Re: [PATCH 2/2] arm64: dts: allwinner: h313: Add Amediatech X96Q

[Andre Przywara <andre.przywara@arm.com>]

On Mon, 15 Sep 2025 13:40:15 +0000
J. Neuschäfer <j.ne@posteo.net> wrote:

> On Fri, Sep 12, 2025 at 10:54:49AM +0100, Andre Przywara wrote:
> > On Fri, 12 Sep 2025 01:52:10 +0200
> > J. Neuschäfer via B4 Relay <devnull+j.ne.posteo.net@kernel.org> wrote:
> > 
> > Hi,
> > 
> > many thanks for posting the DT, I really wish more people would do that!
> >   
> > > From: "J. Neuschäfer" <j.ne@posteo.net>
> > > 
> > > The X96Q is a set-top box with an H313 SoC, AXP305 PMIC, 1 or 2 GiB RAM,
> > > 8 or 16 GiB eMMC flash, 2x USB A, Micro-SD, HDMI, Ethernet, audio/video
> > > output, and infrared input.
> > > 
> > >   https://x96mini.com/products/x96q-tv-box-android-10-set-top-box
> > > 
> > > Tested, works:
> > > - debug UART
> > > - status LED
> > > - USB ports in host mode
> > > - MicroSD
> > > - eMMC
> > > - recovery button hidden behind audio/video port
> > > - analog audio (line out)
> > > 
> > > Does not work:
> > > - Ethernet (requires AC200 MFD/EPHY driver)
> > > - analog video output (requires AC200 driver)
> > > - HDMI audio/video output
> > > 
> > > Untested:
> > > - "OTG" USB port in device mode
> > > - built-in IR receiver
> > > - external IR receiver
> > > - WLAN (requires out-of-tree XRadio driver)
> > > 
> > > Table of regulators on the downstream kernel, for reference:
> > > 
> > >  vcc-5v      1   15      0 unknown  5000mV     0mA  5000mV  5000mV
> > >     dcdca    0    0      0 unknown   900mV     0mA     0mV     0mV
> > >     dcdcb    0    0      0 unknown  1350mV     0mA     0mV     0mV
> > >     dcdcc    0    0      0 unknown   900mV     0mA     0mV     0mV
> > >     dcdcd    0    0      0 unknown  1500mV     0mA     0mV     0mV
> > >     dcdce    0    0      0 unknown  3300mV     0mA     0mV     0mV
> > >     aldo1    0    0      0 unknown  3300mV     0mA     0mV     0mV
> > >     aldo2    0    0      0 unknown   700mV     0mA     0mV     0mV
> > >     aldo3    0    0      0 unknown   700mV     0mA     0mV     0mV
> > >     bldo1    0    0      0 unknown  1800mV     0mA     0mV     0mV
> > >     bldo2    0    0      0 unknown  1800mV     0mA     0mV     0mV
> > >     bldo3    0    0      0 unknown   700mV     0mA     0mV     0mV
> > >     bldo4    0    0      0 unknown   700mV     0mA     0mV     0mV
> > >     cldo1    0    0      0 unknown  2500mV     0mA     0mV     0mV
> > >     cldo2    0    0      0 unknown   700mV     0mA     0mV     0mV
> > >     cldo3    0    0      0 unknown   700mV     0mA     0mV     0mV
> > > 
> > > Signed-off-by: J. Neuschäfer <j.ne@posteo.net>
> > > ---
> > >  arch/arm64/boot/dts/allwinner/Makefile             |   1 +
> > >  arch/arm64/boot/dts/allwinner/sun50i-h313-x96q.dts | 235 +++++++++++++++++++++
> > >  2 files changed, 236 insertions(+)
> > > 
> > > diff --git a/arch/arm64/boot/dts/allwinner/Makefile b/arch/arm64/boot/dts/allwinner/Makefile
> > > index 780aeba0f3a4e14d69c9602e37b8d299165507b9..2edfa7bf4ab31c4aa934da98e5e042edc9aaf600 100644
> > > --- a/arch/arm64/boot/dts/allwinner/Makefile
> > > +++ b/arch/arm64/boot/dts/allwinner/Makefile
> > > @@ -41,6 +41,7 @@ dtb-$(CONFIG_ARCH_SUNXI) += sun50i-h6-pine-h64-model-b.dtb
> > >  dtb-$(CONFIG_ARCH_SUNXI) += sun50i-h6-tanix-tx6.dtb
> > >  dtb-$(CONFIG_ARCH_SUNXI) += sun50i-h6-tanix-tx6-mini.dtb
> > >  dtb-$(CONFIG_ARCH_SUNXI) += sun50i-h313-tanix-tx1.dtb
> > > +dtb-$(CONFIG_ARCH_SUNXI) += sun50i-h313-x96q.dtb
> > >  dtb-$(CONFIG_ARCH_SUNXI) += sun50i-h616-bigtreetech-cb1-manta.dtb
> > >  dtb-$(CONFIG_ARCH_SUNXI) += sun50i-h616-bigtreetech-pi.dtb
> > >  dtb-$(CONFIG_ARCH_SUNXI) += sun50i-h616-orangepi-zero2.dtb
> > > diff --git a/arch/arm64/boot/dts/allwinner/sun50i-h313-x96q.dts b/arch/arm64/boot/dts/allwinner/sun50i-h313-x96q.dts
> > > new file mode 100644
> > > index 0000000000000000000000000000000000000000..9534eb03b89557f2545af5af7cf43390be722cf0
> > > --- /dev/null
> > > +++ b/arch/arm64/boot/dts/allwinner/sun50i-h313-x96q.dts
> > > @@ -0,0 +1,235 @@
> > > +// SPDX-License-Identifier: (GPL-2.0-or-later OR MIT)
> > > +/*
> > > + * Copyright (C) 2025 J. Neuschäfer <j.ne@posteo.net>
> > > + */
> > > +
> > > +/dts-v1/;
> > > +
> > > +#include "sun50i-h616.dtsi"
> > > +#include "sun50i-h616-cpu-opp.dtsi"
> > > +
> > > +#include <dt-bindings/gpio/gpio.h>
> > > +#include <dt-bindings/interrupt-controller/arm-gic.h>
> > > +#include <dt-bindings/input/linux-event-codes.h>
> > > +#include <dt-bindings/leds/common.h>
> > > +
> > > +/ {
> > > +	model = "X96Q";
> > > +	compatible = "amediatech,x96q", "allwinner,sun50i-h616";
> > > +
> > > +	aliases {
> > > +		mmc0 = &mmc0;
> > > +		mmc1 = &mmc1;
> > > +		mmc2 = &mmc2;  
> > 
> > We don't do mmc aliases in the upstream DTs. Long story, but you should
> > not need them. I guess you want to disagree ;-),  
> 
> I've seen mmc aliases before, and I've seen the annoying effects of an
> unstable probe order, so this is mostly of thing of habit.

