Re: [PATCH] platform/x86: tpacpi: Mark charging LEDs as safe

[Josh Triplett]

Rong Zhang wrote:
> On Sat, 2026-11-07 at 12:49 -0800, Josh Triplett wrote:
>> If that is the primary concern, would you accept a DMI-identity-based
>> allowlist?
>
> I don't think you need this if we take the module parameter approach.
> Usually we don't want to introduce a infinitely-growing DMI quirk table
> for some niche use cases.

The module parameter approach seems fine, as long as it doesn't taint the
kernel to use.

> Meanwhile, I can't totally trust your LLM's conclusion.

Extremely valid.

> The ACPI part may
> be correct as understanding decompiled ASL code is not hard for an LLM,
> while reverse-engineering EC firmware is completely on another dimension.

As it turns out, the EC firmware appeared to be unencrypted ARM Thumb code,
and LLMs are pretty good at disassembling these days, too. But still, full
support for "can't totally trust". The ideal solution would be if someone
from Lenovo could confirm the conclusion.