Re: [RESEND PATCH] arm64: fix alternatives with LLVM's integrated assembler

[Nick Desaulniers]

On Thu, Nov 14, 2019 at 8:55 AM Will Deacon <will@xxxxxxxxxx> wrote:
>
> Hi Sami,
>
> Sorry -- I thought I'd already replied to this, but it had actually
> slipped through the cracks.
>
> On Thu, Oct 31, 2019 at 12:46:52PM -0700, Sami Tolvanen wrote:
> > LLVM's integrated assembler fails with the following error when
> > building KVM:
> >
> >   <inline asm>:12:6: error: expected absolute expression
> >    .if kvm_update_va_mask == 0
> >        ^
> >   <inline asm>:21:6: error: expected absolute expression
> >    .if kvm_update_va_mask == 0
> >        ^
> >   <inline asm>:24:2: error: unrecognized instruction mnemonic
> >           NOT_AN_INSTRUCTION
> >           ^
> >   LLVM ERROR: Error parsing inline asm
> >
> > These errors come from ALTERNATIVE_CB and __ALTERNATIVE_CFG,
> > which test for the existence of the callback parameter in inline
> > assembly using the following expression:
> >
> >   " .if " __stringify(cb) " == 0\n"
> >
> > This works with GNU as, but isn't supported by LLVM. This change
> > splits __ALTERNATIVE_CFG and ALTINSTR_ENTRY into separate macros
> > to fix the LLVM build.
>
> Please could you explain a bit more about the failure and why LLVM's
> integrated assembler rejects this?

There are currently more than one issue with `.if` assembler
directives we're tracking against Clang's integrated assembler
currently, particularly around the handling of special cases related
to "fragments."
Recommended reading:
https://eli.thegreenplace.net/2013/01/03/assembler-relaxation
This particular case looks like the error is related to referring to
section before it has been seen.  My current understanding is that
Clang's integrated assembler is one pass, unlike GAS, so it chokes on
references to symbols it has not yet seen.

> Could we use something like .ifb or
> .ifeqs instead?
>
> Thanks,
>
> Will



-- 
Thanks,
~Nick Desaulniers