Re: [PATCH 02/22] KVM: x86: move pdptrs out of the MMU

[Paolo Bonzini]

On 5/11/26 17:06, Paolo Bonzini wrote:
> PDPTRs are part of the CPU state.  A bit unconventionally, they are
> reached via vcpu->arch.walk_mmu instead of being stored in vcpu->arch
> directly.  That is nice in principle---it would allow TDP shadow paging
> to have its own PDPTRs---but it is not necessary, because EPT has no
> PDPTRs and NPT does not cache them.
>
> Since kvm_pdptr_read does not otherwise need the MMU, drop the pdptrs
> from the MMU altogether.  There is however a negative effect, in that
> they are now not stored separately in root_mmu and nested_mmu for L1
> and L2 guests.  This means that they are overwritten by nested VM entry
> and exit, and need to be manually marked dirty.
>
> Note that page table PDPTRs are not affected, since they are stored
> in pae_root.

The vmx/nested.c and svm/nested.c changes here are untested and 
completely broken - I mentioned this in the cover letter but should have 
done it here too in case anyone reviews this.  But there are some 
cleanup opportunities now that I looked at it more closely.

Paolo