Re: [PATCH] mfd: kempld-core: Mark kempld-acpi_table as __maybe_unused

[Michael Brunner]

On Tue, 2020-10-06 at 07:53 +0100, Lee Jones wrote:
> On Mon, 05 Oct 2020, Michael Brunner wrote:
> 
> > On Fri, 2020-10-02 at 08:01 +0100, Lee Jones wrote:
> > > On Thu, 01 Oct 2020, Michael Brunner wrote:
> > > 
> > > > The Intel 0-DAY CI Kernel Test Service reports an unused variable
> > > > warning when compiling with clang for PowerPC:
> > > > 
> > > > > > drivers/mfd/kempld-core.c:556:36: warning: unused variable
> > > > > > 'kempld_acpi_table' [-Wunused-const-variable]
> > > >    static const struct acpi_device_id kempld_acpi_table[] = {
> > > > 
> > > > The issue can be fixed by marking kempld_acpi_table as
> > > > __maybe_unused.
> > > > 
> > > > Fixes: e8299c7313af ("[PATCH] mfd: Add ACPI support to Kontron PLD
> > > > driver")
> > > > 
> > > > Reported-by: kernel test robot <lkp@xxxxxxxxx>
> > > > Signed-off-by: Michael Brunner <michael.brunner@xxxxxxxxxxx>
> > > > ---
> > > >  drivers/mfd/kempld-core.c | 2 +-
> > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > 
> > > > diff --git a/drivers/mfd/kempld-core.c b/drivers/mfd/kempld-core.c
> > > > index 1dfe556df038..273481dfaad4 100644
> > > > --- a/drivers/mfd/kempld-core.c
> > > > +++ b/drivers/mfd/kempld-core.c
> > > > @@ -553,7 +553,7 @@ static int kempld_remove(struct platform_device
> > > > *pdev)
> > > >  	return 0;
> > > >  }
> > > >  
> > > > -static const struct acpi_device_id kempld_acpi_table[] = {
> > > > +static const struct acpi_device_id __maybe_unused
> > > > kempld_acpi_table[] = {
> > > >  	{ "KEM0001", (kernel_ulong_t)&kempld_platform_data_generic },
> > > >  	{}
> > > >  };
> > > 
> > > This is not the right fix.  Better just to compile it out completely
> > > in these circumstances.  I already have a fix for this in soak.
> > 
> > Ok - thank you for the other fix you submitted!
> > 
> > But just out of curiosity - in process/coding-style.rst is written that
> > __maybe_unused should be preferred over wrapping in preprocessor
> > conditionals, if a function or variable may potentially go unused in a
> > particular configuration. So why is my patch not the right one here? At
> > least in my tests it seemed to solve the issue.
> 
> It's a bone of contention for sure.  In these kinds of scenarios
> (i.e. ACPI and OF tables) it is way more common to wrap them:
> 
> $ git grep -B3 'acpi_device_id\|of_device_id' | grep 'CONFIG_ACPI\|CONFIG_OF' | wc -l
> 596
> $ git grep -B3 'acpi_device_id\|of_device_id' | grep __maybe_unused | wc -l
> 63
> 
> Parsing them out completely, also has the benefit of saving space,
> reducing the size of the finalised binary.

Doesn't the compiler remove it anyway? At least in my test I didn't see
a difference in the resulting object files.
Doing a crosscheck, by adding __attribute__((used)) to the definition
of kempld_acpi_table, the object file size increased and
kempld_acpi_table showed up in the symbol table.

Nevertheless, I don't want to start a discussion. I am fine with using
the preprocessor. Just wanted to make sure I understand the technical
implications of both solutions.
Thank you for your time!

Best regards,
  Michael