Re: [PATCH v4 3/3] riscv: dts: spacemit: k3-deepcomputing-fml13v05: Enable SD card support
From: Sandie Cao
Date: Mon Nov 16 2026 - 04:52:31 EST
Hi, Yixun:
> From: "Yixun Lan"<dlan@xxxxxxxxxx>
> Date: Mon, Nov 16, 2026, 16:24
> >
> > In version 1 [1], the patch is cd-gpios = <&gpio K3_GPIO(88) GPIO_ACTIVE_LOW>;
> > but K3_GPIO(88) is fake cd-gpio, it's tied low on board.
> So, there is gpio cd pin, but didn't work? it's hard to say without
> the schematic, but can you check with hardware people, if it
> can be fixed with SoC's internal GPIO BIAS pullup or pulldown?
> So, gpio pin is able to reflect card insert status
>
>
I did check with our hardware team, and the answer is:
there is a CD net on the slot, routed to K3_GPIO(88), but it is hard-tied
to GND on the board. (v1 of this series used
cd-gpios = <&gpio K3_GPIO(88) GPIO_ACTIVE_LOW>, which of course
always reports "card present".) It is a direct connection to GND, not a
strap resistor, so the SoC's internal bias cannot pull it up.
> > So In version2 [2], the patch changes to non-removable.
> > And In version3 [3], add comment.
> >
> > The options: cd-gpio, non-removable, broken-cd all work on this board.
> >
> it works, but does not reflect the hardware status
>
> > We just need to detect the SD card when bootup, don't need the driver
> > to detect hot-plug event , non-removable should be suitable and faster,
> > we recomment to use non-removable.
> >
> No, this isn't a valid reason, it's use case should have nothing to do
> with the property.
>
> My point is: if the gpio cd pin still doesn't work, and the SD card is
> actually physical plugable, then should use "broken-cd".
The SD card slot is onboard [1], put the board into casing, there is no
way to plugin/ plugout the card. So there is no card-detect event that can
ever happen at runtime.
If you still prefer broken-cd, I will follow in next version.
[1]: https://deepcomputing.io/dc-roma-risc-v-mainboard-iii-unveiled-at-fosdem-powered-by-spacemit-k3-for-framework-laptop-13/
Best regards
Sandie