Re: [PATCH v3 17/17] KVM: x86/tdp_mmu: Take root types for kvm_tdp_mmu_invalidate_all_roots()

[Edgecombe, Rick P]

On Tue, 2024-06-25 at 14:14 +0800, Yan Zhao wrote:
> > Explain why not to zap invalid mirrored roots?
> No. Explain why not to zap invalid direct roots.
> Just passing KVM_DIRECT_ROOTS will zap only valid direct roots, right?
> The original kvm_tdp_mmu_zap_all() "Zap all roots, including invalid roots".
> It might be better to explain why not to zap invalid direct roots here.

Hmm, right. We don't actually have enum value to iterate all direct roots (valid
and invalid). We could change tdp_mmu_root_match() to:
diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
index 8320b093fef9..bcd771a8835f 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.c
+++ b/arch/x86/kvm/mmu/tdp_mmu.c
@@ -98,8 +98,8 @@ static bool tdp_mmu_root_match(struct kvm_mmu_page *root,
        if (WARN_ON_ONCE(!(types & KVM_VALID_ROOTS)))
                return false;
 
-       if (root->role.invalid)
-               return types & KVM_INVALID_ROOTS;
+       if (root->role.invalid && !(types & KVM_INVALID_ROOTS))
+               return false;
        if (likely(!is_mirror_sp(root)))
                return types & KVM_DIRECT_ROOTS;
 
Maybe better than a comment...? Need to put the whole thing together and test it
still.