Re: [PATCH] ARM: pass -msoft-float to gcc earlier

From: Szabolcs Nagy <szabolcs.nagy@arm.com>
To: Nick Desaulniers <ndesaulniers@google.com>
Cc: Arnd Bergmann <arnd@arndb.de>,
	Linus Walleij <linus.walleij@linaro.org>,
	Russell King <linux@armlinux.org.uk>,
	LKML <linux-kernel@vger.kernel.org>,
	Masahiro Yamada <yamada.masahiro@socionext.com>,
	nd@arm.com, Linux ARM <linux-arm-kernel@lists.infradead.org>
Subject: Re: [PATCH] ARM: pass -msoft-float to gcc earlier
Date: Wed, 20 May 2020 09:36:32 +0100	[thread overview]
Message-ID: <20200520083631.GJ27289@arm.com> (raw)
In-Reply-To: <CAKwvOd=EphyUWNOZ59FkrfDvDwHHU2yrXW+KML-w_6mHowK=QA@mail.gmail.com>

The 05/19/2020 17:38, Nick Desaulniers wrote:
> sorry, hit tab/enter too soon...
> 
> On Tue, May 19, 2020 at 5:37 PM Nick Desaulniers
> <ndesaulniers@google.com> wrote:
> >
> > On Tue, May 19, 2020 at 3:09 PM Arnd Bergmann <arnd@arndb.de> wrote:
> > >
> > > Szabolcs Nagy ran into a kernel build failure with a custom gcc
> > > toochain that sets -mfpu=auto -mfloat-abi=hard:
> > >
> > >  /tmp/ccmNdcdf.s:1898: Error: selected processor does not support `cpsid i' in ARM mode
> > >
> > > The problem is that $(call cc-option, -march=armv7-a) fails before the
> > > kernel overrides the gcc options to also pass -msoft-float.
> >
> > The call to `$(call cc-option, -march=armv7-a) is th
> 
> The call to `$(call cc-option, -march=armv7-a) is the one that fails or...?

the flag -march=armv7-a is invalid if the float abi
is hard and no fpu is specified (since gcc-8).

either an fpu should be specified or -march=armv7-a+fp
(my toolchain was configured with the latter and it does
not work if the kernel overrides it with -march=armv7-a)

because of this cc-option failure the kernel goes on to
pass nonsense flags everywhere (-march=armv5t) and some
compilation eventually fails with an asm error.

> > >
> > > Move the option to the beginning the Makefile, before we call
> >
> > beginning of the
> >
> > > cc-option for the first time.
> > >
> > > Reported-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
> > > Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87302
> > > Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> >
> > Moving this looks harmless enough, though it's not clear to me how the
> > failure you're describing would occur.  I don't see calls to as-instr
> > in arch/arm/Makefile.  Which object is being built before -msoft-float
> > is being set?
> 
> ... ^
> 
> >
> > > ---
> > >  arch/arm/Makefile | 4 +++-
> > >  1 file changed, 3 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/arch/arm/Makefile b/arch/arm/Makefile
> > > index 7d5cd0f85461..e428ea6eb0fa 100644
> > > --- a/arch/arm/Makefile
> > > +++ b/arch/arm/Makefile
> > > @@ -16,6 +16,8 @@ LDFLAGS_vmlinux       += --be8
> > >  KBUILD_LDFLAGS_MODULE  += --be8
> > >  endif
> > >
> > > +KBUILD_CFLAGS  += -msoft-float
> > > +
> > >  ifeq ($(CONFIG_ARM_MODULE_PLTS),y)
> > >  KBUILD_LDS_MODULE      += $(srctree)/arch/arm/kernel/module.lds
> > >  endif
> > > @@ -135,7 +137,7 @@ AFLAGS_ISA  :=$(CFLAGS_ISA)
> > >  endif
> > >
> > >  # Need -Uarm for gcc < 3.x
> > > -KBUILD_CFLAGS  +=$(CFLAGS_ABI) $(CFLAGS_ISA) $(arch-y) $(tune-y) $(call cc-option,-mshort-load-bytes,$(call cc-option,-malignment-traps,)) -msoft-float -Uarm
> > > +KBUILD_CFLAGS  +=$(CFLAGS_ABI) $(CFLAGS_ISA) $(arch-y) $(tune-y) $(call cc-option,-mshort-load-bytes,$(call cc-option,-malignment-traps,)) -Uarm
> > >  KBUILD_AFLAGS  +=$(CFLAGS_ABI) $(AFLAGS_ISA) $(arch-y) $(tune-y) -include asm/unified.h -msoft-float
> > >
> > >  CHECKFLAGS     += -D__arm__
> > > --
> > > 2.26.2
> > >
> 
> -- 
> Thanks,
> ~Nick Desaulniers

-- 

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel