Re: [PATCH] KVM: x86/tdp_mmu: Fix base gfn check when zapping private huge SPTE

[Sean Christopherson <seanjc@google.com>]

On Mon, Mar 09, 2026, pcjer wrote:
> Signed-off-by: pcjer <pcj3195161583@163.com>
> ---
>  arch/x86/kvm/mmu/tdp_mmu.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
> index 1266d5452..8482a85d6 100644
> --- a/arch/x86/kvm/mmu/tdp_mmu.c
> +++ b/arch/x86/kvm/mmu/tdp_mmu.c
> @@ -1025,8 +1025,8 @@ static bool tdp_mmu_zap_leafs(struct kvm *kvm, struct kvm_mmu_page *root,
>  
>  			slot = gfn_to_memslot(kvm, gfn);
>  			if (kvm_hugepage_test_mixed(slot, gfn, iter.level) ||
> -			    (gfn & mask) < start ||
> -			    end < (gfn & mask) + KVM_PAGES_PER_HPAGE(iter.level)) {
> +			    (gfn & ~mask) < start ||
> +			    end < (gfn & ~mask) + KVM_PAGES_PER_HPAGE(iter.level)) {

Somewhat to my surprise, this does indeed look like a legitimate fix, ignoring
that the code in question was never merged and was lasted posted 2+ years ago[*]
(and has long since been replaced).

The bug likely went unnoticed during development because "(gfn & mask) < start"
would almost always be true (mask == 511 for a 2MiB page).  Though mask should
really just be inverted from the get go in this code

+		if (is_private && kvm_gfn_shared_mask(kvm) &&
+		    is_large_pte(iter.old_spte)) {
+			gfn_t gfn = iter.gfn & ~kvm_gfn_shared_mask(kvm);
+			gfn_t mask = KVM_PAGES_PER_HPAGE(iter.level) - 1;
+
+			struct kvm_memory_slot *slot;
+			struct kvm_mmu_page *sp;
+

[*] https://lore.kernel.org/all/c656573ccc68e212416d323d35f884bff25e6e2d.1708933624.git.isaku.yamahata@intel.com