From: Lai Jiangshan <[email protected]>
When local shadow pages are activated, link_shadow_page() might link
a local shadow pages which is the PAE root for PAE paging with its
children.
Add make_pae_pdpte() to handle it.
The code is not activated since local shadow pages are not activated
yet.
Signed-off-by: Lai Jiangshan <[email protected]>
---
arch/x86/kvm/mmu/mmu.c | 6 +++++-
arch/x86/kvm/mmu/spte.c | 7 +++++++
arch/x86/kvm/mmu/spte.h | 1 +
3 files changed, 13 insertions(+), 1 deletion(-)
diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index c941a5931bc3..e1a059dd9621 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -2340,7 +2340,11 @@ static void link_shadow_page(struct kvm_vcpu *vcpu, u64 *sptep,
BUILD_BUG_ON(VMX_EPT_WRITABLE_MASK != PT_WRITABLE_MASK);
- spte = make_nonleaf_spte(sp->spt, sp_ad_disabled(sp));
+ if (unlikely(sp->role.level == PT32_ROOT_LEVEL &&
+ vcpu->arch.mmu->root_role.level == PT32E_ROOT_LEVEL))
+ spte = make_pae_pdpte(sp->spt);
+ else
+ spte = make_nonleaf_spte(sp->spt, sp_ad_disabled(sp));
mmu_spte_set(sptep, spte);
diff --git a/arch/x86/kvm/mmu/spte.c b/arch/x86/kvm/mmu/spte.c
index b5960bbde7f7..5c31fa1d2b61 100644
--- a/arch/x86/kvm/mmu/spte.c
+++ b/arch/x86/kvm/mmu/spte.c
@@ -279,6 +279,13 @@ u64 make_huge_page_split_spte(u64 huge_spte, int huge_level, int index)
return child_spte;
}
+u64 make_pae_pdpte(u64 *child_pt)
+{
+ /* The only ignore bits in PDPTE are 11:9. */
+ BUILD_BUG_ON(!(GENMASK(11,9) & SPTE_MMU_PRESENT_MASK));
+ return __pa(child_pt) | PT_PRESENT_MASK | SPTE_MMU_PRESENT_MASK |
+ shadow_me_value;
+}
u64 make_nonleaf_spte(u64 *child_pt, bool ad_disabled)
{
diff --git a/arch/x86/kvm/mmu/spte.h b/arch/x86/kvm/mmu/spte.h
index 0127bb6e3c7d..2408ba1361d5 100644
--- a/arch/x86/kvm/mmu/spte.h
+++ b/arch/x86/kvm/mmu/spte.h
@@ -426,6 +426,7 @@ bool make_spte(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
u64 old_spte, bool prefetch, bool can_unsync,
bool host_writable, u64 *new_spte);
u64 make_huge_page_split_spte(u64 huge_spte, int huge_level, int index);
+u64 make_pae_pdpte(u64 *child_pt);
u64 make_nonleaf_spte(u64 *child_pt, bool ad_disabled);
u64 make_mmio_spte(struct kvm_vcpu *vcpu, u64 gfn, unsigned int access);
u64 mark_spte_for_access_track(u64 spte);
--
2.19.1.6.gb485710b
The shortlog is very misleading. This patch doesn't "Link PAE root pagetable with
its children", it adds support for creating PAE PDPTEs in order to link them into
shadow pages, but it doesn't do the actual linking.
KVM: x86/mmu: Add support for linking PAE PDPTE shadow pages
On Sat, May 21, 2022, Lai Jiangshan wrote:
> From: Lai Jiangshan <[email protected]>
>
> When local shadow pages are activated, link_shadow_page() might link
> a local shadow pages which is the PAE root for PAE paging with its
> children.
>
> Add make_pae_pdpte() to handle it.
>
> The code is not activated since local shadow pages are not activated
> yet.
And though it's redudnant with other context, IMO it's helpful to again reiterate
why this will be used for per-vCPU (local) shadow pages, i.e. why it's _not_ used
right now.
Add support for installing PDPTEs via link_shadow_page(), PDPTEs have
different layouts than every other entry type and so need a dedicated
helper to make them.
This code will become active when a future patch activates per-vCPU
shadow pages and stops using so called "special" roots (which are
installed at root allocation, not via link_shadow_page()).
> Signed-off-by: Lai Jiangshan <[email protected]>
> ---
> arch/x86/kvm/mmu/mmu.c | 6 +++++-
> arch/x86/kvm/mmu/spte.c | 7 +++++++
> arch/x86/kvm/mmu/spte.h | 1 +
> 3 files changed, 13 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> index c941a5931bc3..e1a059dd9621 100644
> --- a/arch/x86/kvm/mmu/mmu.c
> +++ b/arch/x86/kvm/mmu/mmu.c
> @@ -2340,7 +2340,11 @@ static void link_shadow_page(struct kvm_vcpu *vcpu, u64 *sptep,
>
> BUILD_BUG_ON(VMX_EPT_WRITABLE_MASK != PT_WRITABLE_MASK);
>
> - spte = make_nonleaf_spte(sp->spt, sp_ad_disabled(sp));
> + if (unlikely(sp->role.level == PT32_ROOT_LEVEL &&
> + vcpu->arch.mmu->root_role.level == PT32E_ROOT_LEVEL))
> + spte = make_pae_pdpte(sp->spt);
> + else
> + spte = make_nonleaf_spte(sp->spt, sp_ad_disabled(sp));
>
> mmu_spte_set(sptep, spte);
>
> diff --git a/arch/x86/kvm/mmu/spte.c b/arch/x86/kvm/mmu/spte.c
> index b5960bbde7f7..5c31fa1d2b61 100644
> --- a/arch/x86/kvm/mmu/spte.c
> +++ b/arch/x86/kvm/mmu/spte.c
> @@ -279,6 +279,13 @@ u64 make_huge_page_split_spte(u64 huge_spte, int huge_level, int index)
> return child_spte;
> }
>
> +u64 make_pae_pdpte(u64 *child_pt)
> +{
> + /* The only ignore bits in PDPTE are 11:9. */
s/ignore/ignored, though it might be worth calling out that unlike 64-bit paging,
the upper bits bits are reserved (I always forget this).
/*
* Only PDPTE bits 11:9 are ignored by hardware. Unlike 64-bit paging,
* bits above the PA bits are reserved.
*/
> + BUILD_BUG_ON(!(GENMASK(11,9) & SPTE_MMU_PRESENT_MASK));
> + return __pa(child_pt) | PT_PRESENT_MASK | SPTE_MMU_PRESENT_MASK |
> + shadow_me_value;
> +}
>
> u64 make_nonleaf_spte(u64 *child_pt, bool ad_disabled)
> {
> diff --git a/arch/x86/kvm/mmu/spte.h b/arch/x86/kvm/mmu/spte.h
> index 0127bb6e3c7d..2408ba1361d5 100644
> --- a/arch/x86/kvm/mmu/spte.h
> +++ b/arch/x86/kvm/mmu/spte.h
> @@ -426,6 +426,7 @@ bool make_spte(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp,
> u64 old_spte, bool prefetch, bool can_unsync,
> bool host_writable, u64 *new_spte);
> u64 make_huge_page_split_spte(u64 huge_spte, int huge_level, int index);
> +u64 make_pae_pdpte(u64 *child_pt);
> u64 make_nonleaf_spte(u64 *child_pt, bool ad_disabled);
> u64 make_mmio_spte(struct kvm_vcpu *vcpu, u64 gfn, unsigned int access);
> u64 mark_spte_for_access_track(u64 spte);
> --
> 2.19.1.6.gb485710b
>