Re: [PATCH V4 23/23] crypto: hisilicon/qm: Workaround to enable build with RISC-V clang

[Sunil V L]

Hi Conor,

On Tue, Apr 04, 2023 at 10:59:41PM +0100, Conor Dooley wrote:
> Hey Sunil,
> 
> This one made me scratch my head for a bit..
> 
> On Tue, Apr 04, 2023 at 11:50:37PM +0530, Sunil V L wrote:
> > With CONFIG_ACPI enabled for RISC-V, this driver gets enabled in
> > allmodconfig build. The gcc tool chain builds this driver removing the
> > inline arm64 assembly code. However, clang for RISC-V tries to build
> > the arm64 assembly and below error is seen.
> 
> There's actually nothing RISC-V specific about that behaviour, that's
> just how clang works. Quoting Nathan:
> "Clang performs semantic analysis (i.e., validates assembly) before
> dead code elimination, so IS_ENABLED() is not sufficient for avoiding
> that error."
> 
Huh, It never occurred to me that this issue could be known already since I
always thought we are hitting this first time since ACPI is enabled only
now for RISC-V. Thank you very much!. 

> > drivers/crypto/hisilicon/qm.c:627:10: error: invalid output constraint '+Q' in asm
> >                        "+Q" (*((char __iomem *)fun_base))
> >                        ^
> > It appears that RISC-V clang is not smart enough to detect
> > IS_ENABLED(CONFIG_ARM64) and remove the dead code.
> 
> So I think this statement is just not true, it can remove dead code, but
> only after it has done the semantic analysis.
>
Yes, with more details now, let me update the commit message.
 
> The reason that this has not been seen before, again quoting Nathan, is:
> "arm64 and x86_64 both support the Q constraint, we cannot build
> LoongArch yet (although it does not have support for Q either so same
> boat as RISC-V), and ia64 is dead/unsupported in LLVM. Those are the
> only architectures that support ACPI, so I guess that explains why we
> have seen no issues aside from RISC-V so far."
> 
> > As a workaround, move this check to preprocessing stage which works
> > with the RISC-V clang tool chain.
> 
> I don't think there's much else you can do!
> Reviewed-by: Conor Dooley <conor.dooley@xxxxxxxxxxxxx>
> 
> Perhaps it is also worth adding:
> Link: https://github.com/ClangBuiltLinux/linux/issues/999
> 
Sure, Thank you very much for digging this!

Thanks,
Sunil