On Wed, Jul 03, 2013 at 10:57:35AM +0200, Paolo Bonzini wrote: > Il 03/07/2013 10:53, Gleb Natapov ha scritto: > >>> > > Please wait a while. I can not understand it very clearly. > >>> > > > >>> > > This conditional check will cause caching a overflow value into mmio spte. > >>> > > The simple case is that kvm adds new slots for many times, the mmio-gen is easily > >>> > > more than MMIO_MAX_GEN. > >>> > > > >> > > >> > Actually, the double zapping can be avoided by moving kvm_mmu_invalidate_mmio_sptes to > >> > the end of install_new_memslots(). > >> > > > Exactly. Why should we hide it in obscure functions? > > Because kvm_mmu_invalidate_mmio_sptes is an x86-specific function, and > we already have a good arch API to hook into __kvm_set_memory_region. > >From this patch it can be seen that the API is not good enough. Generation update API is missing. > Another possible implementation is to cache the last memslot generation, > check it in kvm_arch_prepare/commit_memory_region, and call > kvm_mmu_invalidate_mmio_sptes if it changed. This would avoid the need > to keep "((change == KVM_MR_DELETE) || (change == KVM_MR_MOVE))" in sync > between the two places. But I don't think it is a substantial improvement. > Why are you trying to overcome API shortcomings instead of fixing the API. -- Gleb. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
