On Thu, Apr 24, 2025 at 8:52 AM Shutemov, Kirill <kirill.shutemov@xxxxxxxxx> wrote: > > On Wed, Apr 23, 2025 at 04:09:50PM +0200, Paolo Bonzini wrote: > > On Sat, Apr 19, 2025 at 12:16 AM Edgecombe, Rick P > > <rick.p.edgecombe@xxxxxxxxx> wrote: > > > TDG.VP.VMCALL<INSTRUCTION.WBINVD> - Missing > > > TDG.VP.VMCALL<INSTRUCTION.PCONFIG> - Missing > > > > WBINVD and PCONFIG need to be implemented (PCONFIG can be a stub). > > On the guest side I actively avoided using WBINVD as the only way for VMM > to implement it is to do WBINVD on the host side which is too disruptive > for the system. It is possible way to DoS the host. > > Do we really want to implement it on KVM side? It is good incentive for > guests to avoid WBINVD. KVM does not do anything for WBINVD unless you have assigned devices with non-coherent DMA, so in practice WBINVD would basically be a nop. Paolo