Yes, as I said, it's a long story ;-) IIUC Rockchip does it, we
traditionally don't. There are arguments for both ways, and the decision
for sunxi was to not do it.

> > in this case U-Boot has
> > you covered, by adding the aliases during build time: just use
> > $fdtcontroladdr, as you should do anyway.  
> 
> I'm using a recent upstream U-Boot with a FIT image ('make image.fit' in
> the Linux source tree). In this case U-Boot does not add mmc aliases.
> Is there a special Kconfig option I need to enable in U-Boot?

U-Boot now automatically sync's the kernel DTs into the U-Boot tree, so
U-Boot is using the exact same DT for itself - and can pass that on to any
kernel. The trick is to just not load a DTB, but instead forward the
address of the one U-Boot is already using: $fdtcontroladdr.
If you are using EFI boot, this is done automatically, if you are using
the traditional booti method, you just change the last parameter:
=> booti $kernel_addr_r $ramdisk_addr_r:$filesize $fdtcontroladdr

I don't know how this works exactly with FIT images, so if you can
override the DTB in there or not include it in the first place. The
concept of bundling a particular DTB with a kernel is admittedly a long
standing practise in the embedded world, but does not make much sense
outside of it (in a single image, distribution world), and I like to treat
the embedded use case as a special case of a more generic approach.

