Devicetree
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Bob Hui" <bg9oxa@163.com>
Cc: devicetree@vger.kernel.org, robh@kernel.org, conor+dt@kernel.org
Subject: Re: [PATCH v9 2/2] arm64: dts: rockchip: Add ALIENTEK QuarkPi-CA2
Date: Wed, 21 Oct 2026 14:33:29 +0000	[thread overview]
Message-ID: <sashiko-outbox-180381@kernel.org> (raw)
In-Reply-To: <20261021-quarkpi-ca2-v9-2-b75006b22bc5@163.com>

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [High] arm64: dts: rockchip: missing rknn_mmu_0/1/2 enablement breaks NPU open()
- [High] arm64: dts: rockchip: missing domain-supply on pd_gpu and pd_npu
--

commit e4632e8324e154e7c3c642a2039af079523bc6b6
Author: Bob Hui <bg9oxa@163.com>

arm64: dts: rockchip: Add ALIENTEK QuarkPi-CA2

This commit adds the device tree for the ALIENTEK QuarkPi-CA2, an RK3588S
based board, enabling its onboard storage, regulators, display outputs,
networking, USB, audio, GPU, and NPU peripherals.

> diff --git a/arch/arm64/boot/dts/rockchip/rk3588s-quarkpi-ca2.dts b/arch/arm64/boot/dts/rockchip/rk3588s-quarkpi-ca2.dts
> new file mode 100644
> index 0000000000000..68ee2c0ad9d80
> --- /dev/null
> +++ b/arch/arm64/boot/dts/rockchip/rk3588s-quarkpi-ca2.dts

[ ... ]

> +&gpu {
> +	mali-supply = <&vdd_gpu_s0>;
> +	status = "okay";
> +};

[Severity: High]
Should &pd_gpu and &pd_npu also specify domain-supply = <&vdd_gpu_s0> and
domain-supply = <&vdd_npu_s0>?

In drivers/pmdomain/rockchip/pm-domains.c, rk3588_pm_domains marks both
RK3588_PD_GPU (&pd_gpu) and RK3588_PD_NPU (&pd_npu) with
need_regulator = true, and rockchip_pm_add_one_domain() powers those domains
off during PMU probe. When a consumer attaches or resumes via
genpd_power_on() -> rockchip_pd_power_on() -> rockchip_pd_regulator_enable(),
the PMU driver looks up domain-supply on the power-domain node:

drivers/pmdomain/rockchip/pm-domains.c:rockchip_pd_regulator_enable() {
	if (!pd->info->need_regulator)
		return 0;

	if (IS_ERR_OR_NULL(pd->supply)) {
		pd->supply = devm_of_regulator_get(pmu->dev, pd->node, "domain");

		if (IS_ERR(pd->supply))
			return PTR_ERR(pd->supply);
	}

	return regulator_enable(pd->supply);
}

In arch/arm64/boot/dts/rockchip/rk3588s-quarkpi-ca2.dts, vdd_gpu_s0 is only
regulator-boot-on (without regulator-always-on) and both vdd_gpu_s0 and
vdd_npu_s0 configure regulator-off-in-suspend. Without domain-supply on
&pd_gpu and &pd_npu, the regulator core falls back to a dummy regulator with
a warning and rockchip_pd_power() attempts the PMU power-on handshake before
vdd_gpu_s0 or vdd_npu_s0 has been enabled (such as when panthor probes as a
module after unused boot-on regulators are disabled, or across system suspend
and resume).

> +
> +&rknn_core_0 {
> +	npu-supply = <&vdd_npu_s0>;
> +	sram-supply = <&vdd_npu_mem_s0>;
> +	status = "okay";
> +};
> +
> +&rknn_core_1 {
> +	npu-supply = <&vdd_npu_s0>;
> +	sram-supply = <&vdd_npu_mem_s0>;
> +	status = "okay";
> +};
> +
> +&rknn_core_2 {
> +	npu-supply = <&vdd_npu_s0>;
> +	sram-supply = <&vdd_npu_mem_s0>;
> +	status = "okay";
> +};

[Severity: High]
Will opening the NPU accel device fail with -ENODEV unless &rknn_mmu_0,
&rknn_mmu_1, and &rknn_mmu_2 are also enabled?

In arch/arm64/boot/dts/rockchip/rk3588-base.dtsi, rknn_core_0, rknn_core_1,
and rknn_core_2 reference iommus = <&rknn_mmu_0>, <&rknn_mmu_1>, and
<&rknn_mmu_2>, which default to status = "disabled":

	rknn_mmu_0: iommu@fdab9000 {
		...
		power-domains = <&power RK3588_PD_NPUTOP>;
		status = "disabled";
	};

Because arch/arm64/boot/dts/rockchip/rk3588s-quarkpi-ca2.dts enables the
three rknn_core_* nodes without setting status = "okay" on &rknn_mmu_0,
&rknn_mmu_1, and &rknn_mmu_2, of_iommu_xlate() in drivers/iommu/of_iommu.c
returns -ENODEV:

	if (!of_device_is_available(iommu_spec->np))
		return -ENODEV;

This leaves the NPU core devices without an attached IOMMU. When userspace
opens the NPU device, rocket_open() in drivers/accel/rocket/rocket_drv.c
calls rocket_iommu_domain_create() -> iommu_paging_domain_alloc() ->
__iommu_paging_domain_alloc_flags():

drivers/accel/rocket/rocket_drv.c:rocket_open() {
	...
	rocket_priv->domain = rocket_iommu_domain_create(rdev->cores[0].dev);
	if (IS_ERR(rocket_priv->domain)) {
		ret = PTR_ERR(rocket_priv->domain);
		goto err_free;
	}
	...
}

which checks !dev_has_iommu(dev) and fails unconditionally with -ENODEV.

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261021-quarkpi-ca2-v9-0-b75006b22bc5@163.com?part=2

      reply	other threads:[~2026-10-21 14:33 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-10-21 14:05 [PATCH v9 0/2] arm64: dts: rockchip: Add ALIENTEK QuarkPi-CA2 Bob Hui
2026-10-21 14:05 ` [PATCH v9 1/2] dt-bindings: arm: " Bob Hui
2026-10-22  2:58   ` Rob Herring
2026-10-21 14:05 ` [PATCH v9 2/2] arm64: dts: " Bob Hui
2026-10-21 14:33   ` sashiko-bot [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=sashiko-outbox-180381@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=bg9oxa@163.com \
    --cc=conor+dt@kernel.org \
    --cc=devicetree@vger.kernel.org \
    --cc=robh@kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox