arm64: dts: allwinner: a523: x96qpro+: overvolting eMMC
Subsystems (2)
PATCH 1/1 arm64: dts: allwinner: a523: x96qpro+: overvolting eMMC
On the X96QPro+ TV box, the eMMC wasn't working properly: the chip would be detected, but most requests failed, leaving it unusable. Some "solutions" were circulating downstream, one "popular" one is severely limiting the clock frequency, with a drastic performance impact.
Instead slightly bumping the I/O voltage on PortC from the canonical 1.8V to 1.9V seemed to do the trick: the eMMC now works and reaches the expected 120MB/s performance.
Adjust the respective regulator voltage in the .dts to enable eMMC operation on that device.
---
Hi,
I am not super happy with this patch, but it seems to be a pragmatic and the least hacky solution I have seen.
Cheers,
Andre
arch/arm64/boot/dts/allwinner/sun55i-h728-x96qpro+.dts | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
arch/arm64/boot/dts/allwinner/sun55i-h728-x96qpro+.dts +2 -2
@@ -194,8 +194,8 @@ reg_bldo4: bldo4 { reg_cldo1: cldo1 { regulator-always-on; - regulator-min-microvolt = <1800000>; - regulator-max-microvolt = <1800000>; + regulator-min-microvolt = <1900000>; + regulator-max-microvolt = <1900000>; regulator-name = "vcc-codec-sd"; };
Signature · 1 line
2.46.4
[email protected]
AI reviewThank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [High] Bumping the eMMC IO regulator (`reg_cldo1`) to 1.9V inadvertently misconfigures the SoC's IO bank…
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Bumping the eMMC IO regulator (reg_cldo1) to 1.9V inadvertently misconfigures the SoC's IO bank (Port C) into 3.3V power mode instead of 1.8V mode, creating a logic threshold mismatch. --
--- Patch [1]: [PATCH] arm64: dts: allwinner: a523: x96qpro+: overvolting eMMC --- commit e4c261c388d6e406babdf0d56ac61d104d537230
Author: Andre Przywara <[email protected]> Date: Mon Jul 20 21:38:32 2026 +0000
arm64: dts: allwinner: a523: x96qpro+: overvolting eMMC
This commit changes the eMMC IO regulator voltage on PortC from 1.8V to 1.9V for the X96QPro+ device. The goal is to improve eMMC stability and performance without drastically limiting the clock frequency.
Quoted 3 lines
arch/arm64/boot/dts/allwinner/sun55i-h728-x96qpro+.dts +0
-0
[ ... ]
arch/arm64/boot/dts/allwinner/sun55i-h728-x96qpro+.dts reg_bldo4: bldo4 { · line 198
Quoted 8 lines
arch/arm64/boot/dts/allwinner/sun55i-h728-x96qpro+.dts +2
-2@@ -194,8 +194,8 @@ reg_bldo4: bldo4 { reg_cldo1: cldo1 { regulator-always-on; - regulator-min-microvolt = <1800000>; - regulator-max-microvolt = <1800000>; + regulator-min-microvolt = <1900000>; + regulator-max-microvolt = <1900000>;
[Severity: High]
Does bumping this regulator to 1.9V inadvertently force the SoC's IO bank (Port C) into 3.3V power mode?
Looking at sunxi_pinctrl_set_io_bias_cfg() in drivers/pinctrl/sunxi/pinctrl-sunxi.c, the driver uses a strict check when evaluating the voltage:
case BIAS_VOLTAGE_PIO_POW_MODE_SEL:
val = uV <= 1800000 ? 1 : 0;Since 1900000 is greater than 1800000, this evaluates to 0 and clears the PIO_POW_MOD_SEL register bit, incorrectly configuring the IO bank for 3.3V mode while being powered at 1.9V.
Could this mismatch alter input receiver thresholds and drive strengths, potentially causing signal integrity issues or long-term hardware degradation?
Quoted 2 lines
arch/arm64/boot/dts/allwinner/sun55i-h728-x96qpro+.dts +0
-0regulator-name = "vcc-codec-sd"; };
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Quoted 2 lines · On Mon, 20 Jul 2026 21:50:35 +0000 [email protected] wrote:
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Bumping the eMMC IO regulator (reg_cldo1) to 1.9V inadvertently misconfigures the SoC's IO bank (Port C) into 3.3V power mode instead of 1.8V mode, creating a logic threshold mismatch.
