Re: [PATCH v3 6/7] ACPI: bus: implement acpi_device_hid when !ACPI

[Mauro Carvalho Chehab <mchehab+huawei@kernel.org>]

Em Wed, 11 Dec 2024 08:48:51 +0000
Sakari Ailus <sakari.ailus@linux.intel.com> escreveu:

> Hi Mauro,
> 
> On Wed, Dec 11, 2024 at 09:40:37AM +0100, Mauro Carvalho Chehab wrote:
> > Em Wed, 11 Dec 2024 07:57:06 +0000
> > Sakari Ailus <sakari.ailus@linux.intel.com> escreveu:
> >   
> > > Hi Ricardo,
> > > 
> > > On Tue, Dec 10, 2024 at 11:35:35PM +0100, Ricardo Ribalda wrote:  
> > > > On Tue, 10 Dec 2024 at 22:01, Sakari Ailus <sakari.ailus@linux.intel.com> wrote:    
> > > > >
> > > > > Hi Ricardo,
> > > > >
> > > > > On Tue, Dec 10, 2024 at 07:56:03PM +0000, Ricardo Ribalda wrote:    
> > > > > > Provide an implementation of acpi_device_hid that can be used when
> > > > > > CONFIG_ACPI is not set.
> > > > > >
> > > > > > Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
> > > > > > ---
> > > > > >  include/acpi/acpi_bus.h | 5 +++++
> > > > > >  1 file changed, 5 insertions(+)
> > > > > >
> > > > > > diff --git a/include/acpi/acpi_bus.h b/include/acpi/acpi_bus.h
> > > > > > index 4f1b3a6f107b..c25914a152ee 100644
> > > > > > --- a/include/acpi/acpi_bus.h
> > > > > > +++ b/include/acpi/acpi_bus.h
> > > > > > @@ -1003,6 +1003,11 @@ static inline int unregister_acpi_bus_type(void *bus) { return 0; }
> > > > > >
> > > > > >  static inline int acpi_wait_for_acpi_ipmi(void) { return 0; }
> > > > > >
> > > > > > +static inline const char *acpi_device_hid(struct acpi_device *device)
> > > > > > +{
> > > > > > +     return "";
> > > > > > +}    
> > > > >
> > > > > I wonder if any caller might expect something of a string if provided?
> > > > > Valid _HIDs are either 7 or 8 characters whereas the proper version of the
> > > > > function returns "device" when one cannot be found (dummy_hid in
> > > > > drivers/acpi/scan.c). Unlikely to be a problem perhaps.    
> > > > 
> > > > Good point. I changed it to return "device"    
> > > 
> > > When ACPI is disabled, it's unlikely that string would be used anyway, vs.
> > > the case when ACPI is enabled but there's no _HID. So I think an empty
> > > string should be fine. I wonder what others think.
> > >   
> > Returning "" also caused me some attention at the original patch. IMO,
> > placing a pseudo-valid HID would be better, but I guess "device" is also
> > invalid, as, at least I always saw HIDs in uppercase. Also, I guess it
> > is always a vendor ID + a 4 digit number.
> > 
> > so, IMHO, something like "DEVC9999" would be a better name if we fill it.  
> 
> How about post a patch changing "device" in drivers/acpi/scan.c? :-)

Yeah, keeping it coherent makes sense, but see:

	static const char *dummy_hid = "device";

This is compiled for production kernels, and not just for COMPILE_TEST,
while:

	static inline const char *acpi_device_hid(struct acpi_device *device)
	{
		return "foo";
	}

is only COMPILE_TEST. They don't need to be aligned.

> But I
> think the string also needs to be an invalid as a _HID object so it's not
> masking an actual hardware ID used by a real device.

It doesn't matter if if ever conflicts to a real device, as this is
for COMPILE_TEST only.

Anyway, from my side, I'm just giving my 2 cents. I'm ok either way: 
"", "device", "DEVC999", ...

Thanks,
Mauro