Re: [PATCH 24/28] KVM: x86/mmu: Add dedicated helper to zap TDP MMU root shadow page

[Ben Gardon]

On Fri, Nov 19, 2021 at 8:51 PM Sean Christopherson <seanjc@xxxxxxxxxx> wrote:
>
> Convert tdp_mmu_zap_root() into its own dedicated flow instead of simply
> redirecting into zap_gfn_range().  In addition to hardening zapping of
> roots, this will allow future simplification of zap_gfn_range() by having
> it zap only leaf SPTEs, and by removing its tricky "zap all" heuristic.
> By having all paths that truly need to free _all_ SPs flow through the
> dedicated root zapper, the generic zapper can be freed of those concerns.
>
> Signed-off-by: Sean Christopherson <seanjc@xxxxxxxxxx>
> ---
>  arch/x86/kvm/mmu/tdp_mmu.c | 91 +++++++++++++++++++++++++++-----------
>  1 file changed, 66 insertions(+), 25 deletions(-)
>
> diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
> index 99ea19e763da..0e5a0d40e54a 100644
> --- a/arch/x86/kvm/mmu/tdp_mmu.c
> +++ b/arch/x86/kvm/mmu/tdp_mmu.c
> @@ -53,10 +53,6 @@ void kvm_mmu_uninit_tdp_mmu(struct kvm *kvm)
>         rcu_barrier();
>  }
>
> -static bool zap_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root,
> -                         gfn_t start, gfn_t end, bool can_yield, bool flush,
> -                         bool shared);
> -
>  static void tdp_mmu_free_sp(struct kvm_mmu_page *sp)
>  {
>         free_page((unsigned long)sp->spt);
> @@ -79,11 +75,8 @@ static void tdp_mmu_free_sp_rcu_callback(struct rcu_head *head)
>         tdp_mmu_free_sp(sp);
>  }
>
> -static bool tdp_mmu_zap_root(struct kvm *kvm, struct kvm_mmu_page *root,
> -                            bool shared)
> -{
> -       return zap_gfn_range(kvm, root, 0, -1ull, true, false, shared);
> -}
> +static void tdp_mmu_zap_root(struct kvm *kvm, struct kvm_mmu_page *root,
> +                            bool shared, bool root_is_unreachable);
>
>  /*
>   * Note, putting a root might sleep, i.e. the caller must have IRQs enabled and
> @@ -120,13 +113,8 @@ void kvm_tdp_mmu_put_root(struct kvm *kvm, struct kvm_mmu_page *root,
>          * invalidates any paging-structure-cache entries, i.e. TLB entries for
>          * intermediate paging structures, that may be zapped, as such entries
>          * are associated with the ASID on both VMX and SVM.
> -        *
> -        * WARN if a flush is reported for an invalid root, as its child SPTEs
> -        * should have been zapped by kvm_tdp_mmu_zap_invalidated_roots(), and
> -        * inserting new SPTEs under an invalid root is a KVM bug.
>          */
> -       if (tdp_mmu_zap_root(kvm, root, shared))
> -               WARN_ON_ONCE(root->role.invalid);
> +       tdp_mmu_zap_root(kvm, root, shared, true);
>
>         call_rcu(&root->rcu_head, tdp_mmu_free_sp_rcu_callback);
>  }
> @@ -766,6 +754,65 @@ static inline bool tdp_mmu_iter_cond_resched(struct kvm *kvm,
>         return false;
>  }
>
> +static inline gfn_t tdp_mmu_max_gfn_host(void)
> +{
> +       /*
> +        * Bound TDP MMU walks at host.MAXPHYADDR, guest accesses beyond that
> +        * will hit a #PF(RSVD) and never hit an EPT Violation/Misconfig / #NPF,
> +        * and so KVM will never install a SPTE for such addresses.
> +        */
> +       return 1ULL << (shadow_phys_bits - PAGE_SHIFT);
> +}
> +
> +static void tdp_mmu_zap_root(struct kvm *kvm, struct kvm_mmu_page *root,
> +                            bool shared, bool root_is_unreachable)
> +{
> +       struct tdp_iter iter;
> +
> +       gfn_t end = tdp_mmu_max_gfn_host();
> +       gfn_t start = 0;
> +
> +       kvm_lockdep_assert_mmu_lock_held(kvm, shared);
> +
> +       rcu_read_lock();
> +
> +       /*
> +        * No need to try to step down in the iterator when zapping an entire
> +        * root, zapping an upper-level SPTE will recurse on its children.
> +        */
> +       for_each_tdp_pte_min_level(iter, root->spt, root->role.level,
> +                                  root->role.level, start, end) {
> +retry:
> +               if (tdp_mmu_iter_cond_resched(kvm, &iter, false, shared))
> +                       continue;
> +
> +               if (!is_shadow_present_pte(iter.old_spte))
> +                       continue;
> +
> +               if (!shared) {
> +                       tdp_mmu_set_spte(kvm, &iter, 0);
> +               } else if (!tdp_mmu_set_spte_atomic(kvm, &iter, 0)) {

