Devicetree
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Krzysztof Kozlowski" <krzysztof.kozlowski@oss.qualcomm.com>,
	devicetree@vger.kernel.org
Cc: robh@kernel.org, conor+dt@kernel.org
Subject: Re: [PATCH] arm64: dts: mediatek: Remove undocumented, unapproved aliases
Date: Tue, 17 Nov 2026 13:19:27 +0000	[thread overview]
Message-ID: <sashiko-outbox-218859@kernel.org> (raw)
In-Reply-To: <20261117130556.234785-2-krzysztof.kozlowski@oss.qualcomm.com>

Thank you for your contribution! Sashiko AI review found 4 potential
issues to consider:
- [High] arm64: dts: mediatek: removing gce aliases causes unclocked
  MMIO in cmdq_init()
- [High] arm64: dts: mediatek: removing ovl_adaptor aliases breaks
  ovl_adaptor_comp_init()
- [High] arm64: dts: mediatek: removing ddp and mutex aliases breaks
  mtk_drm_probe()
- [Low] arm64: dts: mediatek: mt8173: incomplete removal of undocumented
  mdp-* aliases
--

commit 0c6075438efac906bb8687313653fb8c36767eef
Author: Krzysztof Kozlowski <krzysztof.kozlowski@oss.qualcomm.com>

arm64: dts: mediatek: Remove undocumented, unapproved aliases

This commit removes undocumented display and mailbox device tree aliases from
several MediaTek SoC and board dtsi files.

[ ... ]

> diff --git a/arch/arm64/boot/dts/mediatek/mt8173.dtsi b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> index 6ec50125cea02..868b0fb844499 100644
> --- a/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> +++ b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> @@ -22,22 +22,6 @@ / {
>  	#size-cells = <2>;
>  
>  	aliases {
> -		ovl0 = &ovl0;
> -		ovl1 = &ovl1;
> -		rdma0 = &rdma0;
> -		rdma1 = &rdma1;
> -		rdma2 = &rdma2;
> -		wdma0 = &wdma0;
> -		wdma1 = &wdma1;
> -		color0 = &color0;
> -		color1 = &color1;
> -		split0 = &split0;
> -		split1 = &split1;
> -		dpi0 = &dpi0;
> -		dsi0 = &dsi0;
> -		dsi1 = &dsi1;
> -		mdp-rdma0 = &mdp_rdma0;
> -		mdp-rdma1 = &mdp_rdma1;
>  		mdp-rsz0 = &mdp_rsz0;
>  		mdp-rsz1 = &mdp_rsz1;
>  		mdp-rsz2 = &mdp_rsz2;

[Severity: Low]
This isn't a bug, but should mdp-rsz0..2, mdp-wdma0, and mdp-wrot0..1 also be
dropped here?  Like mdp-rdma0 and mdp-rdma1, they are undocumented custom
aliases that are unused by drivers/media/platform/mediatek/mdp/.

[ ... ]

> diff --git a/arch/arm64/boot/dts/mediatek/mt8188.dtsi b/arch/arm64/boot/dts/mediatek/mt8188.dtsi
> index 14a320f0c70cb..f14b23dd57aa7 100644
> --- a/arch/arm64/boot/dts/mediatek/mt8188.dtsi
> +++ b/arch/arm64/boot/dts/mediatek/mt8188.dtsi
> @@ -23,39 +23,6 @@ / {
>  	#address-cells = <2>;
>  	#size-cells = <2>;
>  
> -	aliases {
> -		dp-intf0 = &dp_intf0;
> -		dp-intf1 = &dp_intf1;
> -		dpi1 = &dpi1;
> -		dsc0 = &dsc0;
> -		ethdr0 = &ethdr0;
> -		gce0 = &gce0;
> -		gce1 = &gce1;

[Severity: High]
Does removing gce0 and gce1 on MT8188 and MT8195 cause cmdq_init() to perform
MMIO writes with the GCE clocks disabled?

On both SoCs, cmdq->pdata->gce_num is 2, so cmdq_get_clocks() looks up the
clocks via the "gce" alias and silently skips nodes with a negative alias_id:

drivers/mailbox/mtk-cmdq-mailbox.c:cmdq_get_clocks() {
	...
	for_each_child_of_node_scoped(parent, node) {
		int alias_id = of_alias_get_id(node, gce_name);

		if (alias_id < 0 || alias_id >= cmdq->pdata->gce_num)
			continue;

		clks = &cmdq->clocks[alias_id];
	...
	return 0;
}

Without gce0 and gce1, cmdq_get_clocks() returns 0 with cmdq->clocks[].clk
left NULL.  clk_bulk_prepare() and clk_bulk_enable() then succeed as no-ops
before cmdq_init() and cmdq_runtime_resume() write to GCE registers, hanging
the bus or triggering an SError on boot.