Huh, that's interesting, but might actually point to the real root
cause: it seems like the meaning of the bits in the withstand register is inverted on the A523: 0 encodes 1.8V, a 1 encodes 3.3V. But that's the opposite of the H6 definition, which we use in the kernel for the A523 as well. Which means they were wrongly programmed before, but now the wrong voltage selection turns out to trigger the really correct setting.
So I will try playing around with those registers instead.
Cheers,
Andre.
Quoted 48 lines
--
--- Patch [1]: [PATCH] arm64: dts: allwinner: a523: x96qpro+: overvolting eMMC --- commit
e4c261c388d6e406babdf0d56ac61d104d537230Author: Andre Przywara <[email protected]> Date: Mon Jul 20 21:38:32 2026 +0000arm64: dts: allwinner: a523: x96qpro+: overvolting eMMC
This commit changes the eMMC IO regulator voltage on PortC from 1.8V to 1.9V for the X96QPro+ device. The goal is to improve eMMC stability and performance without drastically limiting the clock frequency.
Quoted 3 lines
arch/arm64/boot/dts/allwinner/sun55i-h728-x96qpro+.dts +0
-0[ ... ]
arch/arm64/boot/dts/allwinner/sun55i-h728-x96qpro+.dts reg_bldo4: bldo4 { · line 198
Quoted 8 lines
arch/arm64/boot/dts/allwinner/sun55i-h728-x96qpro+.dts +2
-2@@ -194,8 +194,8 @@ reg_bldo4: bldo4 { reg_cldo1: cldo1 { regulator-always-on; - regulator-min-microvolt = <1800000>; - regulator-max-microvolt = <1800000>; + regulator-min-microvolt = <1900000>; + regulator-max-microvolt = <1900000>;[Severity: High]
Does bumping this regulator to 1.9V inadvertently force the SoC's IO bank (Port C) into 3.3V power mode?Looking at
sunxi_pinctrl_set_io_bias_cfg()indrivers/pinctrl/sunxi/pinctrl-sunxi.c, the driver uses a strict check when evaluating the voltage:case BIAS_VOLTAGE_PIO_POW_MODE_SEL: val = uV <= 1800000 ? 1 : 0;Since 1900000 is greater than 1800000, this evaluates to 0 and clears the
PIO_POW_MOD_SELregister bit, incorrectly configuring the IO bank for 3.3V mode while being powered at 1.9V.Could this mismatch alter input receiver thresholds and drive strengths, potentially causing signal integrity issues or long-term hardware degradation?
Quoted 2 lines
arch/arm64/boot/dts/allwinner/sun55i-h728-x96qpro+.dts +0
-0regulator-name = "vcc-codec-sd"; };
On Mon, 20 Jul 2026 23:38:32 +0200
Andre Przywara <[email protected]> wrote:
Hi,
Quoted 8 lines
On the X96QPro+ TV box, the eMMC wasn't working properly: the chip would be detected, but most requests failed, leaving it unusable. Some "solutions" were circulating downstream, one "popular" one is severely limiting the clock frequency, with a drastic performance impact.
Instead slightly bumping the I/O voltage on PortC from the canonical 1.8V to 1.9V seemed to do the trick: the eMMC now works and reaches the expected 120MB/s performance.
So for the records: please disregard this patch, it was more voodoo than a proper fix.
As Sashiko pointed out, this slight voltage bump changes the GPIO bank withstand voltage, which turned out to be the actual culprit: the encoding was wrong, and the new voltage here now changed the value to inadvertently use the right bit setting.
I will send another patch to fix this properly.
Cheers,
Andre
Quoted 32 lines
Adjust the respective regulator voltage in the .dts to enable eMMC operation on that device.
Signed-off-by Andre Przywara <[email protected]>---
Hi,I am not super happy with this patch, but it seems to be a pragmatic and the least hacky solution I have seen.
Cheers,
Andrearch/arm64/boot/dts/allwinner/sun55i-h728-x96qpro+.dts | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)arch/arm64/boot/dts/allwinner/sun55i-h728-x96qpro+.dts +2
-2@@ -194,8 +194,8 @@ reg_bldo4: bldo4 { reg_cldo1: cldo1 { regulator-always-on; - regulator-min-microvolt = <1800000>; - regulator-max-microvolt = <1800000>; + regulator-min-microvolt = <1900000>; + regulator-max-microvolt = <1900000>; regulator-name = "vcc-codec-sd"; };