Re: [PATCH 2/2] KVM: VMX: Fix VPID capability detection

From: Wanpeng Li
Date: Mon Oct 16 2017 - 22:20:22 EST


Hi Jim,
2017-10-17 3:53 GMT+08:00 Jim Mattson <jmattson@xxxxxxxxxx>:
> According to the SDM,
>
> The IA32_VMX_EPT_VPID_CAP MSR exists only on processors that support
> the 1-setting of the âactivate secondary
> controlsâ VM-execution control (only if bit 63 of the
> IA32_VMX_PROCBASED_CTLS MSR is 1) and that support
> either the 1-setting of the âenable EPTâ VM-execution control (only if
> bit 33 of the IA32_VMX_PROCBASED_CTLS2
> MSR is 1) or the 1-setting of the âenable VPIDâ VM-execution control
> (only if bit 37 of the
> IA32_VMX_PROCBASED_CTLS2 MSR is 1).
>
> Therefore, it seems dangerous to hoist the
> rdmsr(MSR_IA32_VMX_EPT_VPID_CAP, ...) outside of the conditional,
> unless you change it to rdmsr_safe.
>
> I don't think Intel has ever shipped a CPU with VPID but without EPT,
> so the motivation for this change is unclear to me.

My fault, I need to add more explanation to both patch 1/2, 2/2,
actually both patches are try to handle issues observed on L1 when EPT
is not exposed to L1.

For patch 1/2:
EPT switching is emulated for nested, however, I still can use vmxcap
tool to observe "EPTP Switching yes" even if EPT is not exposed to
L1.
For patch 2/2:
EPT is not exposed to L1, the VPID capability is exposed and can be
observed by vmxcap tool:
INVVPID supported yes
Individual-address INVVPID yes
Single-context INVVPID yes
All-context INVVPID yes
Single-context-retaining-globals INVVPID yes

However, the module parameter of VPID observed in L1 is always N, the
cpu_has_vmx_invvpid() check in L1 KVM fails since vmx_capability.vpid
is 0 and it is not read from MSR due to EPT is not exposed. This is
what patch 2/2 tries to fix.

Regards,
Wanpeng Li

>
> On Sat, Oct 14, 2017 at 7:31 PM, Wanpeng Li <kernellwp@xxxxxxxxx> wrote:
>> From: Wanpeng Li <wanpeng.li@xxxxxxxxxxx>
>>
>> According to the Intel SDM, volume 3, section 28.3.2: Creating and
>> Using Cached Translation Information, "The following items describe the
>> creation of mappings while EPT is not in use":
>> - Linear mappings may be created. They are derived from the paging
>> structures referenced (directly or indirectly) by the current value
>> of CR3 and are associated with the current VPID and the current PCID.
>>
>> The VPID is used to tag linear mappings when EPT is not enabled. However,
>> current logic just detects VPID capability if EPT is enabled, this patch
>> fixes it.
>>
>> Cc: Paolo Bonzini <pbonzini@xxxxxxxxxx>
>> Cc: Radim KrÄmÃÅ <rkrcmar@xxxxxxxxxx>
>> Signed-off-by: Wanpeng Li <wanpeng.li@xxxxxxxxxxx>
>> ---
>> arch/x86/kvm/vmx.c | 10 +++++++---
>> 1 file changed, 7 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
>> index 3644540..800d08c 100644
>> --- a/arch/x86/kvm/vmx.c
>> +++ b/arch/x86/kvm/vmx.c
>> @@ -3681,15 +3681,19 @@ static __init int setup_vmcs_config(struct vmcs_config *vmcs_conf)
>> SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE |
>> SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY);
>>
>> + rdmsr(MSR_IA32_VMX_EPT_VPID_CAP,
>> + vmx_capability.ept, vmx_capability.vpid);
>> +
>> if (_cpu_based_2nd_exec_control & SECONDARY_EXEC_ENABLE_EPT) {
>> /* CR3 accesses and invlpg don't need to cause VM Exits when EPT
>> enabled */
>> _cpu_based_exec_control &= ~(CPU_BASED_CR3_LOAD_EXITING |
>> CPU_BASED_CR3_STORE_EXITING |
>> CPU_BASED_INVLPG_EXITING);
>> - rdmsr(MSR_IA32_VMX_EPT_VPID_CAP,
>> - vmx_capability.ept, vmx_capability.vpid);
>> - }
>> + } else
>> + vmx_capability.ept = 0;
>> + if (!(_cpu_based_2nd_exec_control & SECONDARY_EXEC_ENABLE_VPID))
>> + vmx_capability.vpid = 0;
>>
>> min = VM_EXIT_SAVE_DEBUG_CONTROLS | VM_EXIT_ACK_INTR_ON_EXIT;
>> #ifdef CONFIG_X86_64
>> --
>> 2.7.4
>>