Re: [alsa-devel] [PATCH] soundwire: Don't build sound.o without CONFIG_ACPI

[Nathan Chancellor <natechancellor@gmail.com>]

On Tue, Aug 13, 2019 at 09:22:29AM -0500, Pierre-Louis Bossart wrote:
> On 8/13/19 1:10 AM, Nathan Chancellor wrote:
> > clang warns when CONFIG_ACPI is unset:
> > 
> > ../drivers/soundwire/slave.c:16:12: warning: unused function
> > 'sdw_slave_add' [-Wunused-function]
> > static int sdw_slave_add(struct sdw_bus *bus,
> >             ^
> > 1 warning generated.
> > 
> > Before commit 8676b3ca4673 ("soundwire: fix regmap dependencies and
> > align with other serial links"), this code would only be compiled when
> > ACPI was set because it was only selected by SOUNDWIRE_INTEL, which
> > depends on ACPI.
> > 
> > Now, this code can be compiled without CONFIG_ACPI, which causes the
> > above warning. The IS_ENABLED(CONFIG_ACPI) guard could be moved to avoid
> > compiling the function; however, slave.c only contains three functions,
> > two of which are static. Only compile slave.o when CONFIG_ACPI is set,
> > where it will actually be used. bus.h contains a stub for
> > sdw_acpi_find_slaves so there will be no issues with an undefined
> > function.
> > 
> > This has been build tested with CONFIG_ACPI set and unset in combination
> > with CONFIG_SOUNDWIRE unset, built in, and a module.
> 
> Thanks for the patch. Do you have a .config you can share offline so that we
> add it to our tests?

I just took the arm64 defconfig and deleted CONFIG_ACPI and added
CONFIG_SOUNDWIRE=y or =m to produce this warning. I initially found this
on an arm64 allyesconfig build.

> 
> > 
> > Fixes: 8676b3ca4673 ("soundwire: fix regmap dependencies and align with other serial links")
> > Link: https://github.com/ClangBuiltLinux/linux/issues/637
> > Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
> > ---
> >   drivers/soundwire/Makefile | 6 +++++-
> >   drivers/soundwire/slave.c  | 3 ---
> >   2 files changed, 5 insertions(+), 4 deletions(-)
> > 
> > diff --git a/drivers/soundwire/Makefile b/drivers/soundwire/Makefile
> > index 45b7e5001653..226090902716 100644
> > --- a/drivers/soundwire/Makefile
> > +++ b/drivers/soundwire/Makefile
> > @@ -4,9 +4,13 @@
> >   #
> >   #Bus Objs
> > -soundwire-bus-objs := bus_type.o bus.o slave.o mipi_disco.o stream.o
> > +soundwire-bus-objs := bus_type.o bus.o mipi_disco.o stream.o
> >   obj-$(CONFIG_SOUNDWIRE) += soundwire-bus.o
> > +ifdef CONFIG_ACPI
> > +soundwire-bus-objs += slave.o
> > +endif
> 
> I am fine with the change, but we might as well rename the file acpi_slave.c
> then?

Sure, I can do that rename and send a v2.

> 
> > +
> >   #Cadence Objs
> >   soundwire-cadence-objs := cadence_master.o
> >   obj-$(CONFIG_SOUNDWIRE_CADENCE) += soundwire-cadence.o
> > diff --git a/drivers/soundwire/slave.c b/drivers/soundwire/slave.c
> > index f39a5815e25d..0dc188e6873b 100644
> > --- a/drivers/soundwire/slave.c
> > +++ b/drivers/soundwire/slave.c
> > @@ -60,7 +60,6 @@ static int sdw_slave_add(struct sdw_bus *bus,
> >   	return ret;
> >   }
> > -#if IS_ENABLED(CONFIG_ACPI)
> >   /*
> >    * sdw_acpi_find_slaves() - Find Slave devices in Master ACPI node
> >    * @bus: SDW bus instance
> > @@ -110,5 +109,3 @@ int sdw_acpi_find_slaves(struct sdw_bus *bus)
> >   	return 0;
> >   }
> > -
> > -#endif
> > 
>