On Wed, May 5, 2021 at 2:38 AM Kai Huang <kai.huang@xxxxxxxxx> wrote:
>
> TDP MMU iterator's level is identical to page table's actual level. For
> instance, for the last level page table (whose entry points to one 4K
> page), iter->level is 1 (PG_LEVEL_4K), and in case of 5 level paging,
> the iter->level is mmu->shadow_root_level, which is 5. However, struct
> kvm_mmu_page's level currently is not set correctly when it is allocated
> in kvm_tdp_mmu_map(). When iterator hits non-present SPTE and needs to
> allocate a new child page table, currently iter->level, which is the
> level of the page table where the non-present SPTE belongs to, is used.
> This results in struct kvm_mmu_page's level always having its parent's
> level (excpet root table's level, which is initialized explicitly using
> mmu->shadow_root_level). This is kinda wrong, and not consistent with
> existing non TDP MMU code. Fortuantely the sp->role.level is only used
> in handle_removed_tdp_mmu_page(), which apparently is already aware of
> this, and handles correctly. However to make it consistent with non TDP
> MMU code (and fix the issue that both root page table and any child of
> it having shadow_root_level), fix this by using iter->level - 1 in
> kvm_tdp_mmu_map(). Also modify handle_removed_tdp_mmu_page() to handle
> such change.
Ugh. Thank you for catching this. This is going to take me a bit to
review as I should audit the code more broadly for this problem in the
TDP MMU.
It would probably also be a good idea to add a comment on the level
field to say that it represents the level of the SPTEs in the
associated page, not the level of the SPTE that links to the
associated page.
Hopefully that will prevent similar future misunderstandings.
>
> Signed-off-by: Kai Huang <kai.huang@xxxxxxxxx>
> ---
> arch/x86/kvm/mmu/tdp_mmu.c | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
> index debe8c3ec844..bcfb87e1c06e 100644
> --- a/arch/x86/kvm/mmu/tdp_mmu.c
> +++ b/arch/x86/kvm/mmu/tdp_mmu.c
> @@ -335,7 +335,7 @@ static void handle_removed_tdp_mmu_page(struct kvm *kvm, tdp_ptep_t pt,
>
> for (i = 0; i < PT64_ENT_PER_PAGE; i++) {
> sptep = rcu_dereference(pt) + i;
> - gfn = base_gfn + (i * KVM_PAGES_PER_HPAGE(level - 1));
> + gfn = base_gfn + i * KVM_PAGES_PER_HPAGE(level);
>
> if (shared) {
> /*
> @@ -377,12 +377,12 @@ static void handle_removed_tdp_mmu_page(struct kvm *kvm, tdp_ptep_t pt,
> WRITE_ONCE(*sptep, REMOVED_SPTE);
> }
> handle_changed_spte(kvm, kvm_mmu_page_as_id(sp), gfn,
> - old_child_spte, REMOVED_SPTE, level - 1,
> + old_child_spte, REMOVED_SPTE, level,
> shared);
> }
>
> kvm_flush_remote_tlbs_with_address(kvm, gfn,
> - KVM_PAGES_PER_HPAGE(level));
> + KVM_PAGES_PER_HPAGE(level + 1));
>
> call_rcu(&sp->rcu_head, tdp_mmu_free_sp_rcu_callback);
> }
> @@ -1009,7 +1009,7 @@ int kvm_tdp_mmu_map(struct kvm_vcpu *vcpu, gpa_t gpa, u32 error_code,
> }
>
> if (!is_shadow_present_pte(iter.old_spte)) {
> - sp = alloc_tdp_mmu_page(vcpu, iter.gfn, iter.level);
> + sp = alloc_tdp_mmu_page(vcpu, iter.gfn, iter.level - 1);
> child_pt = sp->spt;
>
> new_spte = make_nonleaf_spte(child_pt,
> --
> 2.31.1
>
