On Tue, 20 May 2025, dm...@proton.me wrote: > On Tue, May 20, 2025 at 05:24:33PM +0200, Jan Beulich wrote: > > On 16.05.2025 04:29, dm...@proton.me wrote: > > > --- a/xen/arch/x86/include/asm/domain.h > > > +++ b/xen/arch/x86/include/asm/domain.h > > > @@ -494,6 +494,12 @@ struct arch_domain > > > X86_EMU_PIT | X86_EMU_USE_PIRQ | \ > > > X86_EMU_VPCI) > > > > > > +/* User-selectable features. */ > > > +#define X86_EMU_OPTIONAL (X86_EMU_USE_PIRQ) > > > + > > > +#define X86_EMU_BASELINE (X86_EMU_ALL & ~(X86_EMU_VPCI | \ > > > + X86_EMU_OPTIONAL)) > > > > That is, VPCI is neither baseline nor optional. Certainly at least strange.
I think Denis tried to keep the code more similar to the original. This way it is easier to review the code change and it seems correct. But at the same time it is easier to spot inconsistency that were present even before the patch. > IMO, X86_EMU_OPTIONAL should include both VPCI and PIRQ. It looks that way to me too. However, then we need to be careful as the check would differ from the original, but maybe that's OK. We want vPCI to be potentially exposed to DomUs as well. > But that will be a behavior change: AFAIU, VPCI is injected implicitly for > dom0 > case only, "default" xl toolstack currently excludes VPCI for HVM domains. > > Do I understand it correctly that "BASELINE" in the symbol name is a concern?
