When trying to boot a nanopi-m4 board with an SDHC-class uSD card, the boot
comes to a full stop shortly after initializing the mmc subsystem. The boot
can be cajoled into continuing if, after waiting a minute or so, the uSD
card is ejected and re-inserted. Waiting a minute or so before ejecting and
re-inserting the uSD card is crucial since the boot will not continue if
the card is ejected/re-inserted too soon after the boot has stopped.
The nanopi-m4 has a uSD card and an optional eMMC module, either of which
can be used for booting. In my case I don't have the optional eMMC module,
therefore I'm booting from the uSD card. When booting from the uSD card,
its regulator needs to be enabled at boot.
Curiously, this should have been an issue from day one, but it only started
to become a problem after commit 98e48cd9283d ("regulator: core: resolve
supply for boot-on/always-on regulators") was merged. Additionally, by
coincidence, I happened to be using an SDHC-class card in my device, and
saw the failure. However, if I use an SDXC-class uSD card the problem does
not occur.
Much thanks to Mark Brown and Javier Martinez Canillas for their assistance
on irc!
Signed-off-by: Trevor Woerner <[email protected]>
---
arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi | 1 +
1 file changed, 1 insertion(+)
diff --git a/arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi b/arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi
index 8c0ff6c96e03..5cf02e2ef9b3 100644
--- a/arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi
+++ b/arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi
@@ -71,6 +71,7 @@ vcc3v0_sd: vcc3v0-sd {
pinctrl-names = "default";
pinctrl-0 = <&sdmmc0_pwr_h>;
regulator-always-on;
+ regulator-boot-on;
regulator-min-microvolt = <3000000>;
regulator-max-microvolt = <3000000>;
regulator-name = "vcc3v0_sd";
--
2.30.0.rc0
On Wed, Oct 06, 2021 at 12:08:17PM -0400, Trevor Woerner wrote:
> @@ -71,6 +71,7 @@ vcc3v0_sd: vcc3v0-sd {
> pinctrl-names = "default";
> pinctrl-0 = <&sdmmc0_pwr_h>;
> regulator-always-on;
> + regulator-boot-on;
This shouldn't be required for devices that can read their status, the
fact that it's always on should cause it to be powered up if it wasn't
already when applying constraints. I'm therefore confused as to what
this change is doing and there's probably something else going on here.