Re: [PATCH 2/2] KVM: x86/mmu: Don't step down in the TDP iterator when zapping all SPTEs

[Ben Gardon]

On Wed, Aug 11, 2021 at 10:07 PM Sean Christopherson <seanjc@xxxxxxxxxx> wrote:
>
> Set the min_level for the TDP iterator at the root level when zapping all
> SPTEs so that the _iterator_ only processes top-level SPTEs.  Zapping a
> non-leaf SPTE will recursively zap all its children, thus there is no
> need for the iterator to attempt to step down.  This avoids rereading all
> the top-level SPTEs after they are zapped by causing try_step_down() to
> short-circuit.
>
> Cc: Ben Gardon <bgardon@xxxxxxxxxx>
> Signed-off-by: Sean Christopherson <seanjc@xxxxxxxxxx>

This change looks functionally correct, but I'm not sure it's worth
adding more code special cased on zap-all for what seems like a small
performance improvement in a context which shouldn't be particularly
performance sensitive.
Change is a correct optimization though and it's not much extra code,
so I'm happy to give a:
Reviewed-by: Ben Gardon <bgardon@xxxxxxxxxx>

> ---
>  arch/x86/kvm/mmu/tdp_mmu.c | 13 ++++++++++++-
>  1 file changed, 12 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
> index 6566f70a31c1..aec069c18d82 100644
> --- a/arch/x86/kvm/mmu/tdp_mmu.c
> +++ b/arch/x86/kvm/mmu/tdp_mmu.c
> @@ -751,6 +751,16 @@ static bool zap_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root,
>  {
>         bool zap_all = (end == ZAP_ALL_END);
>         struct tdp_iter iter;
> +       int min_level;
> +
> +       /*
> +        * No need to step down in the iterator when zapping all SPTEs, zapping
> +        * the top-level non-leaf SPTEs will recurse on all their children.
> +        */
> +       if (zap_all)
> +               min_level = root->role.level;
> +       else
> +               min_level = PG_LEVEL_4K;
>
>         /*
>          * Bound the walk at host.MAXPHYADDR, guest accesses beyond that will
> @@ -763,7 +773,8 @@ static bool zap_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root,
>
>         rcu_read_lock();
>
> -       tdp_root_for_each_pte(iter, root, start, end) {
> +       for_each_tdp_pte_min_level(iter, root->spt, root->role.level,
> +                                  min_level, start, end) {
>  retry:
>                 if (can_yield &&
>                     tdp_mmu_iter_cond_resched(kvm, &iter, flush, shared)) {
> --
> 2.33.0.rc1.237.g0d66db33f3-goog
>