Re: acpi_device_notify() binding devices that don't seem like they should be bound

[Daniel Scally <djrscally@xxxxxxxxx>]

On 10/12/2020 16:05, Rafael J. Wysocki wrote:
> All the adevs with matching _ADR also have both _STA and _HID
> unfortunately. Sorry; let me stop half-arsing this and show you
> something useful:
>
> [    0.219953] acpi_find_child_device(PNP0A08:00, 0x00, false)
> [    0.220818] INT3472:00: _STA 0x00, _ADR=0x00000000, _HID=INT3472
> [    0.220821] INT3472:01: _STA 0x00, _ADR=0x00000000, _HID=INT3472
> [    0.220870] INT3472:02: _STA 0x00, _ADR=0x00000000, _HID=INT3472
> [    0.220892] INT3472:03: _STA 0x00, _ADR=0x00000000, _HID=INT3472
> [    0.220916] INT3472:04: _STA 0x00, _ADR=0x00000000, _HID=INT3472
> [    0.220941] INT3472:05: _STA 0x00, _ADR=0x00000000, _HID=INT3472
> [    0.220965] INT3472:06: _STA 0x00, _ADR=0x00000000, _HID=INT3472
> [    0.220990] INT3472:07: _STA 0x00, _ADR=0x00000000, _HID=INT3472
> These will be ignored with -ENODEV.
>
>> [    0.221038] INT3472:08: _STA 0x0f, _ADR=0x00000000, _HID=INT3472
> For this acpi_find_child_device() will return FIND_CHILD_MIN_SCORE if
> I'm not mistaken.
It does - this is the one that binds, being the first.
>> [    0.221051] OVTI5648:00: _STA 0x0f, _ADR=0x00000000, _HID=OVTI5648
>> [    0.221061] INT3472:09: _STA 0x0f, _ADR=0x00000000, _HID=INT3472
>> [    0.221070] OVTI2680:00: _STA 0x0f, _ADR=0x00000000, _HID=OVTI2680
> As well as for the three above.
>
>> [    0.221079] INT3471:00: _STA 0x00, _ADR=0x00000000, _HID=INT3471
>> [    0.221105] INT33BE:00: _STA 0x00, _ADR=0x00000000, _HID=INT33BE
>> [    0.221130] INT3471:01: _STA 0x00, _ADR=0x00000000, _HID=INT3471
>> [    0.221156] INT33BE:01: _STA 0x00, _ADR=0x00000000, _HID=INT33BE
> And the rest will be ignored.
>
>> That's the debug output I included for each adev that's assessed as a
>> child of PNP0A08:00. _STA, _ADR and _HID present for all, _ADR 0x00 for
>> all, _STA 0x0f for the 2 sensors and their PMIC's and 0x00 for the rest.
>> The same situation holds on both of my devices.
> So in fact we don't want to have an ACPI companion for (PNP0A08:00,
> 0x00, false).
Yeah, I think that's right
> This is a hostbridge special case and let me think about this for a while.
Sure - thanks very much for your help.