Re: [PATCH] xen/acpi-processor: fix _CST detection using undersized evaluation buffer

[Jan Beulich]

On 24.02.2026 10:10, David Thomson (dt@xxxxxxxxx) wrote:
>> I understand you reflect original behavior in this regard, but why involve any
>> ACPI function here at all when pblk is non-zero? I.e. why not swap the operands
>> of && ? Object evaluation could have wanted side effects (in which case,
>> however, some different change would be needed here), but checking for method
>> presence surely hasn't.
>>
> You're right on both counts. The original evaluate_object() call at least had the appearance of wanting the result (even though buf was undersized and the result was never used). A pure presence check has no such pretense. Swapping the operands is the obvious improvement. I think dropping the _CST check entirely is also defensible since C-state support is confirmed when pblk is set. But I'm not certain there's no edge case where pblk is non-zero and _CST is also not present, so I'd defer to your judgement.
> 
> Would you prefer:
> a) if (pblk && acpi_has_method(handle, "_CST"))
> b) just if (pblk)

I don't think b) is correct, so a) please (albeit suitably adjusted to really only
flip the operands from what your original patch had).

Jan