vcpu_allowed_register_width() checks if all the VCPUs are either all 32bit or all 64bit. Since the checking is done even for vCPUs that are not initialized (KVM_ARM_VCPU_INIT has not been done) yet, the non-initialized vCPUs are erroneously treated as 64bit vCPU, which causes the function to incorrectly detect a mixed-width VM. Fix vcpu_allowed_register_width() to skip the check for vCPUs that are not initialized yet. Fixes: 66e94d5cafd4 ("KVM: arm64: Prevent mixed-width VM creation") Signed-off-by: Reiji Watanabe <reijiw@xxxxxxxxxx> --- arch/arm64/kvm/reset.c | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/arch/arm64/kvm/reset.c b/arch/arm64/kvm/reset.c index 426bd7fbc3fd..ef78bbc7566a 100644 --- a/arch/arm64/kvm/reset.c +++ b/arch/arm64/kvm/reset.c @@ -180,8 +180,19 @@ static bool vcpu_allowed_register_width(struct kvm_vcpu *vcpu) if (kvm_has_mte(vcpu->kvm) && is32bit) return false; + /* + * Make sure vcpu->arch.target setting is visible from others so + * that the width consistency checking between two vCPUs is done + * by at least one of them at KVM_ARM_VCPU_INIT. + */ + smp_mb(); + /* Check that the vcpus are either all 32bit or all 64bit */ kvm_for_each_vcpu(i, tmp, vcpu->kvm) { + /* Skip if KVM_ARM_VCPU_INIT is not done for the vcpu yet */ + if (tmp->arch.target == -1) + continue; + if (vcpu_has_feature(tmp, KVM_ARM_VCPU_EL1_32BIT) != is32bit) return false; } base-commit: df0cc57e057f18e44dac8e6c18aba47ab53202f9 -- 2.34.1.575.g55b058a8bb-goog