On 08/10/2018 21:28, Vitaly Kuznetsov wrote: > Change since v3 [Sean Christopherson]: > - Add Reviewed-by tags (thanks!). > - Drop stale role initializer in kvm_calc_shadow_ept_root_page_role > (interim change in PATCH4, the end result is the same). > - Use '!!' instead of '!= 0' for kvm_read_cr4_bits() readings. > > Also, rebased to the current kvm/queue. > > Original description: > > Currently, when we switch from L1 to L2 (VMX) we do the following: > - Re-initialize L1 MMU as shadow EPT MMU (nested_ept_init_mmu_context()) > - Re-initialize 'nested' MMU (nested_vmx_load_cr3() -> init_kvm_nested_mmu()) > > When we switch back we do: > - Re-initialize L1 MMU (nested_vmx_load_cr3() -> init_kvm_tdp_mmu()) > > This seems to be sub-optimal. Initializing MMU is expensive (thanks to > update_permission_bitmask(), update_pkru_bitmask(),..) Try solving the > issue by splitting L1-normal and L1-nested MMUs and checking if MMU reset > is really needed. This spares us about 1000 cpu cycles on nested vmexit. > > Brief look at SVM makes me think it can be optimized the exact same way, > I'll do this in a separate series. > > Paolo Bonzini (1): > x86/kvm/mmu: get rid of redundant kvm_mmu_setup() > > Vitaly Kuznetsov (8): > x86/kvm/mmu: make vcpu->mmu a pointer to the current MMU > x86/kvm/mmu.c: set get_pdptr hook in kvm_init_shadow_ept_mmu() > x86/kvm/mmu.c: add kvm_mmu parameter to kvm_mmu_free_roots() > x86/kvm/mmu: introduce guest_mmu > x86/kvm/mmu: make space for source data caching in struct kvm_mmu > x86/kvm/nVMX: introduce source data cache for > kvm_init_shadow_ept_mmu() > x86/kvm/mmu: check if tdp/shadow MMU reconfiguration is needed > x86/kvm/mmu: check if MMU reconfiguration is needed in > init_kvm_nested_mmu() > > arch/x86/include/asm/kvm_host.h | 44 +++- > arch/x86/kvm/mmu.c | 357 +++++++++++++++++++------------- > arch/x86/kvm/mmu.h | 8 +- > arch/x86/kvm/mmu_audit.c | 12 +- > arch/x86/kvm/paging_tmpl.h | 15 +- > arch/x86/kvm/svm.c | 14 +- > arch/x86/kvm/vmx.c | 55 +++-- > arch/x86/kvm/x86.c | 22 +- > 8 files changed, 322 insertions(+), 205 deletions(-) > Queued, thanks. Paolo
