On Mon, Nov 19, 2018 at 06:25:35PM +0100, Michael Mueller wrote: > @@ -3521,6 +3523,10 @@ static int vcpu_post_run(struct kvm_vcpu *vcpu, int exit_reason) > vcpu->run->s.regs.gprs[15] = vcpu->arch.sie_block->gg15; > > atomic_dec(&vcpu->kvm->arch.vcpus_in_sie); > + if (vcpu->kvm->arch.gib_in_use && > + !in_alert_list(vcpu->kvm->arch.gisa) && > + !atomic_fetch_andnot(0, &vcpu->kvm->arch.vcpus_in_sie)) > + vcpu->kvm->arch.gisa->iam = vcpu->kvm->arch.iam; This looks a bit odd to me. Why not simply !atomic_read(&vcpu->kvm->arch.vcpus_in_sie)) instead of the atomic_fetch_andnot() construct? Did I miss something?
