]> err.no Git - linux-2.6/commitdiff
[PATCH] KVM: MMU: Write protect guest pages when a shadow is created for them
authorAvi Kivity <avi@qumranet.com>
Sat, 6 Jan 2007 00:36:43 +0000 (16:36 -0800)
committerLinus Torvalds <torvalds@woody.osdl.org>
Sat, 6 Jan 2007 07:55:25 +0000 (23:55 -0800)
When we cache a guest page table into a shadow page table, we need to prevent
further access to that page by the guest, as that would render the cache
incoherent.

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/mmu.c
drivers/kvm/paging_tmpl.h

index 47c699c21c08ddeff894924ece62fab36de4c78c..ba813f49f8aa4148e062827882f3c76de81a92d0 100644 (file)
@@ -274,6 +274,35 @@ static void rmap_remove(struct kvm *kvm, u64 *spte)
        }
 }
 
+static void rmap_write_protect(struct kvm *kvm, u64 gfn)
+{
+       struct page *page;
+       struct kvm_memory_slot *slot;
+       struct kvm_rmap_desc *desc;
+       u64 *spte;
+
+       slot = gfn_to_memslot(kvm, gfn);
+       BUG_ON(!slot);
+       page = gfn_to_page(slot, gfn);
+
+       while (page->private) {
+               if (!(page->private & 1))
+                       spte = (u64 *)page->private;
+               else {
+                       desc = (struct kvm_rmap_desc *)(page->private & ~1ul);
+                       spte = desc->shadow_ptes[0];
+               }
+               BUG_ON(!spte);
+               BUG_ON((*spte & PT64_BASE_ADDR_MASK) !=
+                      page_to_pfn(page) << PAGE_SHIFT);
+               BUG_ON(!(*spte & PT_PRESENT_MASK));
+               BUG_ON(!(*spte & PT_WRITABLE_MASK));
+               rmap_printk("rmap_write_protect: spte %p %llx\n", spte, *spte);
+               rmap_remove(kvm, spte);
+               *spte &= ~(u64)PT_WRITABLE_MASK;
+       }
+}
+
 static void kvm_mmu_free_page(struct kvm_vcpu *vcpu, hpa_t page_hpa)
 {
        struct kvm_mmu_page *page_head = page_header(page_hpa);
@@ -444,6 +473,8 @@ static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
        page->gfn = gfn;
        page->role = role;
        hlist_add_head(&page->hash_link, bucket);
+       if (!metaphysical)
+               rmap_write_protect(vcpu->kvm, gfn);
        return page;
 }
 
@@ -705,6 +736,7 @@ static void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu)
 
 static void paging_new_cr3(struct kvm_vcpu *vcpu)
 {
+       pgprintk("%s: cr3 %lx\n", __FUNCTION__, vcpu->cr3);
        mmu_free_roots(vcpu);
        mmu_alloc_roots(vcpu);
        kvm_mmu_flush_tlb(vcpu);
@@ -727,24 +759,11 @@ static inline void set_pte_common(struct kvm_vcpu *vcpu,
        *shadow_pte |= access_bits << PT_SHADOW_BITS_OFFSET;
        if (!dirty)
                access_bits &= ~PT_WRITABLE_MASK;
-       if (access_bits & PT_WRITABLE_MASK) {
-               struct kvm_mmu_page *shadow;
 
-               shadow = kvm_mmu_lookup_page(vcpu, gaddr >> PAGE_SHIFT);
-               if (shadow)
-                       pgprintk("%s: found shadow page for %lx, marking ro\n",
-                                __FUNCTION__, (gfn_t)(gaddr >> PAGE_SHIFT));
-               if (shadow)
-                       access_bits &= ~PT_WRITABLE_MASK;
-       }
-
-       if (access_bits & PT_WRITABLE_MASK)
-               mark_page_dirty(vcpu->kvm, gaddr >> PAGE_SHIFT);
+       paddr = gpa_to_hpa(vcpu, gaddr & PT64_BASE_ADDR_MASK);
 
        *shadow_pte |= access_bits;
 
-       paddr = gpa_to_hpa(vcpu, gaddr & PT64_BASE_ADDR_MASK);
-
        if (!(*shadow_pte & PT_GLOBAL_MASK))
                mark_pagetable_nonglobal(shadow_pte);
 
@@ -752,11 +771,28 @@ static inline void set_pte_common(struct kvm_vcpu *vcpu,
                *shadow_pte |= gaddr;
                *shadow_pte |= PT_SHADOW_IO_MARK;
                *shadow_pte &= ~PT_PRESENT_MASK;
-       } else {
-               *shadow_pte |= paddr;
-               page_header_update_slot(vcpu->kvm, shadow_pte, gaddr);
-               rmap_add(vcpu->kvm, shadow_pte);
+               return;
        }
+
+       *shadow_pte |= paddr;
+
+       if (access_bits & PT_WRITABLE_MASK) {
+               struct kvm_mmu_page *shadow;
+
+               shadow = kvm_mmu_lookup_page(vcpu, gaddr >> PAGE_SHIFT);
+               if (shadow) {
+                       pgprintk("%s: found shadow page for %lx, marking ro\n",
+                                __FUNCTION__, (gfn_t)(gaddr >> PAGE_SHIFT));
+                       access_bits &= ~PT_WRITABLE_MASK;
+                       *shadow_pte &= ~PT_WRITABLE_MASK;
+               }
+       }
+
+       if (access_bits & PT_WRITABLE_MASK)
+               mark_page_dirty(vcpu->kvm, gaddr >> PAGE_SHIFT);
+
+       page_header_update_slot(vcpu->kvm, shadow_pte, gaddr);
+       rmap_add(vcpu->kvm, shadow_pte);
 }
 
 static void inject_page_fault(struct kvm_vcpu *vcpu,
index f7cce443ca6ff7f7b917eb02564b9fcd1d2eb48c..cd71973c780c958ab355368ba1295fc98268ce25 100644 (file)
@@ -133,6 +133,7 @@ static void FNAME(walk_addr)(struct guest_walker *walker,
                         walker->level - 1, table_gfn);
        }
        walker->ptep = ptep;
+       pgprintk("%s: pte %llx\n", __FUNCTION__, (u64)*ptep);
 }
 
 static void FNAME(release_walker)(struct guest_walker *walker)