On Wed, Jun 28, 2023 at 07:08:47PM +0800, Huacai Chen wrote:
After commit 0e96ea5c3eb5904e5dc2f ("MIPS: Loongson64: Clean up use of cc-ifversion") we get a build error when make modules_install:
cc1: error: '-mloongson-mmi' must be used with '-mhard-float'
The reason is when make modules_install, 'call cc-option' doesn't work in $(KBUILD_CFLAGS) of 'CHECKFLAGS'. Then there is no -mno-loongson-mmi applied and -march=loongson3a enable MMI instructions.
To be detail, the error message comes from the CHECKFLAGS invocation of $(CC) but it has no impact on the final result of make modules_install, it is purely a cosmetic issue. The error occurs because cc-option is defined in scripts/Makefile.compiler, which is not included in Makefile when running 'make modules_install', as install targets are not supposed to require the compiler; see commit 805b2e1d427aab4b ("kbuild: include Makefile.compiler only when compiler is needed"). As a result, the call to check for '-mno-loongson-mmi' just never happens.
Fix this by partially reverting to the old logic, use 'call cc-option' to conditionally apply -march=loongson3a and -march=mips64r2.
By the way, Loongson-2E/2F is also broken in commit 13ceb48bc19c563e05f4 ("MIPS: Loongson2ef: Remove unnecessary {as,cc}-option calls") so fix it together.
Fixes: 13ceb48bc19c563e05f4 ("MIPS: Loongson2ef: Remove unnecessary {as,cc}-option calls") Fixes: 0e96ea5c3eb5904e5dc2 ("MIPS: Loongson64: Clean up use of cc-ifversion") Cc: stable@vger.kernel.org Cc: Feiyang Chen chenfeiyang@loongson.cn Cc: Nathan Chancellor nathan@kernel.org Cc: Nick Desaulniers ndesaulniers@google.com Signed-off-by: Huacai Chen chenhuacai@loongson.cn
Reviewed-by: Nathan Chancellor nathan@kernel.org
V2: Update commit message and fix for LOONGSON2EF together.
arch/mips/Makefile | 10 +++------- 1 file changed, 3 insertions(+), 7 deletions(-)
diff --git a/arch/mips/Makefile b/arch/mips/Makefile index a7a4ee66a9d3..35a1b9b34734 100644 --- a/arch/mips/Makefile +++ b/arch/mips/Makefile @@ -181,16 +181,12 @@ endif cflags-$(CONFIG_CAVIUM_CN63XXP1) += -Wa,-mfix-cn63xxp1 cflags-$(CONFIG_CPU_BMIPS) += -march=mips32 -Wa,-mips32 -Wa,--trap -cflags-$(CONFIG_CPU_LOONGSON2E) += -march=loongson2e -Wa,--trap -cflags-$(CONFIG_CPU_LOONGSON2F) += -march=loongson2f -Wa,--trap +cflags-$(CONFIG_CPU_LOONGSON2E) += $(call cc-option,-march=loongson2e) -Wa,--trap +cflags-$(CONFIG_CPU_LOONGSON2F) += $(call cc-option,-march=loongson2f) -Wa,--trap +cflags-$(CONFIG_CPU_LOONGSON64) += $(call cc-option,-march=loongson3a,-march=mips64r2) -Wa,--trap # Some -march= flags enable MMI instructions, and GCC complains about that # support being enabled alongside -msoft-float. Thus explicitly disable MMI. cflags-$(CONFIG_CPU_LOONGSON2EF) += $(call cc-option,-mno-loongson-mmi) -ifdef CONFIG_CPU_LOONGSON64 -cflags-$(CONFIG_CPU_LOONGSON64) += -Wa,--trap -cflags-$(CONFIG_CC_IS_GCC) += -march=loongson3a -cflags-$(CONFIG_CC_IS_CLANG) += -march=mips64r2 -endif cflags-$(CONFIG_CPU_LOONGSON64) += $(call cc-option,-mno-loongson-mmi) cflags-$(CONFIG_CPU_R4000_WORKAROUNDS) += $(call cc-option,-mfix-r4000,) -- 2.39.3