[PATCH] KVM: arm64: nv: Skip vCPUs without a pseudo-TLB in invalidate_vncr_va()

[Marc Zyngier]

On Sun, 07 Jun 2026 09:43:53 +0100,
Hyunwoo Kim <imv4bel at gmail.com> wrote:
> 
> vncr_tlb is not allocated before a vCPU runs for the first time, so
> vcpu->arch.vncr_tlb is NULL for a vCPU that has been created but not yet
> run. Code that iterates over every vCPU's pseudo-TLB must skip those.
> 
> invalidate_vncr_va() iterates over the vCPUs with kvm_for_each_vcpu() and
> dereferences vt->valid without checking whether vncr_tlb is NULL.
> 
> While iterating, skip vCPUs whose pseudo-TLB has not been allocated.
>
> Fixes: 4ffa72ad8f37 ("KVM: arm64: nv: Add S1 TLB invalidation primitive for VNCR_EL2")
> Signed-off-by: Hyunwoo Kim <imv4bel at gmail.com>
> ---
>  arch/arm64/kvm/nested.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/arch/arm64/kvm/nested.c b/arch/arm64/kvm/nested.c
> index 6f7bc9a9992e..063e079d1d1a 100644
> --- a/arch/arm64/kvm/nested.c
> +++ b/arch/arm64/kvm/nested.c
> @@ -969,6 +969,10 @@ static void invalidate_vncr_va(struct kvm *kvm,
>  		struct vncr_tlb *vt = vcpu->arch.vncr_tlb;
>  		u64 va_start, va_end, va_size;
>  
> +		/* Skip vCPUs whose pseudo-TLB hasn't been allocated yet */
> +		if (!vt)
> +			continue;
> +
>  		if (!vt->valid)
>  			continue;
>  

This looks correct and matches what we already have for
invalidate_vncr_ipa().

But I think this misses the opportunity to squash a whole class of
similar bugs, should we ever have the need for another function that
iterates over all *valid* VNCR pseudo-TLBs.

Since I'm on a train and have nothing better to do, I've written the
following hack.

Thoughts?

	M.

diff --git a/arch/arm64/kvm/nested.c b/arch/arm64/kvm/nested.c
index 38f672e940878..f0a9f81a08302 100644
--- a/arch/arm64/kvm/nested.c
+++ b/arch/arm64/kvm/nested.c
@@ -897,9 +897,21 @@ static void invalidate_vncr(struct vncr_tlb *vt)
 		clear_fixmap(vncr_fixmap(vt->cpu));
 }
 
+/*
+ * VNCR TLB invalidation occurs from MMU notifiers or TLBI instructions, and
+ * either can race against a vcpu not being onlined yet (no pseudo-TLB
+ * allocated). Similarly, the TLB might be invalid.  Skip those, as they
+ * obviously don't participate in the invalidation at this stage.
+ */
+#define kvm_for_each_vncr_tlb(idx, vcpup, tlbp, kvm)	\
+	kvm_for_each_vcpu(idx, vcpu, kvm)		\
+		if (((tlbp) = vcpu->arch.vncr_tlb) &&	\
+		    (tlbp)->valid)
+
 static void kvm_invalidate_vncr_ipa(struct kvm *kvm, u64 start, u64 end)
 {
 	struct kvm_vcpu *vcpu;
+	struct vncr_tlb *vt;
 	unsigned long i;
 
 	lockdep_assert_held_write(&kvm->mmu_lock);
@@ -907,24 +919,9 @@ static void kvm_invalidate_vncr_ipa(struct kvm *kvm, u64 start, u64 end)
 	if (!kvm_has_feat(kvm, ID_AA64MMFR4_EL1, NV_frac, NV2_ONLY))
 		return;
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
-		struct vncr_tlb *vt = vcpu->arch.vncr_tlb;
+	kvm_for_each_vncr_tlb(i, vcpu, vt, kvm) {
 		u64 ipa_start, ipa_end, ipa_size;
 
-		/*
-		 * Careful here: We end-up here from an MMU notifier,
-		 * and this can race against a vcpu not being onlined
-		 * yet, without the pseudo-TLB being allocated.
-		 *
-		 * Skip those, as they obviously don't participate in
-		 * the invalidation at this stage.
-		 */
-		if (!vt)
-			continue;
-
-		if (!vt->valid)
-			continue;
-
 		ipa_size = ttl_to_size(pgshift_level_to_ttl(vt->wi.pgshift,
 							    vt->wr.level));
 		ipa_start = vt->wr.pa & ~(ipa_size - 1);
@@ -954,17 +951,14 @@ static void invalidate_vncr_va(struct kvm *kvm,
 			       struct s1e2_tlbi_scope *scope)
 {
 	struct kvm_vcpu *vcpu;
+	struct vncr_tlb *vt;
 	unsigned long i;
 
 	lockdep_assert_held_write(&kvm->mmu_lock);
 
-	kvm_for_each_vcpu(i, vcpu, kvm) {
-		struct vncr_tlb *vt = vcpu->arch.vncr_tlb;
+	kvm_for_each_vncr_tlb(i, vcpu, vt, kvm) {
 		u64 va_start, va_end, va_size;
 
-		if (!vt->valid)
-			continue;
-
 		va_size = ttl_to_size(pgshift_level_to_ttl(vt->wi.pgshift,
 							   vt->wr.level));
 		va_start = vt->gva & ~(va_size - 1);

-- 
Jazz isn't dead. It just smells funny.