2024-05-11 03:46:58

by Hou Wenlong

Subject: [PATCH v2] KVM: x86/mmu: Only allocate shadowed translation cache for sp->role.level <= KVM_MAX_HUGEPAGE_LEVEL

Only the indirect SP with sp->role.level <= KVM_MAX_HUGEPAGE_LEVEL might
have leaf gptes, so allocation of shadowed translation cache is needed
only for it. Then, it can use sp->shadowed_translation to determine
whether to use the information in the shadowed translation cache or not.
Also, extend the WARN in FNAME(sync_spte)() to ensure that this won't
break shadow_mmu_get_sp_for_split().

Suggested-by: Lai Jiangshan <[email protected]>
Signed-off-by: Hou Wenlong <[email protected]>
---
v1->v2:
- Remove the order change in kvm_mmu_page_get_gfn() to keep smallest
diff.
- Drop the helper.
- Extend the WARN in FNAME(sync_spte).

arch/x86/kvm/mmu/mmu.c | 11 +++++------
arch/x86/kvm/mmu/paging_tmpl.h | 2 +-
2 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index fc3b59b59ee1..dc6f6a272e98 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -719,7 +719,7 @@ static gfn_t kvm_mmu_page_get_gfn(struct kvm_mmu_page *sp, int index)
if (sp->role.passthrough)
return sp->gfn;

- if (!sp->role.direct)
+ if (sp->shadowed_translation)
return sp->shadowed_translation[index] >> PAGE_SHIFT;

return sp->gfn + (index << ((sp->role.level - 1) * SPTE_LEVEL_BITS));
@@ -733,7 +733,7 @@ static gfn_t kvm_mmu_page_get_gfn(struct kvm_mmu_page *sp, int index)
*/
static u32 kvm_mmu_page_get_access(struct kvm_mmu_page *sp, int index)
{
- if (sp_has_gptes(sp))
+ if (sp->shadowed_translation)
return sp->shadowed_translation[index] & ACC_ALL;

/*
@@ -754,7 +754,7 @@ static u32 kvm_mmu_page_get_access(struct kvm_mmu_page *sp, int index)
static void kvm_mmu_page_set_translation(struct kvm_mmu_page *sp, int index,
gfn_t gfn, unsigned int access)
{
- if (sp_has_gptes(sp)) {
+ if (sp->shadowed_translation) {
sp->shadowed_translation[index] = (gfn << PAGE_SHIFT) | access;
return;
}
@@ -1697,8 +1697,7 @@ static void kvm_mmu_free_shadow_page(struct kvm_mmu_page *sp)
hlist_del(&sp->hash_link);
list_del(&sp->link);
free_page((unsigned long)sp->spt);
- if (!sp->role.direct)
- free_page((unsigned long)sp->shadowed_translation);
+ free_page((unsigned long)sp->shadowed_translation);
kmem_cache_free(mmu_page_header_cache, sp);
}

@@ -2199,7 +2198,7 @@ static struct kvm_mmu_page *kvm_mmu_alloc_shadow_page(struct kvm *kvm,

sp = kvm_mmu_memory_cache_alloc(caches->page_header_cache);
sp->spt = kvm_mmu_memory_cache_alloc(caches->shadow_page_cache);
- if (!role.direct)
+ if (!role.direct && role.level <= KVM_MAX_HUGEPAGE_LEVEL)
sp->shadowed_translation = kvm_mmu_memory_cache_alloc(caches->shadowed_info_cache);

set_page_private(virt_to_page(sp->spt), (unsigned long)sp);
diff --git a/arch/x86/kvm/mmu/paging_tmpl.h b/arch/x86/kvm/mmu/paging_tmpl.h
index 7a87097cb45b..89b5d73e9e3c 100644
--- a/arch/x86/kvm/mmu/paging_tmpl.h
+++ b/arch/x86/kvm/mmu/paging_tmpl.h
@@ -911,7 +911,7 @@ static int FNAME(sync_spte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp, int
gpa_t pte_gpa;
gfn_t gfn;

- if (WARN_ON_ONCE(!sp->spt[i]))
+ if (WARN_ON_ONCE(!sp->spt[i] || !sp->shadowed_translation))
return 0;

first_pte_gpa = FNAME(get_level1_sp_gpa)(sp);
--
2.31.1

2024-06-04 23:45:42

by Sean Christopherson

Subject: Re: [PATCH v2] KVM: x86/mmu: Only allocate shadowed translation cache for sp->role.level <= KVM_MAX_HUGEPAGE_LEVEL

On Sat, 11 May 2024 11:46:37 +0800, Hou Wenlong wrote:
> Only the indirect SP with sp->role.level <= KVM_MAX_HUGEPAGE_LEVEL might
> have leaf gptes, so allocation of shadowed translation cache is needed
> only for it. Then, it can use sp->shadowed_translation to determine
> whether to use the information in the shadowed translation cache or not.
> Also, extend the WARN in FNAME(sync_spte)() to ensure that this won't
> break shadow_mmu_get_sp_for_split().
>
> [...]

Applied to kvm-x86 mmu, thanks!

[1/1] KVM: x86/mmu: Only allocate shadowed translation cache for sp->role.level <= KVM_MAX_HUGEPAGE_LEVEL
https://github.com/kvm-x86/linux/commit/9ecc1c119b28

--
https://github.com/kvm-x86/linux/tree/next