| Message ID | 20180907115947.1947-1-richard.weiyang@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v3] KVM: x86: don't reset root in kvm_mmu_setup() | expand |
On 07/09/2018 13:59, Wei Yang wrote: > Here is the code path which shows kvm_mmu_setup() is invoked after > kvm_mmu_create(). Since kvm_mmu_setup() is only invoked in this code path, > this means the root_hpa and prev_roots are guaranteed to be invalid. And > it is not necessary to reset it again. > > kvm_vm_ioctl_create_vcpu() > kvm_arch_vcpu_create() > vmx_create_vcpu() > kvm_vcpu_init() > kvm_arch_vcpu_init() > kvm_mmu_create() > kvm_arch_vcpu_setup() > kvm_mmu_setup() > kvm_init_mmu() > > This patch set reset_roots to false in kmv_mmu_setup(). > > Signed-off-by: Wei Yang <richard.weiyang@gmail.com> > Reviewed-by: Liran Alon <liran.alon@oracle.com> > > --- > v3: add a comment in code based on Liran comment. > v2: adjust changelog based on Sean Christopherson comment > --- > arch/x86/kvm/mmu.c | 7 ++++++- > 1 file changed, 6 insertions(+), 1 deletion(-) > > diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c > index b2b88454b1d8..899c029cff0d 100644 > --- a/arch/x86/kvm/mmu.c > +++ b/arch/x86/kvm/mmu.c > @@ -5413,7 +5413,12 @@ void kvm_mmu_setup(struct kvm_vcpu *vcpu) > { > MMU_WARN_ON(VALID_PAGE(vcpu->arch.mmu.root_hpa)); > > - kvm_init_mmu(vcpu, true); > + /* > + * kvm_mmu_setup() is called only on vCPU initialization. > + * Therefore, no need to reset mmu roots as they are not yet > + * initialized. > + */ > + kvm_init_mmu(vcpu, false); > } > > static void kvm_mmu_invalidate_zap_pages_in_memslot(struct kvm *kvm, > Queued, thanks. Paolo
diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c index b2b88454b1d8..899c029cff0d 100644 --- a/arch/x86/kvm/mmu.c +++ b/arch/x86/kvm/mmu.c @@ -5413,7 +5413,12 @@ void kvm_mmu_setup(struct kvm_vcpu *vcpu) { MMU_WARN_ON(VALID_PAGE(vcpu->arch.mmu.root_hpa)); - kvm_init_mmu(vcpu, true); + /* + * kvm_mmu_setup() is called only on vCPU initialization. + * Therefore, no need to reset mmu roots as they are not yet + * initialized. + */ + kvm_init_mmu(vcpu, false); } static void kvm_mmu_invalidate_zap_pages_in_memslot(struct kvm *kvm,
Here is the code path which shows kvm_mmu_setup() is invoked after kvm_mmu_create(). Since kvm_mmu_setup() is only invoked in this code path, this means the root_hpa and prev_roots are guaranteed to be invalid. And it is not necessary to reset it again. kvm_vm_ioctl_create_vcpu() kvm_arch_vcpu_create() vmx_create_vcpu() kvm_vcpu_init() kvm_arch_vcpu_init() kvm_mmu_create() kvm_arch_vcpu_setup() kvm_mmu_setup() kvm_init_mmu() This patch set reset_roots to false in kmv_mmu_setup(). Signed-off-by: Wei Yang <richard.weiyang@gmail.com> Reviewed-by: Liran Alon <liran.alon@oracle.com> --- v3: add a comment in code based on Liran comment. v2: adjust changelog based on Sean Christopherson comment --- arch/x86/kvm/mmu.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)