On 18.07.2023 19:01, Stewart Hildebrand wrote: > On 7/7/23 02:59, Jan Beulich wrote: >> On 07.07.2023 03:47, Stewart Hildebrand wrote: >>> --- a/xen/arch/arm/include/asm/domain.h >>> +++ b/xen/arch/arm/include/asm/domain.h >>> @@ -298,8 +298,7 @@ static inline void arch_vcpu_block(struct vcpu *v) {} >>> >>> #define arch_vm_assist_valid_mask(d) (1UL << >>> VMASST_TYPE_runstate_update_flag) >>> >>> -/* vPCI is not available on Arm */ >>> -#define has_vpci(d) ({ (void)(d); false; }) >>> +#define has_vpci(d) ({ IS_ENABLED(CONFIG_HAS_VPCI) && >>> is_hardware_domain(d); }) >> >> While likely not much of a problem here, I think we should strive to >> write macros such that their arguments are evaluated exactly once in >> all cases (for side effects to occur exactly once). When that's not >> easily possible, so be it, but here it doesn't look problematic to >> swap both sides of the &&. > > Thanks for pointing this out. Hmm... I'm considering turning it into a static > inline function. This would also satisfy MISRA C:2012 Dir 4.9: "A function > should be used in preference to a function-like macro where they are > interchangeable" [1].
I don't think that'll work prior to us splitting type definitions into separate headers. You simply cannot deref d at this point (or in fact at any point within this header), as struct domain hasn't been defined yet. Jan > [1] > https://gitlab.com/MISRA/MISRA-C/MISRA-C-2012/Example-Suite/-/blob/master/D_04_09.c