KVM: MMU: large page support

Create large pages mappings if the guest PTE's are marked as such and
the underlying memory is hugetlbfs backed.  If the largepage contains
write-protected pages, a large pte is not used.

Gives a consistent 2% improvement for data copies on ram mounted
filesystem, without NPT/EPT.

Anthony measures a 4% improvement on 4-way kernbench, with NPT.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>

1 files changed, 26 insertions, 6 deletions

diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h
index 4b55f462e2b..17f9d160ca3 100644
--- a/arch/x86/kvm/paging_tmpl.h
+++ b/arch/x86/kvm/paging_tmpl.h
@@ -248,6 +248,7 @@ static void FNAME(update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *page,
 	pt_element_t gpte;
 	unsigned pte_access;
 	struct page *npage;
+	int largepage = vcpu->arch.update_pte.largepage;
 
 	gpte = *(const pt_element_t *)pte;
 	if (~gpte & (PT_PRESENT_MASK | PT_ACCESSED_MASK)) {
@@ -264,7 +265,8 @@ static void FNAME(update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *page,
 		return;
 	get_page(npage);
 	mmu_set_spte(vcpu, spte, page->role.access, pte_access, 0, 0,
-		     gpte & PT_DIRTY_MASK, NULL, gpte_to_gfn(gpte), npage);
+		     gpte & PT_DIRTY_MASK, NULL, largepage, gpte_to_gfn(gpte),
+		     npage);
 }
 
 /*
@@ -272,8 +274,8 @@ static void FNAME(update_pte)(struct kvm_vcpu *vcpu, struct kvm_mmu_page *page,
  */
 static u64 *FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr,
 			 struct guest_walker *walker,
-			 int user_fault, int write_fault, int *ptwrite,
-			 struct page *page)
+			 int user_fault, int write_fault, int largepage,
+			 int *ptwrite, struct page *page)
 {
 	hpa_t shadow_addr;
 	int level;
@@ -301,11 +303,19 @@ static u64 *FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr,
 		shadow_ent = ((u64 *)__va(shadow_addr)) + index;
 		if (level == PT_PAGE_TABLE_LEVEL)
 			break;
-		if (is_shadow_present_pte(*shadow_ent)) {
+
+		if (largepage && level == PT_DIRECTORY_LEVEL)
+			break;
+
+		if (is_shadow_present_pte(*shadow_ent)
+		    && !is_large_pte(*shadow_ent)) {
 			shadow_addr = *shadow_ent & PT64_BASE_ADDR_MASK;
 			continue;
 		}
 
+		if (is_large_pte(*shadow_ent))
+			rmap_remove(vcpu->kvm, shadow_ent);
+
 		if (level - 1 == PT_PAGE_TABLE_LEVEL
 		    && walker->level == PT_DIRECTORY_LEVEL) {
 			metaphysical = 1;
@@ -339,7 +349,7 @@ static u64 *FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr,
 	mmu_set_spte(vcpu, shadow_ent, access, walker->pte_access & access,
 		     user_fault, write_fault,
 		     walker->ptes[walker->level-1] & PT_DIRTY_MASK,
-		     ptwrite, walker->gfn, page);
+		     ptwrite, largepage, walker->gfn, page);
 
 	return shadow_ent;
 }
@@ -369,6 +379,7 @@ static int FNAME(page_fault)(struct kvm_vcpu *vcpu, gva_t addr,
 	int write_pt = 0;
 	int r;
 	struct page *page;
+	int largepage = 0;
 
 	pgprintk("%s: addr %lx err %x\n", __FUNCTION__, addr, error_code);
 	kvm_mmu_audit(vcpu, "pre page fault");
@@ -396,6 +407,14 @@ static int FNAME(page_fault)(struct kvm_vcpu *vcpu, gva_t addr,
 	}
 
 	down_read(&current->mm->mmap_sem);
+	if (walker.level == PT_DIRECTORY_LEVEL) {
+		gfn_t large_gfn;
+		large_gfn = walker.gfn & ~(KVM_PAGES_PER_HPAGE-1);
+		if (is_largepage_backed(vcpu, large_gfn)) {
+			walker.gfn = large_gfn;
+			largepage = 1;
+		}
+	}
 	page = gfn_to_page(vcpu->kvm, walker.gfn);
 	up_read(&current->mm->mmap_sem);
 
@@ -410,7 +429,8 @@ static int FNAME(page_fault)(struct kvm_vcpu *vcpu, gva_t addr,
 	spin_lock(&vcpu->kvm->mmu_lock);
 	kvm_mmu_free_some_pages(vcpu);
 	shadow_pte = FNAME(fetch)(vcpu, addr, &walker, user_fault, write_fault,
-				  &write_pt, page);
+				  largepage, &write_pt, page);
+
 	pgprintk("%s: shadow pte %p %llx ptwrite %d\n", __FUNCTION__,
 		 shadow_pte, *shadow_pte, write_pt);