Re: [PATCH 3/7] kvm: vmx: pass MSR_IA32_SPEC_CTRL and MSR_IA32_PRED_CMD down to the guest

[Liran Alon <LIRAN.ALON@ORACLE.COM>]

On 08/01/18 21:18, Jim Mattson wrote:
> Guest usage of MSR_IA32_SPEC_CTRL and MSR_IA32_PRED_CMD should be
> predicated on guest_cpuid_has(vcpu, X86_FEATURE_SPEC_CTRL).
>
> On Mon, Jan 8, 2018 at 10:08 AM, Paolo Bonzini <pbonzini@redhat.com> wrote:
>> Direct access to MSR_IA32_SPEC_CTRL and MSR_IA32_PRED_CMD is important
>> for performance.  Allow load/store of MSR_IA32_SPEC_CTRL, restore guest
>> IBRS on VM entry and set it to 0 on VM exit (because Linux does not use
>> it yet).
>>
>> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
>> ---
>>   arch/x86/kvm/vmx.c | 32 ++++++++++++++++++++++++++++++++
>>   1 file changed, 32 insertions(+)
>>
>> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
>> index 669f5f74857d..d00bcad7336e 100644
>> --- a/arch/x86/kvm/vmx.c
>> +++ b/arch/x86/kvm/vmx.c
>> @@ -120,6 +120,8 @@
>>   module_param_named(preemption_timer, enable_preemption_timer, bool, S_IRUGO);
>>   #endif
>>
>> +static bool __read_mostly have_spec_ctrl;
>> +
>>   #define KVM_GUEST_CR0_MASK (X86_CR0_NW | X86_CR0_CD)
>>   #define KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST (X86_CR0_WP | X86_CR0_NE)
>>   #define KVM_VM_CR0_ALWAYS_ON                                           \
>> @@ -609,6 +611,8 @@ struct vcpu_vmx {
>>          u64                   msr_host_kernel_gs_base;
>>          u64                   msr_guest_kernel_gs_base;
>>   #endif
>> +       u64                   spec_ctrl;
>> +
>>          u32 vm_entry_controls_shadow;
>>          u32 vm_exit_controls_shadow;
>>          u32 secondary_exec_control;
>> @@ -3361,6 +3365,9 @@ static int vmx_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
>>          case MSR_IA32_TSC:
>>                  msr_info->data = guest_read_tsc(vcpu);
>>                  break;
>> +       case MSR_IA32_SPEC_CTRL:
>> +               msr_info->data = to_vmx(vcpu)->spec_ctrl;
>> +               break;
>
> I have:
>
> if (!have_spec_ctrl ||
>      (!msr_info->host_initiated &&
>       !guest_cpuid_has(vcpu, X86_FEATURE_SPEC_CTRL)))
> return 1;
> msr_info->data = to_vmx(vcpu)->msr_ia32_spec_ctrl;
> break;
>
>>          case MSR_IA32_SYSENTER_CS:
>>                  msr_info->data = vmcs_read32(GUEST_SYSENTER_CS);
>>                  break;
>> @@ -3500,6 +3507,9 @@ static int vmx_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
>>          case MSR_IA32_TSC:
>>                  kvm_write_tsc(vcpu, msr_info);
>>                  break;
>> +       case MSR_IA32_SPEC_CTRL:
>> +               to_vmx(vcpu)->spec_ctrl = msr_info->data;
>> +               break;
>
> I have:
>
> if (!have_spec_ctrl ||
>      (!msr_info->host_initiated &&
>       !guest_cpuid_has(vcpu, X86_FEATURE_SPEC_CTRL)))
> return 1;
> to_vmx(vcpu)->msr_ia32_spec_ctrl = data;
> break;
>
>>          case MSR_IA32_CR_PAT:
>>                  if (vmcs_config.vmentry_ctrl & VM_ENTRY_LOAD_IA32_PAT) {
>>                          if (!kvm_mtrr_valid(vcpu, MSR_IA32_CR_PAT, data))
>> @@ -7062,6 +7072,17 @@ static __init int hardware_setup(void)
>>                  goto out;
>>          }
>>
>> +       /*
>> +        * FIXME: this is only needed until SPEC_CTRL is supported
