Re: [PATCH] arm64: armv8_deprecated: fix unused-function error

[Ren Zhijie]

在 2022/11/23 19:01, Mark Rutland 写道:
> On Wed, Nov 23, 2022 at 10:06:03AM +0800, Ren Zhijie wrote:
> > 在 2022/11/23 0:48, Mark Rutland 写道:
> > > On Tue, Nov 22, 2022 at 03:20:10AM +0000, Ren Zhijie wrote:
> > > > If CONFIG_SWP_EMULATION is not set and
> > > > CONFIG_CP15_BARRIER_EMULATION is not set,
> > > > aarch64-linux-gnu complained about unused-function :
> > > >
> > > > arch/arm64/kernel/armv8_deprecated.c:67:21: error: ‘aarch32_check_condition’ defined but not used [-Werror=unused-function]
> > > >    static unsigned int aarch32_check_condition(u32 opcode, u32 psr)
> > > >                        ^~~~~~~~~~~~~~~~~~~~~~~
> > > > cc1: all warnings being treated as errors
> > > >
> > > > To fix this error, warp the definition of
> > > > aarch32_check_condition() by defined(CONFIG_SWP_EMULATION) ||
> > > > defined(CONFIG_CP15_BARRIER_EMULATION)
> > > >
> > > > Fixes: 0c5f416219da ("arm64: armv8_deprecated: move aarch32 helper earlier")
> > > This also depends on building with additional options to turn warnings into
> > > errors, no?
> > No，i just run the normal command as follow:
> >
> > make ARCH="arm64" CROSS_COMPILE="aarch64-linux-gnu-"
> I think you must also have CONFIG_WERROR enabled?
>
> Just building defconfig + CONFIG_ARMV8_DEPRECATED=y gives me a warning, but not
> an error, and the kernel builds just fine.
>
> So this is a problem to fix, and I appreciate that in test configs this might
> be broken, but it's not a full build-time failure for most users.

I get it, thanks a lot!

Thanks,

Ren

> Thanks,
> Mark.
>
> > > > Signed-off-by: Ren Zhijie <renzhijie2@xxxxxxxxxx>
> > > > ---
> > > >    arch/arm64/kernel/armv8_deprecated.c | 2 ++
> > > >    1 file changed, 2 insertions(+)
> > > >
> > > > diff --git a/arch/arm64/kernel/armv8_deprecated.c b/arch/arm64/kernel/armv8_deprecated.c
> > > > index ed0788cf6bbb..3f29ceb6653a 100644
> > > > --- a/arch/arm64/kernel/armv8_deprecated.c
> > > > +++ b/arch/arm64/kernel/armv8_deprecated.c
> > > > @@ -64,6 +64,7 @@ struct insn_emulation {
> > > >    #define	ARM_OPCODE_CONDITION_UNCOND	0xf
> > > > +#if defined(CONFIG_SWP_EMULATION) || defined(CONFIG_CP15_BARRIER_EMULATION)
> > > >    static unsigned int aarch32_check_condition(u32 opcode, u32 psr)
> > > >    {
> > > >    	u32 cc_bits  = opcode >> 28;
> > > > @@ -76,6 +77,7 @@ static unsigned int aarch32_check_condition(u32 opcode, u32 psr)
> > > >    	}
> > > >    	return ARM_OPCODE_CONDTEST_UNCOND;
> > > >    }
> > > > +#endif
> > > Could we mark this as '__maybe_unused' or 'inline' instead? I think that's
> > > preferable to the ifdeferry.
> > sure, i will use __maybe_unused in v2.
> >
> > Thanks,
> >
> > Ren.
> >
> > > Thanks,
> > > Mark.
> > >
> > > >    #ifdef CONFIG_SWP_EMULATION
> > > >    /*
> > > > --
> > > > 2.17.1
> > > .
> .