On Tue, 2024-06-18 at 07:34 -0700, Sean Christopherson wrote: > There's also option: > > c) Init disabled_quirks based on VM type. > > I.e. let userspace enable the quirk. If the VMM wants to shoot its TDX VM > guests, > then so be it. That said, I don't like this option because it would create a > very > bizarre ABI. I think we actually need to force it on for TDX because kvm_mmu_zap_all_fast() only zaps the direct (shared) root. If userspace decides to not enable the quirk, mirror/private memory will not be zapped on memslot deletion. Then later if there is a hole punch it will skip zapping that range because there is no memslot. Then won't it let the pages get freed while they are still mapped in the TD? If I got that right (not 100% sure on the gmem hole punch page freeing), I think KVM needs to force the behavior for TDs. > > > > > > > I'd prefer to go with option (a) here. Because we don't have any behavior > > > defined yet for KVM_X86_TDX_VM, we don't really need to "disable a quirk" > > > of it. > > I vote for (a) as well. > > > > Instead we could just define KVM_X86_QUIRK_SLOT_ZAP_ALL to be about the > > > behavior > > > of the existing vm_types. It would be a few lines of documentation to save > > > implementing and maintaining a whole interface with special logic for TDX. > > > So to > > > me it doesn't seem worth it, unless there is some other user for a new > > > more > > > complex quirk interface. > > What about introducing a forced disabled_quirk field? > > Nah, it'd require manual opt-in for every VM type for almost no benefit. In > fact, > IMO the code itself would be a net negative versus: > > return kvm->arch.vm_type == KVM_X86_DEFAULT_VM && > kvm_check_has_quirk(kvm, KVM_X86_QUIRK_SLOT_ZAP_ALL); > > because explicitly checking for KVM_X86_DEFAULT_VM would directly match the > documentation (which would state that the quirk only applies to DEFAULT_VM). Ok, I updated (and posted on this series) the TDX integration patch.