> -		merge0 = &merge0;
> -		merge1 = &merge1;
> -		merge2 = &merge2;
> -		merge3 = &merge3;
> -		merge4 = &merge4;
> -		merge5 = &merge5;
> -		mutex0 = &mutex0;
> -		mutex1 = &mutex1;

[Severity: High]
Does removing the mutex and DDP component aliases break multi-instance
component matching and leak device_node references in mtk_drm_probe()?

In mtk_drm_probe(), MTK_DISP_MUTEX nodes are matched with:

drivers/gpu/drm/mediatek/mtk_drm_drv.c:mtk_drm_probe() {
	...
		if (comp_type == MTK_DISP_MUTEX) {
			int id;

			id = of_alias_get_id(node, "mutex");
			if (id < 0 || id == private->data->mmsys_id) {
				private->mutex_node = of_node_get(node);
	...
}

When mutex0 and mutex1 are removed on MT8188 and MT8195, id is -ENODEV for
both sibling mutex nodes, so VDOSYS0 (mmsys_id 0) matches mutex0 and then
overwrites private->mutex_node with mutex1, binding VDOSYS0 to VDOSYS1's
hardware disp-mutex and leaking mutex0's OF node reference.

Similarly, mtk_ddp_comp_get_id() accepts id < 0 as matching the first entry
of that component type in mtk_ddp_matches[]:

drivers/gpu/drm/mediatek/mtk_ddp_comp.c:mtk_ddp_comp_get_id() {
	int id = of_alias_get_id(node, mtk_ddp_comp_stem[comp_type]);
	int i;

	for (i = 0; i < ARRAY_SIZE(mtk_ddp_matches); i++) {
		if (comp_type == mtk_ddp_matches[i].type &&
		    (id < 0 || id == mtk_ddp_matches[i].alias_id))
			return i;
	}
	...
}

Without the DDP aliases (ovl*, ovl-2l*, rdma*, color*, dpi*, dsi*, dp-intf*,
merge*, etc.), secondary component instances collapse onto instance 0's
comp_id, overwriting private->comp_node[comp_id] (leaking the prior
device_node reference) and private->ddp_comp[comp_id] while leaving their own
component slots unpopulated.

> -		padding0 = &padding0;
> -		padding1 = &padding1;
> -		padding2 = &padding2;
> -		padding3 = &padding3;
> -		padding4 = &padding4;
> -		padding5 = &padding5;
> -		padding6 = &padding6;
> -		padding7 = &padding7;
> -		vdo1-rdma0 = &vdo1_rdma0;
> -		vdo1-rdma1 = &vdo1_rdma1;
> -		vdo1-rdma2 = &vdo1_rdma2;
> -		vdo1-rdma3 = &vdo1_rdma3;
> -		vdo1-rdma4 = &vdo1_rdma4;
> -		vdo1-rdma5 = &vdo1_rdma5;
> -		vdo1-rdma6 = &vdo1_rdma6;
> -		vdo1-rdma7 = &vdo1_rdma7;
> -	};

[Severity: High]
Does removing the ethdr, merge, padding, and vdo1-rdma aliases break probing
of mediatek-disp-ovl-adaptor on MT8188 and MT8195?

ovl_adaptor_comp_get_id() requires a non-negative alias ID matching
comp_matches[]:

drivers/gpu/drm/mediatek/mtk_disp_ovl_adaptor.c:ovl_adaptor_comp_get_id() {
	int alias_id = of_alias_get_id(node, private_comp_stem[type]);
	int i;

	for (i = 0; i < ARRAY_SIZE(comp_matches); i++)
		if (comp_matches[i].type == type &&
		    comp_matches[i].alias_id == alias_id)
			return i;

	dev_warn(dev, "Failed to get id. type: %d, alias: %d\n", type, alias_id);
	return -EINVAL;
}

When those aliases are removed, of_alias_get_id() returns -ENODEV for every
sub-component, causing ovl_adaptor_comp_init() to skip all children and fail
probe with -ENODEV.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261117130556.234785-2-krzysztof.kozlowski@oss.qualcomm.com?part=1

      reply	other threads:[~2026-11-17 13:19 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-17 13:05 [PATCH] arm64: dts: mediatek: Remove undocumented, unapproved aliases Krzysztof Kozlowski
2026-11-17 13:19 ` 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-218859@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=conor+dt@kernel.org \
    --cc=devicetree@vger.kernel.org \
    --cc=krzysztof.kozlowski@oss.qualcomm.com \
    --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