Building the Linux kernel with Clang and LLVM
 help / color / mirror / Atom feed
From: Nathan Chancellor <nathan@kernel.org>
To: Masahiro Yamada <masahiroy@kernel.org>
Cc: linux-kbuild@vger.kernel.org,
	Nick Desaulniers <ndesaulniers@google.com>,
	Tom Rix <trix@redhat.com>,
	llvm@lists.linux.dev, linux-kernel@vger.kernel.org,
	Fangrui Song <maskray@google.com>,
	Nicolas Schier <nicolas@fjasle.eu>
Subject: Re: [PATCH] kbuild: clang: do not use CROSS_COMPILE for target triple
Date: Wed, 19 Apr 2023 12:24:46 -0700	[thread overview]
Message-ID: <20230419192446.GA1993276@dev-arch.thelio-3990X> (raw)
In-Reply-To: <CAK7LNAQsgB-SC2t+9ZpDwJbME+BVt+Gur1ydTnHfBfMAfn_y0w@mail.gmail.com>

On Sun, Apr 16, 2023 at 10:02:12PM +0900, Masahiro Yamada wrote:
> On Mon, Apr 3, 2023 at 11:48 PM Nathan Chancellor <nathan@kernel.org> wrote:
> >
> > On Sun, Apr 02, 2023 at 02:01:17AM +0900, Masahiro Yamada wrote:
> > > The target triple is overridden by the user-supplied CROSS_COMPILE,
> > > but I do not see a good reason to support it. Users can use a new
> > > architecture without adding CLANG_TARGET_FLAGS_*, but that would be
> > > a rare case.
> > >
> > > Use the hard-coded and deterministic target triple all the time.
> > >
> > > Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
> >
> > I know of one bug where the value of '--target' matters:
> >
> > https://github.com/ClangBuiltLinux/linux/issues/1244
> 
> 
> I did not look into it closely, but if we say

Heh, neither did I when I wrote the initial response to this patch,
because that issue turns out to be entirely irrelevant for this patch :)

> "
> Using either CROSS_COMPILE=powerpc64-linux-gnu- or
> CROSS_COMPILE=powerpc-linux-gnu- fixes it.
> Using KCFLAGS=-v reveals that powerpc64le-linux-gnu-as is not getting
> the endianness information.
> ", why didn't we fix it like the following?

It is because this already technically happens under the hood with
clang. '--target=powerpc64le-linux-gnu -mbig-endian' is functionally
equivalent to '--target=powerpc64-linux-gnu'. The issue is that
'--target=powerpc64-linux-gnu' and '--target=powerpc-linux-gnu' were not
passing along '-mbig-endian' to the assembler, so
powerpc64le-linux-gnu-ld was expecting big endian object files due to
'-EB' in LDFLAGS but was getting little endian object files, since the
assembler's triple was little endian by default.

We could work around this in the kernel by explicitly passing
'-Wa,-mlittle-endian' for these older versions of clang but I do not
think it is worth it, especially since it will just get removed once we
no longer support clang-11, which is our next uprev.

TL;DR: Patch should be fine, issue is tangential.

>  diff --git a/scripts/Makefile.clang b/scripts/Makefile.clang
>  index 70b354fa1cb4..8dda7dc69c93 100644
>  --- a/scripts/Makefile.clang
>  +++ b/scripts/Makefile.clang
>  @@ -6,7 +6,7 @@ CLANG_TARGET_FLAGS_arm64        := aarch64-linux-gnu
>   CLANG_TARGET_FLAGS_hexagon     := hexagon-linux-musl
>   CLANG_TARGET_FLAGS_m68k                := m68k-linux-gnu
>   CLANG_TARGET_FLAGS_mips                := mipsel-linux-gnu
>  -CLANG_TARGET_FLAGS_powerpc     := powerpc64le-linux-gnu
>  +CLANG_TARGET_FLAGS_powerpc     := powerpc64-linux-gnu
>   CLANG_TARGET_FLAGS_riscv       := riscv64-linux-gnu
>   CLANG_TARGET_FLAGS_s390                := s390x-linux-gnu
>   CLANG_TARGET_FLAGS_x86         := x86_64-linux-gnu
> 
> 
> 
> We do not need to test all possible target triples.
> 
> 
> We can just use the one that is known to work.
> Anyway, I will apply this patch. Thanks.
> 
> 
> >
> > This was fixed in LLVM 12.0.0. We are not testing this in our CI though,
> > so we would not get bit by this (we could bump the minimum supported
> > version of LLVM to 12.0.0 for this, we have talked recently about doing
> > it for other reasons).
> >
> > I guess I cannot really think of a good reason not to do this aside from
> > that; the target triple should only affect code generation, rather than
> > tool selection (i.e., this does not take away the ability to use a
> > custom set of binutils with clang).
> >
> > However, Nick is currently OOO and I would like his opinion voiced
> > before we commit to this. Consider this a tentative:
> >
> > Acked-by: Nathan Chancellor <nathan@kernel.org>
> >
> > > ---
> > >
> > >  scripts/Makefile.clang | 8 ++------
> > >  1 file changed, 2 insertions(+), 6 deletions(-)
> > >
> > > diff --git a/scripts/Makefile.clang b/scripts/Makefile.clang
> > > index 70b354fa1cb4..9076cc939e87 100644
> > > --- a/scripts/Makefile.clang
> > > +++ b/scripts/Makefile.clang
> > > @@ -13,15 +13,11 @@ CLANG_TARGET_FLAGS_x86            := x86_64-linux-gnu
> > >  CLANG_TARGET_FLAGS_um                := $(CLANG_TARGET_FLAGS_$(SUBARCH))
> > >  CLANG_TARGET_FLAGS           := $(CLANG_TARGET_FLAGS_$(SRCARCH))
> > >
> > > -ifeq ($(CROSS_COMPILE),)
> > >  ifeq ($(CLANG_TARGET_FLAGS),)
> > > -$(error Specify CROSS_COMPILE or add '--target=' option to scripts/Makefile.clang)
> > > +$(error add '--target=' option to scripts/Makefile.clang)
> > >  else
> > >  CLANG_FLAGS  += --target=$(CLANG_TARGET_FLAGS)
> > > -endif # CLANG_TARGET_FLAGS
> > > -else
> > > -CLANG_FLAGS  += --target=$(notdir $(CROSS_COMPILE:%-=%))
> > > -endif # CROSS_COMPILE
> > > +endif
> > >
> > >  ifeq ($(LLVM_IAS),0)
> > >  CLANG_FLAGS  += -fno-integrated-as
> > > --
> > > 2.37.2
> > >
> 
> 
> 
> -- 
> Best Regards
> Masahiro Yamada

      reply	other threads:[~2023-04-19 19:24 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-04-01 17:01 [PATCH] kbuild: clang: do not use CROSS_COMPILE for target triple Masahiro Yamada
2023-04-03 14:47 ` Nathan Chancellor
2023-04-07 19:54   ` Nick Desaulniers
2023-04-16 13:02   ` Masahiro Yamada
2023-04-19 19:24     ` Nathan Chancellor [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20230419192446.GA1993276@dev-arch.thelio-3990X \
    --to=nathan@kernel.org \
    --cc=linux-kbuild@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=llvm@lists.linux.dev \
    --cc=masahiroy@kernel.org \
    --cc=maskray@google.com \
    --cc=ndesaulniers@google.com \
    --cc=nicolas@fjasle.eu \
    --cc=trix@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox