Re: [PATCH v9 4/5] x86/cpuid: fix unbootable VMs by really inlining memcmp() in hypervisor_cpuid_base()

[Mauricio Faria de Oliveira]

On 2026-09-15 12:48, Borislav Petkov wrote:
> On Mon, Sep 14, 2026 at 10:11:55PM -0700, Borislav Petkov wrote:
>> Hmm, never heard that before but that doesn't mean anything. Probably should
>> mention it in one of the commit messages so that we know for the future...
> 
> My compiler friends are telling me that gcc doesn't do that now but it might
> in the future, perhaps.
> 
> -fno-builtin-memcmp
> 
> will stop any of that behavior so if, perhaps, building the pvh code with it
> looks better as a solution, we probably could use it.

It looks like clang does [1], but according to an LLM that is also
disabled with -fno-builtin-memcmp.

IMO a fix in the source file is clearer, less obfuscated than in a
Makefile (and the source file has to be changed anyway). It also has the
potential advantage that, if the Makefile change slips in a future
change/refactor, the issue shouldn't reappear.

Either way, please let me know and I can certainly change the patch.

Thanks,

[1]
https://github.com/llvm/llvm-project/commit/9473c01e968bc4ecf5ea7c54ef0c84b687d365ea

-- 
Mauricio