Re: [PATCH V4 14/14] LoongArch: Adjust build infrastructure for 32BIT/64BIT

[Nathan Chancellor <nathan@kernel.org>]

Hi Huacai,

Now that this is in -next as commit 3d9aba6618d1 ("LoongArch: Adjust
build infrastructure for 32BIT/64BIT"), I am seeing two issues:

On Thu, Nov 27, 2025 at 11:48:32PM +0800, Huacai Chen wrote:
> Adjust build infrastructure (Kconfig, Makefile and ld scripts) to let
> us enable both 32BIT/64BIT kernel build.
...
> diff --git a/arch/loongarch/Kconfig b/arch/loongarch/Kconfig
> index 730f34214519..4bacde9f46d1 100644
> --- a/arch/loongarch/Kconfig
> +++ b/arch/loongarch/Kconfig
> +menu "Kernel type and options"
> +
> +choice
> +	prompt "Kernel type"
>  
>  config 32BIT
> -	bool
> +	bool "32-bit kernel"
> +	help
> +	  Select this option if you want to build a 32-bit kernel.
>  
>  config 64BIT
> -	def_bool y
> +	bool "64-bit kernel"
> +	help
> +	  Select this option if you want to build a 64-bit kernel.
> +
> +endchoice

The way this is written results in a 32-bit kernel when building
ARCH=loongarch allmodconfig or having a previously 64-bit configuration
and running olddefconfig on it, which seems surprising to me. I think it
would be good to either adjust the ordering such that 64BIT is first or
include an explicit default statement like so:

diff --git a/arch/loongarch/Kconfig b/arch/loongarch/Kconfig
index 3b042dbb2c41..3addd06b3f5a 100644
--- a/arch/loongarch/Kconfig
+++ b/arch/loongarch/Kconfig
@@ -220,6 +220,7 @@ menu "Kernel type and options"
 
 choice
 	prompt "Kernel type"
+	default 64BIT
 
 config 32BIT
 	bool "32-bit kernel"
--

> diff --git a/arch/loongarch/Makefile b/arch/loongarch/Makefile
> index 8d45b860fe56..47516aeea9d2 100644
> --- a/arch/loongarch/Makefile
> +++ b/arch/loongarch/Makefile
...
> @@ -62,9 +66,19 @@ ifneq ($(SUBARCH),$(ARCH))
>    endif
>  endif
>  
> +ifdef CONFIG_32BIT
> +ifdef CONFIG_32BIT_STANDARD
> +ld-emul			= $(32bit-emul)
> +cflags-y		+= -march=la32v1.0 -mabi=ilp32s -mcmodel=normal
> +else  # CONFIG_32BIT_REDUCED
> +ld-emul			= $(32bit-emul)
> +cflags-y		+= -march=la32rv1.0 -mabi=ilp32s -mcmodel=normal
> +endif
> +endif

As a result of the above problem, I see

  clang: error: ignoring '-mabi=ilp32s' as it conflicts with that implied by '-msoft-float' (lp64s) [-Werror,-Woption-ignored]
  clang: error: ignoring '-mabi=ilp32s' as it conflicts with that implied by '-msoft-float' (lp64s) [-Werror,-Woption-ignored]
  clang: error: ignoring '-mabi=ilp32s' as it conflicts with that implied by '-msoft-float' (lp64s) [-Werror,-Woption-ignored]

when building configurations that I expected to be 64-bit but had been
turned into 32-bit ones with LLVM. This was also reported by the test
robot with allnoconfig.

  https://lore.kernel.org/202604232041.ESJDwVG4-lkp@intel.com/

Maybe some change is needed on the LLVM side?

Cheers,
Nathan