Re: Results of more testing on Orange PI zero/one boards

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Mon, Sep 11, 2023 at 11:02:30PM +0100, Robert Swindells wrote:
> 
> Dave Tyson <dtyson%anduin.org.uk@localhost> wrote:
> > The orangepi-one board fails to boot successfully under NetBSD-10 or
> > NetBSD-current. The problem is due to the kernel not finding a root
> > device as the sdcard is not detected/attached.
> >
> > There is a workaround: If the orangepi-one dtb on the above media is
> > replaced by the dtb supplied under NetBSD-9 then the board boots 
> > correctly. Looking at the differences between the decompiled dtbs
> > from NetBSD-9 and NetBSD-10 there are a few significant differences
> > and I wonder if something is tripping up the code which parses the
> > in-core dtb to enumerate the device tree. If someone could point me
> > to where in the source tree this code lives then I could try adding
> > some debug code to see what is happening...
> 
> I think it is unlikely there are bugs in the code to parse dtb files,
> it is used for a lot of other boards.
> 
> The dts sources look wrong to me, the sun8i-h3-orangepi-one.dts patch
> file references a local sun8i-h3.dtsi file that has been deleted.
> 
> The built dtb file under -current ends up with all the mmc nodes
> disabled, could you try the patch below.
> 
> There seem to be a number of other nodes disabled too, like emac and
> usb.
> 
> Index: sun8i-h3-orangepi-one.dts
> ===================================================================
> RCS file: /cvsroot/src/sys/arch/arm/dts/sun8i-h3-orangepi-one.dts,v
> retrieving revision 1.3
> diff -u -r1.3 sun8i-h3-orangepi-one.dts
> --- sun8i-h3-orangepi-one.dts   30 Nov 2017 21:39:35 -0000      1.3
> +++ sun8i-h3-orangepi-one.dts   11 Sep 2023 21:58:41 -0000
> @@ -27,5 +27,11 @@
>   */
>  
>  #include "../../../external/gpl2/dts/dist/arch/arm/boot/dts/sun8i-h3-orangepi-one.dts"
> -#include "sun8i-h3.dtsi"
>  
> +/ {
> +       soc {
> +               mmc@1c0f000 {
> +                       status = "okay";
> +               };
> +       };
> +};

Good catch, removing the sun8i-h3.dtsi could do it
Forcing the okay status on the mmc shouldn't be needed because the
sun8i-h3-orangepi-one.dts file does it.

So maybe sun8i-h3-orangepi-one.dts should just be deleted from
src/sys/arch/arm/dts/ and let the build pick up the one from src/sys/external
instead.

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--