[PATCH] KVM: MMU: Never free a shadow page actively serving as a root
authorAvi Kivity <avi@qumranet.com>
Sat, 6 Jan 2007 00:36:51 +0000 (16:36 -0800)
committerLinus Torvalds <torvalds@woody.osdl.org>
Sat, 6 Jan 2007 07:55:26 +0000 (23:55 -0800)
We always need cr3 to point to something valid, so if we detect that we're
freeing a root page, simply push it back to the top of the active list.

Signed-off-by: Avi Kivity <avi@qumranet.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
drivers/kvm/kvm.h
drivers/kvm/mmu.c

index 201b2735ca9194919b135614acb8b25184578fd9..b24a86e1f434cb267416c0c63608393a1d145378 100644 (file)
@@ -134,6 +134,7 @@ struct kvm_mmu_page {
                                    */
        int global;              /* Set if all ptes in this page are global */
        int multimapped;         /* More than one parent_pte? */
+       int root_count;          /* Currently serving as active root */
        union {
                u64 *parent_pte;               /* !multimapped */
                struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
index 0e44aca9eee7e422f32a34b331c917b83741d198..f16321498093391bd681028d73555f3218b87ad7 100644 (file)
@@ -550,8 +550,13 @@ static void kvm_mmu_zap_page(struct kvm_vcpu *vcpu,
                *parent_pte = 0;
        }
        kvm_mmu_page_unlink_children(vcpu, page);
-       hlist_del(&page->hash_link);
-       kvm_mmu_free_page(vcpu, page->page_hpa);
+       if (!page->root_count) {
+               hlist_del(&page->hash_link);
+               kvm_mmu_free_page(vcpu, page->page_hpa);
+       } else {
+               list_del(&page->link);
+               list_add(&page->link, &vcpu->kvm->active_mmu_pages);
+       }
 }
 
 static int kvm_mmu_unprotect_page(struct kvm_vcpu *vcpu, gfn_t gfn)
@@ -667,12 +672,15 @@ static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, hpa_t p)
 static void mmu_free_roots(struct kvm_vcpu *vcpu)
 {
        int i;
+       struct kvm_mmu_page *page;
 
 #ifdef CONFIG_X86_64
        if (vcpu->mmu.shadow_root_level == PT64_ROOT_LEVEL) {
                hpa_t root = vcpu->mmu.root_hpa;
 
                ASSERT(VALID_PAGE(root));
+               page = page_header(root);
+               --page->root_count;
                vcpu->mmu.root_hpa = INVALID_PAGE;
                return;
        }
@@ -682,6 +690,8 @@ static void mmu_free_roots(struct kvm_vcpu *vcpu)
 
                ASSERT(VALID_PAGE(root));
                root &= PT64_BASE_ADDR_MASK;
+               page = page_header(root);
+               --page->root_count;
                vcpu->mmu.pae_root[i] = INVALID_PAGE;
        }
        vcpu->mmu.root_hpa = INVALID_PAGE;
@@ -691,6 +701,8 @@ static void mmu_alloc_roots(struct kvm_vcpu *vcpu)
 {
        int i;
        gfn_t root_gfn;
+       struct kvm_mmu_page *page;
+
        root_gfn = vcpu->cr3 >> PAGE_SHIFT;
 
 #ifdef CONFIG_X86_64
@@ -700,6 +712,8 @@ static void mmu_alloc_roots(struct kvm_vcpu *vcpu)
                ASSERT(!VALID_PAGE(root));
                root = kvm_mmu_get_page(vcpu, root_gfn, 0,
                                        PT64_ROOT_LEVEL, 0, NULL)->page_hpa;
+               page = page_header(root);
+               ++page->root_count;
                vcpu->mmu.root_hpa = root;
                return;
        }
@@ -715,6 +729,8 @@ static void mmu_alloc_roots(struct kvm_vcpu *vcpu)
                root = kvm_mmu_get_page(vcpu, root_gfn, i << 30,
                                        PT32_ROOT_LEVEL, !is_paging(vcpu),
                                        NULL)->page_hpa;
+               page = page_header(root);
+               ++page->root_count;
                vcpu->mmu.pae_root[i] = root | PT_PRESENT_MASK;
        }
        vcpu->mmu.root_hpa = __pa(vcpu->mmu.pae_root);