Re: [PATCH] ARM: do not assemble iwmmxt.S with LLVM toolchain

From: Nathan Chancellor <natechancellor@gmail.com>
To: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Linus Walleij <linus.walleij@linaro.org>,
	Peter.Smith@arm.com, stefan@agner.ch,
	David Howells <dhowells@redhat.com>,
	Mauro Carvalho Chehab <mchehab+samsung@kernel.org>,
	manojgupta@google.com,
	Benjamin Gaignard <benjamin.gaignard@linaro.org>,
	"Joel Fernandes \(Google\)" <joel@joelfernandes.org>,
	Jian Cai <caij2003@gmail.com>,
	Clang-Built-Linux ML <clang-built-linux@googlegroups.com>,
	ilie.halip@gmail.com, Masahiro Yamada <masahiroy@kernel.org>,
	Russell King <linux@armlinux.org.uk>,
	Krzysztof Kozlowski <krzk@kernel.org>,
	Bartosz Golaszewski <bgolaszewski@baylibre.com>,
	samitolvanen@google.com,
	"Eric W. Biederman" <ebiederm@xmission.com>,
	Arnd Bergmann <arnd@arndb.de>,
	"Steven Rostedt \(VMware\)" <rostedt@goodmis.org>,
	jiancai@google.com, Doug Anderson <armlinux@m.disordat.com>,
	Dan Williams <dan.j.williams@intel.com>,
	linux-arm-kernel@lists.infradead.org,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Nick Desaulniers <ndesaulniers@google.com>,
	linux-kernel@vger.kernel.org,
	Patrick Bellasi <patrick.bellasi@arm.com>,
	Masami Hiramatsu <mhiramat@kernel.org>, Tejun Heo <tj@kernel.org>,
	Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [PATCH] ARM: do not assemble iwmmxt.S with LLVM toolchain
Date: Fri, 10 Apr 2020 00:44:40 -0700	[thread overview]
Message-ID: <20200410074440.GA35316@ubuntu-s3-xlarge-x86> (raw)
In-Reply-To: <CA+icZUWJLGfp-UVhXDaCR=Xnce7phE1ffPHC4RzM8mXPhBaV9g@mail.gmail.com>

On Fri, Apr 10, 2020 at 08:38:05AM +0200, Sedat Dilek wrote:
> On Fri, Apr 10, 2020 at 1:28 AM Jian Cai <caij2003@gmail.com> wrote:
> >
> > iwmmxt.S contains XScale instructions LLVM ARM backend does not support.
> > Skip this file if LLVM integrated assemmbler or LLD is used to build ARM
> > kernel.
> >
> > Signed-off-by: Jian Cai <caij2003@gmail.com>
> > ---
> >  arch/arm/Kconfig | 2 +-
> >  init/Kconfig     | 6 ++++++
> >  2 files changed, 7 insertions(+), 1 deletion(-)
> >
> > diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
> > index 66a04f6f4775..39de8fc64a73 100644
> > --- a/arch/arm/Kconfig
> > +++ b/arch/arm/Kconfig
> > @@ -804,7 +804,7 @@ source "arch/arm/mm/Kconfig"
> >
> >  config IWMMXT
> >         bool "Enable iWMMXt support"
> > -       depends on CPU_XSCALE || CPU_XSC3 || CPU_MOHAWK || CPU_PJ4 || CPU_PJ4B
> > +       depends on !AS_IS_CLANG && !LD_IS_LLD && (CPU_XSCALE || CPU_XSC3 || CPU_MOHAWK || CPU_PJ4 || CPU_PJ4B)
> >         default y if PXA27x || PXA3xx || ARCH_MMP || CPU_PJ4 || CPU_PJ4B
> >         help
> >           Enable support for iWMMXt context switching at run time if
> > diff --git a/init/Kconfig b/init/Kconfig
> > index 1c12059e0f7e..b0ab3271e900 100644
> > --- a/init/Kconfig
> > +++ b/init/Kconfig
> > @@ -19,6 +19,12 @@ config GCC_VERSION
> >  config CC_IS_CLANG
> >         def_bool $(success,$(CC) --version | head -n 1 | grep -q clang)
> >
> > +config AS_IS_CLANG
> > +       def_bool $(success,$(AS) --version | head -n 1 | grep -q clang)
> > +
> > +config LD_IS_LLD
> > +       def_bool $(success,$(LD) --version | head -n 1 | grep -q LLD)
> > +
> >  config CLANG_VERSION
> >         int
> >         default $(shell,$(srctree)/scripts/clang-version.sh $(CC))
> > --
> > 2.26.0.110.g2183baf09c-goog
> 
> Yesterday, when looking trough commits in Linus tree, I saw:
> 
> "init/kconfig: Add LD_VERSION Kconfig"
> 
> Nick had a patchset to distinguish LINKER via Kconfig (I cannot find
> it right now).

Probably referring to this?

https://github.com/samitolvanen/linux/commit/61889e01f0ed4f07a9d631f163bba6c6637bfa46

> So we should do all this the way CC_IS_XXX CC_VERSION handling is done.
> 
> I just want to point to [2] where we can rework (simplify) this
> handling for CC and LD handling in a further step.
> In one of Peter Z. tree someone started to do so (I was inspired by that).
> 
> Unfortunately, the hunk from [1] is IMHO a bit mis-placed and CC and
> LD handling should stay together:
> 
> CC_IS_XXX where XXX is GCC or CLANG
> CC_VERSION where CC is GCC or CLANG

Are you suggesting unifying GCC_VERSION and CLANG_VERSION or am I
misunderstanding what you wrote here? Do you mean XXX_VERSION where XXX
is GCC or CLANG?

> LD_IS_XXX where XXX is BFD or GOLD or LLD
> LD_VERSION

ld.gold is no longer allowed to link the kernel so there is no point in
accounting for it in Kconfig. That leaves only ld.bfd and ld.lld to
account for. I do not think there is a point in adding LD_IS_BFD;
!LD_IS_LLD covers that since there is not another linker (at least that
I am aware of) that links the kernel.

Compiler is different because it technically has three options if icc
even still works to build the kernel.

LD_VERISON is explicitly an ld.bfd thing due to the way ld-version.sh
is written:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/scripts/ld-version.sh

There is not much of a reason to try and make LLD work with that given
we do not need it now. I am of the mindset that proactively changing
something only makes life more difficult down the road and makes things
harder to maintain.

We could suggest renaming that config to GNU_LD_VERSION and gnu-ld-version.sh
to be slightly more accurate but I am not sure that is necessary since
again, CONFIG_LD_IS_LLD will handle any incompatibilities that we
encounter with LD_VERSION, just like we do with CLANG_VERSION/GCC_VERSION.
See my commit for the __gnu_mcount_nc thing in ARM for an example of
that (CONFIG_CC_IS_GCC still needs to be specified).

https://git.kernel.org/linus/b0fe66cf095016e0b238374c10ae366e1f087d11

> Just my €0,02.
> 
> Regards,
> - Sedat -
> 
> [1] https://git.kernel.org/linus/9553d16fa671b9621c5e2847d08bd90d3be3349c
> [2] https://github.com/ClangBuiltLinux/linux/issues/941
> 

Cheers,
Nathan

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