[PATCH] acpi: Fix header declaration of acpi_arch_dma_setup() w/o CONFIG_ACPI

[Dave Jiang]

On 6/1/23 03:31, Robin Murphy wrote:
> On 2023-06-01 02:48, Hanjun Guo wrote:
>> Hi Dave,
>>
>> Sorry for the late reply, I have some comments inline.
>>
>> On 2023/5/23 0:06, Dave Jiang wrote:
>>> arm64 build can be done without CONFIG_ACPI. The ifdef bits for
>>> acpi_arch_dma_setup() is placed inside CONFIG_ACPI. When CONFIG_ACPI is
>>> not set, this causes warning reported by kernel test bot. Move the
>>> prototype declaration for acpi_arch_dma_setup() outside of CONFIG_ACPI.
>>
>> ...
>>
>>>
>>>>> drivers/acpi/arm64/dma.c:7:6: warning: no previous prototype for 
>>>>> function 'acpi_arch_dma_setup' [-Wmissing-prototypes]
>>>     void acpi_arch_dma_setup(struct device *dev)
>>>          ^
>>>     drivers/acpi/arm64/dma.c:7:1: note: declare 'static' if the 
>>> function is not intended to be used outside of this translation unit
>>>     void acpi_arch_dma_setup(struct device *dev)
>>>     ^
>>>     static
>>>     1 warning generated.
>>
>> drivers/acpi can only be compiled with CONFIG_ACPI=y, so
>> drivers/acpi/arm64/ will be the same, not sure how to trigger
>> this compile warning.
>>
>> I disable CONFIG_ACPI on my ARM64 machine, but didn't get the
>> warning you reported.
>
> Looking at the linked LKP report, it seems it's it's explicitly trying 
> to build drivers/acpi/arm64/ despite the config:
>
>         mkdir build_dir && cp config build_dir/.config
>         COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross W=1 
> O=build_dir ARCH=arm64 olddefconfig
>         COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross W=1 
> O=build_dir ARCH=arm64 SHELL=/bin/bash drivers/acpi/arm64/
>
> So I guess it's a problem with the LKP setup? In general, trying to 
> build arbitrary parts of the kernel which are configured out can never 
> be expected to work.

It seems to be a problem triggered because I was working on some patches 
that added obj-y at the drivers/Makefile for acpi/ and opened the ACPI 
dir for !CONFIG_ACPI compile. The code has been reworked after comments 
from Rafael. So this is no longer a problem. Please ignore patch. Sorry 
about the noise.




>
> Thanks,
> Robin.