On 14/03/2025 14.42, Will Deacon wrote:
> On Fri, Mar 14, 2025 at 01:05:15PM +0100, Arnd Bergmann wrote:
>> On Fri, Mar 14, 2025, at 12:55, Will Deacon wrote:
>>> On Fri, Mar 14, 2025 at 08:09:39AM +0100, Thomas Huth wrote:
>>>> __ASSEMBLY__ is only defined by the Makefile of the kernel, so
>>>> this is not really useful for uapi headers (unless the userspace
>>>> Makefile defines it, too). Let's switch to __ASSEMBLER__ which
>>>> gets set automatically by the compiler when compiling assembly
>>>> code.
>>>>
>>>> Cc: Catalin Marinas <catalin.marinas@arm.com>
>>>> Cc: Will Deacon <will@kernel.org>
>>>> Signed-off-by: Thomas Huth <thuth@redhat.com>
>>>> ---
>>>> arch/arm64/include/uapi/asm/kvm.h | 2 +-
>>>> arch/arm64/include/uapi/asm/ptrace.h | 4 ++--
>>>> arch/arm64/include/uapi/asm/sigcontext.h | 4 ++--
>>>> 3 files changed, 5 insertions(+), 5 deletions(-)
>>>
>>> Is there a risk of breaking userspace with this? I wonder if it would
>>> be more conservative to do something like:
>>>
>>> #if !defined(__ASSEMBLY__) && !defined(__ASSEMBLER__)
>>>
>>> so that if somebody is doing '#define __ASSEMBLY__' then they get the
>>> same behaviour as today.
>>>
>>> Or maybe we don't care?
>>
>> I think the main risk we would have is user applications relying
>> on the __ASSEMBLER__ checks in new kernel headers and not defining
>> __ASSEMBLY__. This would result in the application not building
>> against old kernel headers that only check against __ASSEMBLY__.
>
> Hmm. I hadn't thought about the case of old headers :/
>
> A quick Debian codesearch shows that glibc might #define __ASSEMBLY__
> for some arch-specific headers:
>
> https://codesearch.debian.net/search?q=%23define+__ASSEMBLY__&literal=1
>
> which is what I was more worried about.
Since both, GCC and Clang, define __ASSEMBLER__ since a long time (Arnd
checked GCC 2.95, and I checked that at least Clang 7.0 still has it), I
think the only problem might be other compiler toolchains that might not set
__ASSEMBLER__ automatically. I just checked Tiny-C 0.9.27, and that also
sets __ASSEMBLER__ already. And according to
https://github.com/IanHarvey/pcc/blob/master/cc/cc/cc.1#L405 it is also set
in PCC.
I haven't spotted it in LCC though (which seems to be an old C89 compiler if
I got it right). So if we are worried about such exotic old compilers, it's
maybe better to check both, __ASSEMBLY__ and __ASSEMBLER__ in the uapi
files? Or would it be ok to force those few people to set __ASSEMBLER__
manually in their Makefiles (just like they had to do before with
__ASSEMBLY__) in case they want to compile assembler code with such exotic
compilers and new kernel headers?
Thomas