> What do you mean by 'build time'? U-Boot's?

When U-Boot builds the DTB it uses itself, it will add some nodes, mostly
from arch/arm/dts/sunxi-u-boot.dtsi.
And because U-Boot relies on the MMC aliases internally, we add them
there. So if you pass on that DT, you will get the aliases.

Cheers,
Andre

> > > +		serial0 = &uart0;
> > > +	};  
> [...]
> > > +&mmc0 {
> > > +	vmmc-supply = <&reg_aldo1>;
> > > +	cd-gpios = <&pio 5 6 GPIO_ACTIVE_LOW>;	/* PF6 */
> > > +	disable-wp;
> > > +	bus-width = <4>;
> > > +	max-frequency = <150000000>;  
> > 
> > That line is already in the .dtsi file, so redundant.  
> 
> Good point, removing.
> >   
> > > +	status = "okay";
> > > +	/* µSD */  
> > 
> > If we really need this comment, it should be above, right after the
> > "&mmc0 {". And I wonder if it should be "microSD" instead.  
> 
> Good points, I'll do both.
> 
> > > +};
> > > +
> > > +&mmc1 {
> > > +	/* TODO: XRadio XR819 WLAN */  
> > 
> > Either you just keep the comment, an mention mmc1, but don't reference the
> > node, or you add the properties that you know of already, like
> > vmmc-supply, vqmmc-supply, mmc-pwrseq, bus-width, non-removable.
> > But this "empty reference with a comment" is somewhat odd.  
> 
> I'll change it to a pure comment for now.
> 
> > > +};
> > > +
> > > +&mmc2 {
> > > +	vmmc-supply = <&reg_aldo1>;
> > > +	vqmmc-supply = <&reg_bldo1>;
> > > +	non-removable;
> > > +	cap-mmc-hw-reset;
> > > +	mmc-ddr-1_8v;
> > > +	mmc-hs200-1_8v;
> > > +	bus-width = <8>;
> > > +	max-frequency = <100000000>;  
> > 
> > Are you sure you need that?  
> 
> I found it in the downstream DT. The eMMC probes fine without it, so
> I'll remove it. The eMMC datasheet (assuming I found the right one),
> claims a maximum of 200 MHz, so the 100 MHz restriction seems
> unnecessary.
> 
> >   
> > > +	status = "okay";
> > > +	/* eMMC */  
> > 
> > Please move that comment up.  
> 
> Will do.
> 
> 
> [...]
> > > +		regulators {
> > > +			reg_dcdca: dcdca {
> > > +				regulator-always-on;
> > > +				regulator-min-microvolt = <810000>;
> > > +				regulator-max-microvolt = <1100000>;
> > > +				regulator-name = "vdd-cpu";
> > > +			};
> > > +
> > > +			dcdcb {
> > > +				/* unused */
> > > +			};
> > > +
> > > +			reg_dcdcc: dcdcc {
> > > +				regulator-always-on;
> > > +				regulator-min-microvolt = <810000>;
> > > +				regulator-max-microvolt = <990000>;
> > > +				regulator-name = "vdd-gpu-sys";
> > > +			};
> > > +
> > > +			dcdcd {
> > > +				regulator-always-on;  
> > 
> > Why is this always on? What happens if you remove that line or turn it off?
> > For always-on regulators we either need a comment saying why, or, better,
> > have an explanatory regulator-name (like above).
> > Is that for DRAM, by any chance (1.5V for DDR3 chips)?  
> 
> The observed behavior is that the system halts when dcdcd is disabled.
> The relationship is a bit hard to determine without schematics, but the
> DDR3 chips (Micron MT41J256M4) do indeed use 1.5V. Since this is the
> only source of 1.5V, I think it's a safe guess that this is indeed the
> DRAM regulator.
> 
> > 
> > Cheers,
> > Andre  
> 
> Thanks for the review!
> J. Neuschäfer