Re: [PATCH 1/1] tools/libacpi: clear ASL warning about PCI0

[Jan Beulich <jbeulich@xxxxxxxx>]

On 16.12.2024 13:19, Ariel Otilibili-Anieli wrote:
> On Monday, December 16, 2024 12:38 CET, Jan Beulich <jbeulich@xxxxxxxx> wrote:
> 
>> On 16.12.2024 12:31, Ariel Otilibili-Anieli wrote:
>>> On Monday, December 16, 2024 12:01 CET, Jan Beulich <jbeulich@xxxxxxxx> 
>>> wrote:
>>>
>>>> On 16.12.2024 11:36, Ariel Otilibili-Anieli wrote:
>>>>> On Monday, December 16, 2024 10:53 CET, Jan Beulich <jbeulich@xxxxxxxx> 
>>>>> wrote:
>>>>>
>>>>>> On 15.12.2024 16:40, Ariel Otilibili wrote:
>>>>>>> * iasl complains _HID and _ADR cannot be used at the same time
>>>>>>>
>>>>>>> ```
>>>>>>> /usr/bin/iasl -vs -p tools/firmware/hvmloader/dsdt_anycpu.tmp -tc 
>>>>>>> tools/firmware/hvmloader/dsdt_anycpu.asl 2>&1 | grep -B10 HID
>>>>>>> tools/firmware/hvmloader/dsdt_anycpu.asl     40:        Device (PCI0)
>>>>>>> Warning  3073 -                                    Multiple types ^  
>>>>>>> (Device object requires either a _HID or _ADR, but not both)
>>>>>>> ```
>>>>>>>
>>>>>>> * generally _HID devices are enumerated and have their drivers loaded 
>>>>>>> by ACPI
>>>>>>> * this is from "ASL 2.0 Introduction and Overview" (page 4).
>>>>>>> * removing _ADR, the warning is cleared out.
>>>>>>
>>>>>> Okay, that's the positive aspect. Yet what about the potential fallout 
>>>>>> thereof?
>>>>>> Can you confirm that there's no risk of regressions with older guest 
>>>>>> OSes, for
>>>>>> example?
>>>>>
>>>>> OSes that were released after ACPI 2.0 should work [1]; including WinXP: 
>>>>> The 2.0 specs says either _HID or _ADR should be included [2], not both 
>>>>> (Section 6.1, page 146).
>>>>
>>>> We must be looking at two different variants of the spec then. My copy says
>>>> "device object must contain either an _HID object or an _ADR object, but 
>>>> can
>>>> contain both." Also still in 2.0c. I agree that in e.g. 6.5 the wording has
>>>> changed. I also agree that the use of "either" doesn't help clarity.
>>>
>>> I looked up 2.0 (July 2000); indeed, it said "can contain both". My bad.
>>>>
>>>>> I chose WinXP because, on another patch, it came up in the discussion [3].
>>>
>>> The change should work down to WinXP: the name _HID is kept.
>>>
>>> ```
>>> $ git grep -B2 -A2 -n PNP0A03
>>> tools/libacpi/dsdt.asl-40-       Device (PCI0)
>>> tools/libacpi/dsdt.asl-41-       {
>>> tools/libacpi/dsdt.asl:42:           Name (_HID, EisaId ("PNP0A03"))
>>> tools/libacpi/dsdt.asl-43-           Name (_UID, 0x00)
>>> tools/libacpi/dsdt.asl-44-           Name (_ADR, 0x00)
>>> ```
>>>
>>> Its EISA ID is "PNP0A03"; the namespace is reserved for Microsoft. 
>>> Microsoft identifies "PNP0A03" as PCI devices [1].
>>
>> You again say "should" without explaining what you derive this from. Is it
>> written down somewhere that no OS we (remotely) care about ever evaluated
>> _ADR when _HID was there? As before, along side mentioning the benefits of
>> the change, I'd like to also see a discussion of risks.
>>
> 
> I derive this knowledge only from the APCI specs. Indeed, I've not researched 
> how every OS interprets _HID and _ADR.
> 
> From your answer, I see you would like to be sure the change will introduce 
> no regression. I do understand you point of view; keeping the backward 
> compatibility. 
> 
> The benefit is that the warning will be cleared. We agree on that.
> 
> The risk is that, if ever any OS relies on _ADR, and cannot read _HID; it 
> would break.
> 
> After thinking about it, the other way is less risky: this _HID name is only 
> recognized by Windows. Every OS should (I did say it again, should) 
> understand _ADR.
> 
> If you think the change makes sense, I can remove _HID instead of _ADR.

But that would remove relevant information, the the PNP ID serves a purpose.

> Otherwise, I think we should end the discussion.

Well, you may decide to withdraw / abandon the patch, or you may decide to
re-submit with an extended description, clarifying why the removal is
expected to be safe. Even if - obviously - you can't inspect e.g. Windows
sources.

Jan