Message ID | E5A1A1E7-449F-4161-87B9-7A6CA0CAA42C@kohlschutter.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | [v6] arm64: dts: rockchip: Fix SD card init on rk3399-nanopi4 | expand |
> Am 18.07.2022 um 14:04 schrieb Christian Kohlschütter <christian@kohlschutter.com>: > > mmc/SD-card initialization may fail on NanoPi R4S with > "mmc1: problem reading SD Status register" / > "mmc1: error -110 whilst initialising SD card" > either on cold boot or after a reboot. ... Walking back on my claim in the commit message that no further patches are needed for the u-boot integration to work. Other than that, the actual patch is unchanged.
Am 18.07.2022 um 14:05 schrieb Christian Kohlschütter <christian@kohlschutter.com>: > >> Am 18.07.2022 um 14:04 schrieb Christian Kohlschütter <christian@kohlschutter.com>: >> >> mmc/SD-card initialization may fail on NanoPi R4S with >> "mmc1: problem reading SD Status register" / >> "mmc1: error -110 whilst initialising SD card" >> either on cold boot or after a reboot. > ... > Walking back on my claim in the commit message that no further patches are needed for the u-boot integration to work. > Other than that, the actual patch is unchanged. I can no longer verify that the v6 patch actually increases use_count to 2. At this point, I don't have a purely device tree-based fix that works reliably, therefore necessitating the code changes in regulator/core.c. If anyone of you knows a way to reliably increase vcc3v0_sd's use_count ("cat /sys/kernel/debug/regulator/vcc3v0_sd/use_count") to at least 2, then we can remove the regulator-always-on statement.
diff --git a/arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi b/arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi index 8c0ff6c96e03..38507a6e3046 100644 --- a/arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi +++ b/arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi @@ -61,7 +61,17 @@ vcc1v8_s3: vcc1v8-s3 { regulator-min-microvolt = <1800000>; regulator-max-microvolt = <1800000>; regulator-name = "vcc1v8_s3"; - vin-supply = <&vcc_1v8>; + + /* + * Workaround to skip setting gpio0 RK_PA1 to LOW upon reboot, + * which may freeze the system. + * + * Adding a reference to vcc3v0_sd increases its num_users + * count to 2, preventing deactivation since this regulator is + * marked "always-on". + */ + // vin-supply = <&vcc_1v8>; // actual supply + vin-supply = <&vcc3v0_sd>; }; vcc3v0_sd: vcc3v0-sd { @@ -70,7 +80,6 @@ vcc3v0_sd: vcc3v0-sd { gpio = <&gpio0 RK_PA1 GPIO_ACTIVE_HIGH>; pinctrl-names = "default"; pinctrl-0 = <&sdmmc0_pwr_h>; - regulator-always-on; regulator-min-microvolt = <3000000>; regulator-max-microvolt = <3000000>; regulator-name = "vcc3v0_sd";
mmc/SD-card initialization may fail on NanoPi R4S with "mmc1: problem reading SD Status register" / "mmc1: error -110 whilst initialising SD card" either on cold boot or after a reboot. Moreover, the system would also sometimes hang upon reboot. This is caused by vcc3v0-sd's "regulator-always-on", which triggers an erroneous double-initialization of the regulator. This causes voltage fluctuations that can, depending on timing, prevent the SD card from initializing correctly. Adding some liberal delay via "off-on-delay-us" is ineffective since that codepath is skipped as long "regulator-always-on" is set. Removing "regulator-always-on" alone is not sufficient because that would allow the system to set GPIO0_A1 to LOW upon reboot, which may cause the system to hang. In order to allow the system to set GPIO0_A1 to HIGH upon initialization but prevent it from changing it back to LOW, this patch increases the usage count of vcc3v0-sd from 1 to 2, whereas the additional reference, "vcc1v8_s3", is marked as "always-on", causing permanent retention. Signed-off-by: Christian Kohlschütter <christian@kohlschutter.com> --- arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-)