Re: [PATCH] gpib: mark pnp_device_id tables as __maybe_unused

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Wed, Feb 05, 2025 at 02:04:37PM +0100, Arnd Bergmann wrote:
> On Wed, Feb 5, 2025, at 13:25, Greg Kroah-Hartman wrote:
> > On Wed, Feb 05, 2025 at 01:12:26PM +0100, Arnd Bergmann wrote:
> >>
> >> @@ -1390,7 +1390,7 @@ static struct pci_driver tnt4882_pci_driver = {
> >>  	.probe = &tnt4882_pci_probe
> >>  };
> >>  
> >> -static const struct pnp_device_id tnt4882_pnp_table[] = {
> >> +static __maybe_unused const struct pnp_device_id tnt4882_pnp_table[] = {
> >
> > I see this happening in many different drivers right now, what is so
> > unique about pnp that causes this?  Shouldn't we fix up the
> > MODULE_DEVICE_TABLE() macro to not require stuff like this instead?
> 
> I think the other drivers that produce a similar warning usually
> have a different bug, they have an incorrect of_match_ptr() or
> ACPI_PTR() around the reference to that table, and the correct
> fix is usually to just remove those macros. I have previously
> sent patches for all of these, and could resend those.
> 
> These two pnp drivers are special because they predate the
> linux-2.6 driver model and there is no reference to the table
> at all in the drivers.

Ah.  Then the variable should just be removed as it's obviously not
doing anything :)

(I know it's doing module loading, but that's not ok to just fake it
this way, it should use that table when it is searching the pnp area.)

So a nice "#if 0" carve out for now perhaps?

thanks,

greg k-h