[PATCH] arm64: dts: mediatek: mt8188-geralt: drop firmware-name from first SCP core

[AngeloGioacchino Del Regno]

Il 19/11/25 09:41, Chen-Yu Tsai ha scritto:
> On Wed, Nov 19, 2025 at 4:31 PM AngeloGioacchino Del Regno
> <angelogioacchino.delregno at collabora.com> wrote:
>>
>> Il 19/11/25 08:02, Chen-Yu Tsai ha scritto:
>>> Arnd pointed out that having firmware-name in the device tree is wrong.
>>> Drop it.
>>>
>>> Fixes: e36be19823f0 ("arm64: dts: mediatek: mt8188-geralt: Enable first SCP core")
>>
>> Fixes here is .. well, right, but wrong - as if you "fix" that in previous kernel
>> releases, you will create regressions :-)
> 
> This is a fix because the commit in question was merged in v6.18-rc1
> and v6.18 hasn't been released yet. So you should pick it now for
> v6.18.
> 
>> For all of the currently supported boards, I was planning to temporarily leave the
>> firmware-name properties in place for one kernel version (v6.19) and remove it from
>> all of them at once, for kernel v6.20.
> 
> OK. So I guess the plan is to have MT8188 SCP working in v6.18 based
> on the current name then?
> 

Any firmware in linux-firmware that needs a rename will still need to retain
compatibility with old kernel versions: it doesn't matter if scp.img will be
renamed because we will anyway need a symlink.

This does effectively mean that, for MT8188, the current "scp.img" will be
"reachable" as both "scp.img" and as "scp_c0.img" - so timing for removal of
the firmware-name property in here doesn't matter at all.

There's no rush.

>> I'd be okay if you send a commit or a series removing firmware-name from all of
>> them, without any Fixes tag and without any "fix" word (so that we avoid autosel
>> backports at all costs!), but keep in mind that I'd pick that in the next merge
>> window, and not right now.
> 
> We can avoid autosel with:
> 
>       Cc: <stable+noautosel at kernel.org> # reason goes here, and must be present
> 
> (See stable-kernel-rules for the whole paragraph.)
> 
> After the blobs in the firmware repo have been renamed, we then explicitly
> backport the change to stable. I think that works out better? And the only
> one that needs to be timed is the MT8188 change anyway.
> 
> How does that sound?
> 

We can avoid autosel with that, sure - but I still don't really judge it as being
a fix, as not specifying firmware-name is something new, even if it was technically
wrong from the beginning.
(Someone may or may not agree with me, but that's another story.)

In the past, it wasn't a problem - it's a problem now, and a resolved one - but
this doesn't mean that having this property was a "broken" behavior in the past.

It worked. And it was fine.

Even if I don't see the actual benefit, I don't mind having this backported to
v6.18 when it'll be time to pick this commit along with the others, but again
I don't understand why you're feeling like this is urgent.

For how I see it: -> it's not urgent! <-

There has to be no regression on the supported boards - with or without - this
property being present.

The only one that can be just renamed without *caring* about symlinks is MT8189,
as it "never worked" upstream (as in, there, a rename without a symlink will not
create any regression) simply because MT8189 never got any firmware-name property
(actually, never got any devicetree at all, even).

Again, timing is not important; the sooner the better, yes, but at the same level
of importance as any other *functionality improvement* (not fixes), nothing more
and nothing less.

Does that clarify the course of action for you?

Cheers,
Angelo

> 
> ChenYu
> 
>> Thanks!
>> Angelo
>>
>>> Signed-off-by: Chen-Yu Tsai <wenst at chromium.org>
>>> ---
>>>    arch/arm64/boot/dts/mediatek/mt8188-geralt.dtsi | 1 -
>>>    1 file changed, 1 deletion(-)
>>>
>>> diff --git a/arch/arm64/boot/dts/mediatek/mt8188-geralt.dtsi b/arch/arm64/boot/dts/mediatek/mt8188-geralt.dtsi
>>> index 7fedbacdac44..8e423504ec05 100644
>>> --- a/arch/arm64/boot/dts/mediatek/mt8188-geralt.dtsi
>>> +++ b/arch/arm64/boot/dts/mediatek/mt8188-geralt.dtsi
>>> @@ -1166,7 +1166,6 @@ &scp_cluster {
>>>    &scp_c0 {
>>>        pinctrl-names = "default";
>>>        pinctrl-0 = <&scp_pins>;
>>> -     firmware-name = "mediatek/mt8188/scp.img";
>>>        memory-region = <&scp_mem_reserved>;
>>>        status = "okay";
>>>    };
>>
>>