On 06/12/18 23:34, Boris Ostrovsky wrote: > On 12/6/18 5:11 PM, Paolo Bonzini wrote: >> On 06/12/18 07:04, Maran Wilson wrote: >>> +config PVH >>> + bool "Support for running PVH guests" >>> + ---help--- >>> + This option enables the PVH entry point for guest virtual machines >>> + as specified in the x86/HVM direct boot ABI. >>> + >> IIUC this breaks "normal" bzImage boot, so we should have something like >> >> The resulting kernel will not boot with most x86 boot loaders >> such as GRUB > > > Grub support for PVH guests (for Xen) is well under way. Oh, nice. :) >> or SYSLINUX. Unless you plan to start the kernel >> using QEMU or Xen, you probably want to say N here. > > I think PVH should not be user-selectable at all. It should be selected > by either XEN_PVH or KVM_GUEST_PVH (which you suggested to drop). KVM_GUEST_PVH is not entirely accurate because it's not just for KVM (it can be used with QEMU and Apple's Hypervisor.framework for example). It's also not necessarily just for QEMU (it could be implemented for kvmtool if desired), but as long as it's in the help I guess it's acceptable. I think we could just drop the sentence about boot loaders from my suggestion. >> >> and also >> >> depends on !EFI >> >> because even though in principle it would be possible to write a PVH >> loader for UEFI, PVH's start info does not support the EFI handover >> protocol. > > But we should be able to build the binary with both EFI and PVH? Can you? It's a completely different binary format, the EFI handover protocol is invoked via a special entry point and needs the Linux header format, not ELF. Paolo