Worth adding a comment about why this is used instead of
tdp_mmu_zap_spte_atomic.

> +                       /*
> +                        * cmpxchg() shouldn't fail if the root is unreachable.
> +                        * to be unreachable.  Re-read the SPTE and retry so as

Repeated phrase.


> +                        * not to leak the page and its children.
> +                        */
> +                       WARN_ONCE(root_is_unreachable,
> +                                 "Contended TDP MMU SPTE in unreachable root.");
> +                       iter.old_spte = kvm_tdp_mmu_read_spte(iter.sptep);

Note this will conflict with the series David sent out Friday.
Hopefully some of the cleanups early in that series get merged, in
which case this line will not be needed.

> +                       goto retry;
> +               }
> +               /*
> +                * WARN if the root is invalid and is unreachable, all SPTEs
> +                * should've been zapped by kvm_tdp_mmu_zap_invalidated_roots(),
> +                * and inserting new SPTEs under an invalid root is a KVM bug.
> +                */
> +               WARN_ON_ONCE(root_is_unreachable && root->role.invalid);
> +       }
> +
> +       rcu_read_unlock();
> +}
> +
>  bool kvm_tdp_mmu_zap_sp(struct kvm *kvm, struct kvm_mmu_page *sp)
>  {
>         u64 old_spte;
> @@ -807,8 +854,7 @@ static bool zap_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root,
>                           gfn_t start, gfn_t end, bool can_yield, bool flush,
>                           bool shared)
>  {
> -       gfn_t max_gfn_host = 1ULL << (shadow_phys_bits - PAGE_SHIFT);
> -       bool zap_all = (start == 0 && end >= max_gfn_host);
> +       bool zap_all = (start == 0 && end >= tdp_mmu_max_gfn_host());
>         struct tdp_iter iter;
>
>         /*
> @@ -817,12 +863,7 @@ static bool zap_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root,
>          */
>         int min_level = zap_all ? root->role.level : PG_LEVEL_4K;
>
> -       /*
> -        * Bound the walk at host.MAXPHYADDR, guest accesses beyond that will
> -        * hit a #PF(RSVD) and never get to an EPT Violation/Misconfig / #NPF,
> -        * and so KVM will never install a SPTE for such addresses.
> -        */
> -       end = min(end, max_gfn_host);
> +       end = min(end, tdp_mmu_max_gfn_host());

tdp_mmu_max_gfn_host and this refactor, could be added in a separate
commit if desired.

>
>         kvm_lockdep_assert_mmu_lock_held(kvm, shared);
>
> @@ -898,7 +939,7 @@ void kvm_tdp_mmu_zap_all(struct kvm *kvm)
>          */
>         for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
>                 for_each_tdp_mmu_root_yield_safe(kvm, root, i, false)
> -                       (void)tdp_mmu_zap_root(kvm, root, false);
> +                       tdp_mmu_zap_root(kvm, root, false, true);
>         }
>  }
>
> @@ -934,7 +975,7 @@ void kvm_tdp_mmu_zap_invalidated_roots(struct kvm *kvm,
>                  * will still flush on yield, but that's a minor performance
>                  * blip and not a functional issue.
>                  */
> -               (void)tdp_mmu_zap_root(kvm, root, true);
> +               tdp_mmu_zap_root(kvm, root, true, false);
>                 kvm_tdp_mmu_put_root(kvm, root, true);
>         }
>  }
> --
> 2.34.0.rc2.393.gf8c9666880-goog
>