------

Signed-off-by: SuperKali &lt;hello@superkali.me&gt;
diff --git a/patch/kernel/archive/rockchip64-6.18/dt/rk3576-nanopi-r76s.dts b/patch/kernel/archive/rockchip64-6.18/dt/rk3576-nanopi-r76s.dts
@@ -753,6 +753,7 @@
 &sdio {
 	max-frequency = <200000000>;
 	bus-width = <4>;
+	cap-sd-highspeed;
 	cap-sdio-irq;
 	disable-wp;
 	keep-power-in-suspend;
diff --git a/patch/kernel/archive/rockchip64-6.18/mmc-sdio-skip-cmd11-voltage-switch-when-already-1v8.patch b/patch/kernel/archive/rockchip64-6.18/mmc-sdio-skip-cmd11-voltage-switch-when-already-1v8.patch
@@ -0,0 +1,42 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: SuperKali <hello@superkali.me>
+Date: Wed, 3 Jun 2026 17:28:00 +0000
+Subject: mmc: sdio: skip the 1.8V switch when the host already signals at 1.8V
+
+A UHS-capable SDIO card can report S18A=1 even when the host signalling
+is already pinned to 1.8V. The core then issues CMD11 anyway, and with no
+real 3.3V to 1.8V transition the controller can stall on the voltage-switch
+clock update and fall back to high speed.
+
+Skip the switch when ios.signal_voltage is already 1.8V, the same guard
+mmc_sd_init_card() uses on the SD path. This lets an RTL8822CS on a
+fixed-1.8V rail come up at SDR104.
+
+Signed-off-by: SuperKali <hello@superkali.me>
+---
+ drivers/mmc/core/sdio.c | 8 +++++++-
+ 1 file changed, 7 insertions(+), 1 deletion(-)
+
+diff --git a/drivers/mmc/core/sdio.c b/drivers/mmc/core/sdio.c
+index 111111111111..222222222222 100644
+--- a/drivers/mmc/core/sdio.c
++++ b/drivers/mmc/core/sdio.c
+@@ -738,8 +738,14 @@ static int mmc_sdio_init_card(struct mmc_host *host, u32 ocr,
+ 	 * fails to check rocr & R4_18V_PRESENT,  but we still need to
+ 	 * try to init uhs card. sdio_read_cccr will take over this task
+ 	 * to make sure which speed mode should work.
++	 *
++	 * If the host already signals at 1.8V there is nothing to switch.
++	 * Some cards still report S18A=1, so guard CMD11 explicitly to avoid
++	 * a redundant voltage switch that can stall the controller. Mirrors
++	 * mmc_sd_init_card().
+ 	 */
+-	if (rocr & ocr & R4_18V_PRESENT) {
++	if ((rocr & ocr & R4_18V_PRESENT) &&
++	    host->ios.signal_voltage != MMC_SIGNAL_VOLTAGE_180) {
+ 		err = mmc_set_uhs_voltage(host, ocr_card);
+ 		if (err == -EAGAIN) {
+ 			mmc_sdio_pre_init(host, ocr_card, card);
+-- 
+Armbian
+
diff --git a/patch/kernel/archive/rockchip64-7.0/dt/rk3576-nanopi-r76s.dts b/patch/kernel/archive/rockchip64-7.0/dt/rk3576-nanopi-r76s.dts
@@ -753,6 +753,7 @@
 &sdio {
 	max-frequency = <200000000>;
 	bus-width = <4>;
+	cap-sd-highspeed;
 	cap-sdio-irq;
 	disable-wp;
 	keep-power-in-suspend;
diff --git a/patch/kernel/archive/rockchip64-7.0/mmc-sdio-skip-cmd11-voltage-switch-when-already-1v8.patch b/patch/kernel/archive/rockchip64-7.0/mmc-sdio-skip-cmd11-voltage-switch-when-already-1v8.patch
@@ -0,0 +1,42 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: SuperKali <hello@superkali.me>
+Date: Wed, 3 Jun 2026 17:28:00 +0000
+Subject: mmc: sdio: skip the 1.8V switch when the host already signals at 1.8V
+
+A UHS-capable SDIO card can report S18A=1 even when the host signalling
+is already pinned to 1.8V. The core then issues CMD11 anyway, and with no
+real 3.3V to 1.8V transition the controller can stall on the voltage-switch
+clock update and fall back to high speed.
+
+Skip the switch when ios.signal_voltage is already 1.8V, the same guard
+mmc_sd_init_card() uses on the SD path. This lets an RTL8822CS on a
+fixed-1.8V rail come up at SDR104.
