On Thu, 2020-06-18 at 21:34 +0200, Michael Nazzareno Trimarchi wrote:
Looks like the pinctrl driver. This *ought* to work (with the caveat that I really ought to make two pinctrl setups and change between them according to the voltage, like the Linux DT and mtk-sd driver do).
--- a/arch/arm/dts/mt7623n-bananapi-bpi-r2.dts +++ b/arch/arm/dts/mt7623n-bananapi-bpi-r2.dts @@ -133,12 +133,14 @@ "MSDC0_DAT2", "MSDC0_DAT3", "MSDC0_DAT4", "MSDC0_DAT5", "MSDC0_DAT6", "MSDC0_DAT7"; input-enable;
bias-pull-up;
drive-strength = <2>;bias-pull-up = <3>; }; conf-clk { pins = "MSDC0_CLK";
bias-pull-down;
drive-strength = <2>;bias-pull-down = <3>; }; conf-rst {But it doesn't work. Partly because the U-Boot mtk pinctrl driver doesn't support the R0 R1 bits encoded in that <3>, and partly because it doesn't *even* get pullup/pulldown right at all; it sets the bit in the GPIO_PULLSEL registers but *not* the PUPD bit in the correct register for MSDC and other pins.
So change them manually, make it work, You have the board. I don't have any MediaTek board at the moment. Can you implement the missed part?
I think I'd rather let the MediaTek folks handle it, as it affects other SoCs too.
It turns out Linux has a pinctrl-mt7623.c driver which looks a lot like the U-Boot one, but *isn't* actually the one being used, because in Linux it only matches "mediatek,mt7623-moore-pinctrl" which isn't in this board's DT.
Linux also has a pinctrl-mt2701.c driver that matches the "mediatek,mt7623-pinctrl" that *is* on this board, and which *does* work correctly.
So I don't know if we just have the wrong driver in U-Boot. Or if the mediatek,mt7623-moore-pinctrl compatible string is a temporary name for the *same* hardware, to invoke the new driver that isn't complete yet so it isn't being used in Linux... but is in U-Boot?
Sean?