On 03/12/19 10:21, Jack Wang wrote:
> Paolo Bonzini <pbonzini@xxxxxxxxxx> 于2019年12月2日周一 下午4:09写道：
>>
>> On 02/12/19 15:51, Greg Kroah-Hartman wrote:
>>> On Mon, Dec 02, 2019 at 03:40:04PM +0100, Jack Wang wrote:
>>>> Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> 于2019年11月27日周三 下午10:30写道：
>>>>>
>>>>> From: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
>>>>>
>>>>> [ Upstream commit 7671ce21b13b9596163a29f4712cb2451a9b97dc ]
>>>>>
>>>>> In preparation of supporting checkpoint/restore for nested state,
>>>>> commit ca0bde28f2ed ("kvm: nVMX: Split VMCS checks from nested_vmx_run()")
>>>>> modified check_vmentry_postreqs() to only perform the guest EFER
>>>>> consistency checks when nested_run_pending is true. But, in the
>>>>> normal nested VMEntry flow, nested_run_pending is only set after
>>>>> check_vmentry_postreqs(), i.e. the consistency check is being skipped.
>>>>>
>>>>> Alternatively, nested_run_pending could be set prior to calling
>>>>> check_vmentry_postreqs() in nested_vmx_run(), but placing the
>>>>> consistency checks in nested_vmx_enter_non_root_mode() allows us
>>>>> to split prepare_vmcs02() and interleave the preparation with
>>>>> the consistency checks without having to change the call sites
>>>>> of nested_vmx_enter_non_root_mode(). In other words, the rest
>>>>> of the consistency check code in nested_vmx_run() will be joining
>>>>> the postreqs checks in future patches.
>>>>>
>>>>> Fixes: ca0bde28f2ed ("kvm: nVMX: Split VMCS checks from nested_vmx_run()")
>>>>> Signed-off-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
>>>>> Cc: Jim Mattson <jmattson@xxxxxxxxxx>
>>>>> Reviewed-by: Jim Mattson <jmattson@xxxxxxxxxx>
>>>>> Signed-off-by: Paolo Bonzini <pbonzini@xxxxxxxxxx>
>>>>> Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
>>>>> ---
>>>>> arch/x86/kvm/vmx.c | 10 +++-------
>>>>> 1 file changed, 3 insertions(+), 7 deletions(-)
>>>>>
>>>>> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
>>>>> index fe7fdd666f091..bdf019f322117 100644
>>>>> --- a/arch/x86/kvm/vmx.c
>>>>> +++ b/arch/x86/kvm/vmx.c
>>>>> @@ -12694,6 +12694,9 @@ static int enter_vmx_non_root_mode(struct kvm_vcpu *vcpu, u32 *exit_qual)
>>>>> if (likely(!evaluate_pending_interrupts) && kvm_vcpu_apicv_active(vcpu))
>>>>> evaluate_pending_interrupts |= vmx_has_apicv_interrupt(vcpu);
>>>>>
>>>>> + if (from_vmentry && check_vmentry_postreqs(vcpu, vmcs12, exit_qual))
>>>>> + return EXIT_REASON_INVALID_STATE;
>>>>> +
>>>>> enter_guest_mode(vcpu);
>>>>>
>>>>> if (!(vmcs12->vm_entry_controls & VM_ENTRY_LOAD_DEBUG_CONTROLS))
>>>>> @@ -12836,13 +12839,6 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
>>>>> */
>>>>> skip_emulated_instruction(vcpu);
>>>>>
>>>>> - ret = check_vmentry_postreqs(vcpu, vmcs12, &exit_qual);
>>>>> - if (ret) {
>>>>> - nested_vmx_entry_failure(vcpu, vmcs12,
>>>>> - EXIT_REASON_INVALID_STATE, exit_qual);
>>>>> - return 1;
>>>>> - }
>>>>> -
>>>>> /*
>>>>> * We're finally done with prerequisite checking, and can start with
>>>>> * the nested entry.
>>>>> --
>>>>> 2.20.1
>>>>>
>>>>>
>>>>>
>>>> Hi all,
>>>>
>>>> This commit caused many kvm-unit-tests regression, cherry-pick
>>>> following commits from 4.20 fix the regression:
>>>> d63907dc7dd1 ("KVM: nVMX: rename enter_vmx_non_root_mode to
>>>> nested_vmx_enter_non_root_mode")
>>>> a633e41e7362 ("KVM: nVMX: assimilate nested_vmx_entry_failure() into
>>>> nested_vmx_enter_non_root_mode()")
>>>
>>> Now queued up, thanks!
>>>
>>> greg k-h
>>>
>>
>> Why was it backported anyway? Can everybody please just stop applying
>> KVM patches to stable kernels unless CCed to stable@xxxxxxxxxxxxxxx?
>>
>> I thought I had already asked Sasha to opt out of the autoselect
>> nonsense after catching another bug that would have been introduced.
>>
>> Paolo
>>
> Hi Paolo,
>
> Should we simply revert the patch, maybe also
> 9fe573d539a8 ("KVM: nVMX: reset cache/shadows when switching loaded VMCS")
>
> Both of them are from one big patchset:
> https://patchwork.kernel.org/cover/10616179/
>
> Revert both patches recover the regression I see on kvm-unit-tests.
Greg already included the patches that the bot missed, so it's okay.
Paolo
