On Fri, Apr 26, 2024 at 07:44:40AM +0200, Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote: > On Thu, Apr 25, 2024 at 6:51 PM Isaku Yamahata <isaku.yamahata@xxxxxxxxx> wrote: > > > AFAIK, unwinding on failure is completely uninteresting, and arguably undesirable, > > > because undoing LAUNCH_UPDATE or PAGE.ADD will affect the measurement, i.e. there > > > is no scenario where deleting pages from guest_memfd would allow a restart/resume > > > of the build process to truly succeed. > > > > > > Just for record. With the following twist to kvm_gmem_populate, > > KVM_TDX_INIT_MEM_REGION can use kvm_gmem_populate(). For those who are curious, > > I also append the callback implementation at the end. > > Nice, thank you very much. Since TDX does not need > HAVE_KVM_GMEM_PREPARE, if I get rid of FGP_CREAT_ONLY it will work for > you, right? Yes, that's right. -- Isaku Yamahata <isaku.yamahata@xxxxxxxxx>
