On 11.08.2017 08:13, Paolo Bonzini wrote:
> On 11/08/2017 07:59, Wanpeng Li wrote:
>> From: Wanpeng Li <wanpeng.li@xxxxxxxxxxx>
>>
>> Commit c016004494b0 (KVM: x86: Avoid guest page table walk when gpa_available
>> is set) avoids the page table walk when cr2 has already contained a valid GPA.
>> However, that is not the truth if ept == 0 and shadow page table is used. In
>> this scenario cr2 can just contains a valid GVA. The commit results in guest
>> stuck during boot due to read/write emulation against GVA instead of GPA of
>> the guest.
>>
>> This patch fixes it by not setting the gpa_available flag if the shadow page table
>> is used.
>
> Yup, I got these on my overnight tests too. A better fix that I was
> going to commit today is:
>
> if (vcpu->arch.mmu.direct_map) {
> vcpu->arch.gpa_available = true;
> vcpu->arch.gpa_val = fault_address;
> }
>
Makes sense to me. Can we add a comment?
/* with shadow page tables, cr2 could contain a GVA, not a GPA */
> because AMD gets here even if npt=1.
>
> Paolo
--
Thanks,
David