>> +        * by upstream Linux in cpufeatures, then it can be replaced
>> +        * with static_cpu_has.
>> +        */
>> +       have_spec_ctrl = cpu_has_spec_ctrl();
>> +       if (have_spec_ctrl)
>> +               pr_info("kvm: SPEC_CTRL available\n");
>> +       else
>> +               pr_info("kvm: SPEC_CTRL not available\n");
>> +
>>          if (boot_cpu_has(X86_FEATURE_NX))
>>                  kvm_enable_efer_bits(EFER_NX);
>>
>> @@ -7131,6 +7152,8 @@ static __init int hardware_setup(void)
>>          vmx_disable_intercept_for_msr(MSR_IA32_SYSENTER_CS, false);
>>          vmx_disable_intercept_for_msr(MSR_IA32_SYSENTER_ESP, false);
>>          vmx_disable_intercept_for_msr(MSR_IA32_SYSENTER_EIP, false);
>> +       vmx_disable_intercept_for_msr(MSR_IA32_SPEC_CTRL, false);
>> +       vmx_disable_intercept_for_msr(MSR_IA32_PRED_CMD, false);
>
> I have a lot of changes to MSR permission bitmap handling, but these
> intercepts should only be disabled when guest_cpuid_has(vcpu,
> X86_FEATURE_SPEC_CTRL).
>
>>          memcpy(vmx_msr_bitmap_legacy_x2apic_apicv,
>>                          vmx_msr_bitmap_legacy, PAGE_SIZE);
>> @@ -9597,6 +9620,9 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu)
>>
>>          pt_guest_enter(vmx);
>>
>> +       if (have_spec_ctrl && vmx->spec_ctrl != 0)
>> +               wrmsrl(MSR_IA32_SPEC_CTRL, vmx->spec_ctrl);
>> +
>
> Here, I have:
>
> /*
> * If the guest is allowed to write to MSR_IA32_SPEC_CTRL,
> * store it on VM-exit.
> */
> if (guest_cpuid_has(vcpu, X86_FEATURE_SPEC_CTRL))
> add_autostore_msr(vmx, MSR_IA32_SPEC_CTRL);
> else
> clear_autostore_msr(vmx, MSR_IA32_SPEC_CTRL);
>
> /*
> * If the guest's IA32_SPEC_CTRL MSR doesn't match the host's
> * IA32_SPEC_CTRL MSR, then add the MSR to the atomic switch
> * MSRs, so that the guest value will be loaded on VM-entry
> * and the host value will be loaded on VM-exit.
> */
> if (vmx->msr_ia32_spec_ctrl != spec_ctrl_enabled())
> add_atomic_switch_msr(vmx, MSR_IA32_SPEC_CTRL,
>        vmx->msr_ia32_spec_ctrl,
>        spec_ctrl_enabled());
> else
> clear_atomic_switch_msr(vmx, MSR_IA32_SPEC_CTRL);
>

I totally agree with this.

This exactly solves the issue I mentioned before of restoring the guest 
value of MSR_IA32_SPEC_CTRL using WRMSR before calling 
atomic_switch_perf_msrs() which does an indirect branch.

>>          atomic_switch_perf_msrs(vmx);
>>
>>          vmx_arm_hv_timer(vcpu);
>> @@ -9707,6 +9733,12 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu)
>>   #endif
>>                );
>>
>> +       if (have_spec_ctrl) {
>> +               rdmsrl(MSR_IA32_SPEC_CTRL, vmx->spec_ctrl);
>> +               if (vmx->spec_ctrl)
>> +                       wrmsrl(MSR_IA32_SPEC_CTRL, 0);
>> +       }
>> +
>
> I know the VM-exit MSR load and store lists are probably slower, but
> I'm a little uncomfortable restoring the host's IA32_SPEC_CTRL MSR
> late if the guest has it clear and the host has it set.

Again, I totally agree. This is a better approach for handling this.

>
>>          /* MSR_IA32_DEBUGCTLMSR is zeroed on vmexit. Restore it if needed */
>>          if (vmx->host_debugctlmsr)
>>                  update_debugctlmsr(vmx->host_debugctlmsr);
>> --
>> 1.8.3.1
>>
>>