+
+Signed-off-by: SuperKali <hello@superkali.me>
+---
+ drivers/mmc/core/sdio.c | 8 +++++++-
+ 1 file changed, 7 insertions(+), 1 deletion(-)
+
+diff --git a/drivers/mmc/core/sdio.c b/drivers/mmc/core/sdio.c
+index 111111111111..222222222222 100644
+--- a/drivers/mmc/core/sdio.c
++++ b/drivers/mmc/core/sdio.c
+@@ -738,8 +738,14 @@ static int mmc_sdio_init_card(struct mmc_host *host, u32 ocr,
+ 	 * fails to check rocr & R4_18V_PRESENT,  but we still need to
+ 	 * try to init uhs card. sdio_read_cccr will take over this task
+ 	 * to make sure which speed mode should work.
++	 *
++	 * If the host already signals at 1.8V there is nothing to switch.
++	 * Some cards still report S18A=1, so guard CMD11 explicitly to avoid
++	 * a redundant voltage switch that can stall the controller. Mirrors
++	 * mmc_sd_init_card().
+ 	 */
+-	if (rocr & ocr & R4_18V_PRESENT) {
++	if ((rocr & ocr & R4_18V_PRESENT) &&
++	    host->ios.signal_voltage != MMC_SIGNAL_VOLTAGE_180) {
+ 		err = mmc_set_uhs_voltage(host, ocr_card);
+ 		if (err == -EAGAIN) {
+ 			mmc_sdio_pre_init(host, ocr_card, card);
+-- 
+Armbian
+
diff --git a/patch/kernel/archive/rockchip64-7.1/dt/rk3576-nanopi-r76s.dts b/patch/kernel/archive/rockchip64-7.1/dt/rk3576-nanopi-r76s.dts
@@ -753,6 +753,7 @@
 &sdio {
 	max-frequency = <200000000>;
 	bus-width = <4>;
+	cap-sd-highspeed;
 	cap-sdio-irq;
 	disable-wp;
 	keep-power-in-suspend;
diff --git a/patch/kernel/archive/rockchip64-7.1/mmc-sdio-skip-cmd11-voltage-switch-when-already-1v8.patch b/patch/kernel/archive/rockchip64-7.1/mmc-sdio-skip-cmd11-voltage-switch-when-already-1v8.patch
@@ -0,0 +1,42 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: SuperKali <hello@superkali.me>
+Date: Wed, 3 Jun 2026 17:28:00 +0000
+Subject: mmc: sdio: skip the 1.8V switch when the host already signals at 1.8V
+
+A UHS-capable SDIO card can report S18A=1 even when the host signalling
+is already pinned to 1.8V. The core then issues CMD11 anyway, and with no
+real 3.3V to 1.8V transition the controller can stall on the voltage-switch
+clock update and fall back to high speed.
+
+Skip the switch when ios.signal_voltage is already 1.8V, the same guard
+mmc_sd_init_card() uses on the SD path. This lets an RTL8822CS on a
+fixed-1.8V rail come up at SDR104.
+
+Signed-off-by: SuperKali <hello@superkali.me>
+---
+ drivers/mmc/core/sdio.c | 8 +++++++-
+ 1 file changed, 7 insertions(+), 1 deletion(-)
+
+diff --git a/drivers/mmc/core/sdio.c b/drivers/mmc/core/sdio.c
+index 111111111111..222222222222 100644
+--- a/drivers/mmc/core/sdio.c
++++ b/drivers/mmc/core/sdio.c
+@@ -738,8 +738,14 @@ static int mmc_sdio_init_card(struct mmc_host *host, u32 ocr,
+ 	 * fails to check rocr & R4_18V_PRESENT,  but we still need to
+ 	 * try to init uhs card. sdio_read_cccr will take over this task
+ 	 * to make sure which speed mode should work.
++	 *
++	 * If the host already signals at 1.8V there is nothing to switch.
++	 * Some cards still report S18A=1, so guard CMD11 explicitly to avoid
++	 * a redundant voltage switch that can stall the controller. Mirrors
++	 * mmc_sd_init_card().
+ 	 */
+-	if (rocr & ocr & R4_18V_PRESENT) {
++	if ((rocr & ocr & R4_18V_PRESENT) &&
++	    host->ios.signal_voltage != MMC_SIGNAL_VOLTAGE_180) {
+ 		err = mmc_set_uhs_voltage(host, ocr_card);
+ 		if (err == -EAGAIN) {
+ 			mmc_sdio_pre_init(host, ocr_card, card);
+-- 
+Armbian
+
