On 25.10.2023 17:58, Julien Grall wrote: > On 25/10/2023 09:18, Jan Beulich wrote: >> On 24.10.2023 21:59, Stefano Stabellini wrote: >>> If I understood correctly I am fine with that. To make sure we are all >>> on the same page, can you provide a couple of samples? >> >> Taking the earlier example, instead of DRIVERS_PASSTHROUGH_VTD_DMAR_H it >> would then be VTD_DMAR_H. arch/x86/pv/mm.h would use PV_MM_H, but then >> you can already see that a hypothetical arch/x86/mm.h would use X86_MM_H, >> thus colliding with what your proposal would also yield for >> arch/x86/include/asm/mm.h. So maybe private header guards should come >> with e.g. a trailing underscore? Or double underscores as component >> separators, where .../include/... use only single underscores? Or >> headers in arch/*/include/asm/ use ASM_<name>_H (i.e. not making the >> architecture explicit in the guard name, on the grounds that headers >> from multiple architectures shouldn't be included at the same time)? > Thanks for providing some details. The proposal for private headers > make sense. For arch/.../include/asm/ headers, I would strongly prefer > if we use prefix them with ASM_*. > > As a side note, I am guessing for asm-generic, we would want to use > ASM_GENERIC_* for the guard prefix. Is that correct?
That was an assumption I was working from, yes. Could also be just GENERIC_ afaic. Jan
