lkml.org 
[lkml]   [2019]   [Oct]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] arm64: lse: fix LSE atomics with LLVM's integrated assembler
On Mon, Oct 07, 2019 at 01:28:19PM -0700, Nick Desaulniers wrote:
> On Mon, Oct 7, 2019 at 1:14 PM 'Sami Tolvanen' via Clang Built Linux
> <clang-built-linux@googlegroups.com> wrote:
> >
> > Unlike gcc, clang considers each inline assembly block to be independent
> > and therefore, when using the integrated assembler for inline assembly,
> > any preambles that enable features must be repeated in each block.
> >
> > Instead of changing all inline assembly blocks that use LSE, this change
> > adds -march=armv8-a+lse to KBUILD_CFLAGS, which works with both clang
> > and gcc.
> >
> > Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
>
> Thanks Sami, looks like this addresses:
> Link: https://github.com/ClangBuiltLinux/linux/issues/671
> I tried adding `.arch armv8-a+lse` directives to all of the inline asm:
> https://github.com/ClangBuiltLinux/linux/issues/573#issuecomment-535098996
> though that quickly ran aground in some failed assertion using the
> alternatives system that I don't quite yet understand.

I think these issues somehow are related to the strange way we used to
jump to the out-of-line fallbacks. Since around addfc38672c7 ("arm64:
atomics: avoid out-of-line ll/sc atomics") this approach was removed.

I reproduced your patch on 5.4-rc2 and no longer get the clang build
errors. Therefore this approach is viable option.

>
> One thing to be careful about is that blankets the entire kernel in
> `+lse`, allowing LSE atomics to be selected at any point. The
> assembler directive in the header (or per inline asm block) is more
> fine grained. I'm not sure that they would be guarded by the
> alternative system. Maybe that's not an issue, but it is the reason I
> tried to localize the assembler directive first.
>
> Note that Clang really wants the target arch specified by either:
> 1. command line argument (as in this patch)

This makes me nervous, as we're telling the compiler that the machine
we're building for has LSE - obviously it would be perfectly acceptable
for the compiler to then throw in some LSE instructions at some point.
Thus something may break further down the line.

> 2. per function function attribute
> 3. per asm statement assembler directive

Keen to hear Will's thoughts - but I'd suggest this is probably the
safest way forward.

Thanks,

Andrew Murray

>
> 1 is the smallest incision.
>
> > ---
> > arch/arm64/Makefile | 2 ++
> > arch/arm64/include/asm/lse.h | 2 --
> > 2 files changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/arm64/Makefile b/arch/arm64/Makefile
> > index 84a3d502c5a5..7a7c0cb8ed60 100644
> > --- a/arch/arm64/Makefile
> > +++ b/arch/arm64/Makefile
> > @@ -36,6 +36,8 @@ lseinstr := $(call as-instr,.arch_extension lse,-DCONFIG_AS_LSE=1)
> > ifeq ($(CONFIG_ARM64_LSE_ATOMICS), y)
> > ifeq ($(lseinstr),)
> > $(warning LSE atomics not supported by binutils)
> > + else
> > +KBUILD_CFLAGS += -march=armv8-a+lse
> > endif
> > endif
> >
> > diff --git a/arch/arm64/include/asm/lse.h b/arch/arm64/include/asm/lse.h
> > index 80b388278149..8603a9881529 100644
> > --- a/arch/arm64/include/asm/lse.h
> > +++ b/arch/arm64/include/asm/lse.h
> > @@ -14,8 +14,6 @@
> > #include <asm/atomic_lse.h>
> > #include <asm/cpucaps.h>
> >
> > -__asm__(".arch_extension lse");
> > -
> > extern struct static_key_false cpu_hwcap_keys[ARM64_NCAPS];
> > extern struct static_key_false arm64_const_caps_ready;
> >
> > --
> > 2.23.0.581.g78d2f28ef7-goog
> >
> > --
> > You received this message because you are subscribed to the Google Groups "Clang Built Linux" group.
> > To unsubscribe from this group and stop receiving emails from it, send an email to clang-built-linux+unsubscribe@googlegroups.com.
> > To view this discussion on the web visit https://groups.google.com/d/msgid/clang-built-linux/20191007201452.208067-1-samitolvanen%40google.com.
>
>
>
> --
> Thanks,
> ~Nick Desaulniers

\
 
 \ /
  Last update: 2019-10-08 10:37    [W:0.152 / U:10.460 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site