Re: [PATCH 8/8] riscv: dts: spacemit: add SpacemiT K3 Pico-ITX board device tree

[Bo Gan]

Hi Guodong,

On 12/16/25 23:13, Guodong Xu wrote:
> On Tue, Dec 16, 2025 at 10:33 PM Heinrich Schuchardt
> <heinrich.schuchardt@xxxxxxxxxxxxx> wrote:
> > On 12/16/25 14:32, Guodong Xu wrote:
> > > K3 Pico-ITX is a 2.5-inch single-board computer equipted with a SpacemiT
> > > K3 SoC.
> > >
> > > This minimal device tree enables booting into a serial console with UART
> > > output.
> > >
> > > Signed-off-by: Guodong Xu <guodong@xxxxxxxxxxxx>
> > > ---
> > >    arch/riscv/boot/dts/spacemit/Makefile        |  1 +
> > >    arch/riscv/boot/dts/spacemit/k3-pico-itx.dts | 25 +++++++++++++++++++++++++
> > >    2 files changed, 26 insertions(+)
> > >
> > > diff --git a/arch/riscv/boot/dts/spacemit/Makefile b/arch/riscv/boot/dts/spacemit/Makefile
> > > index 95889e7269d1bae679b28cd053e1b0a23ae6de68..7e2b877025718113a0e31917eadf7562f488d825 100644
> > > --- a/arch/riscv/boot/dts/spacemit/Makefile
> > > +++ b/arch/riscv/boot/dts/spacemit/Makefile
> > > @@ -4,3 +4,4 @@ dtb-$(CONFIG_ARCH_SPACEMIT) += k1-milkv-jupiter.dtb
> > >    dtb-$(CONFIG_ARCH_SPACEMIT) += k1-musepi-pro.dtb
> > >    dtb-$(CONFIG_ARCH_SPACEMIT) += k1-orangepi-r2s.dtb
> > >    dtb-$(CONFIG_ARCH_SPACEMIT) += k1-orangepi-rv2.dtb
> > > +dtb-$(CONFIG_ARCH_SPACEMIT) += k3-pico-itx.dtb
> > > diff --git a/arch/riscv/boot/dts/spacemit/k3-pico-itx.dts b/arch/riscv/boot/dts/spacemit/k3-pico-itx.dts
> > > new file mode 100644
> > > index 0000000000000000000000000000000000000000..0f9d04dd352f5331e82599285113b86af5b09ebe
> > > --- /dev/null
> > > +++ b/arch/riscv/boot/dts/spacemit/k3-pico-itx.dts
> > > @@ -0,0 +1,25 @@
> > > +// SPDX-License-Identifier: (GPL-2.0 OR MIT)
> > > +/*
> > > + * Copyright (c) 2025 SpacemiT (Hangzhou) Technology Co. Ltd
> > > + * Copyright (c) 2025 Guodong Xu <guodong@xxxxxxxxxxxx>
> > > + */
> > > +
> > > +#include "k3.dtsi"
> > > +
> > > +/ {
> > > +     model = "SpacemiT K3 Pico-ITX";
> > > +     compatible = "spacemit,k3-pico-itx", "spacemit,k3";
> > > +
> > > +     chosen {
> > > +             stdout-path = "serial0";
> > > +     };
> > > +
> > > +     memory@100200000 {
> > > +             device_type = "memory";
> > > +             reg = <0x1 0x00200000 0x3 0xffe00000>;
> >
> > Shouldn't the reserved memory be described as no-map /reserved-memory
> > nodes instead?
>
> Hi Heinrich,
>
> Yes, you are right. Using a reserved-memory node with the no-map property is
> the correct approach. I will update this in the next version to explicitly
> reserve the first 2MB.
>
> > I would assume that 0x1,0000,0000 is the location of OpenSBI.
> > What is at 0x3,ffe0,0000?
>
> Yes, confirmed, 0x100000000 is the start of OpenSBI. The size 0x3ffe00000 in
> my original patch was the result of manually subtracting the 2MB OpenSBI
> offset from the total 16GB memory.
>
> I will correct this in v2 to show the actual memory and put the first 2MB in
> reserved-memory {} for opensbi.

I don't think this is the right approach. The Linux kernel shouldn't care
about the size of the underlying SBI implementation. The size of OpenSBI
also depends on the version and features enabled. Thus, you should just
let the OpenSBI or whatever SBI fix up the device-tree for you. OpenSBI
already does that today. It inserts one or more reserved-memory node and
carves out the memory region of its own. If there's some bootloader in
between SBI <-> Linux, such as U-boot, then it's responsible for copying
over those reserved regions from OpenSBI into the device-tree Linux uses
(U-boot already does). Thus, there's no need to manually specify it.

Bo

> Thanks for the review.
>
> BR,
> Guodong Xu
>
>
> > Best regards
> >
> > Heinrich
> >
> > > +     };
> > > +};
> > > +
> > > +&uart0 {
> > > +     status = "okay";
> > > +};