KVM: MMU: mark page table dirty when a pte is actually modified

Sometime cmpxchg_gpte doesn't modify gpte, in such case, don't mark page table page as dirty. Signed-off-by: Gui Jianfeng <guijianfeng@cn.fujitsu.com> Signed-off-by: Avi Kivity <avi@redhat.com>
mariux64 · Aug 1, 2010 · f3b8c96 · f3b8c96
1 parent eec4b14
commit f3b8c96
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h
@@ -177,10 +177,10 @@ static int FNAME(walk_addr)(struct guest_walker *walker,
 		if (!(pte & PT_ACCESSED_MASK)) {
 			trace_kvm_mmu_set_accessed_bit(table_gfn, index,
 						       sizeof(pte));
-			mark_page_dirty(vcpu->kvm, table_gfn);
 			if (FNAME(cmpxchg_gpte)(vcpu->kvm, table_gfn,
 			    index, pte, pte|PT_ACCESSED_MASK))
 				goto walk;
+			mark_page_dirty(vcpu->kvm, table_gfn);
 			pte |= PT_ACCESSED_MASK;
 		}
 
@@ -217,11 +217,11 @@ static int FNAME(walk_addr)(struct guest_walker *walker,
 		bool ret;
 
 		trace_kvm_mmu_set_dirty_bit(table_gfn, index, sizeof(pte));
-		mark_page_dirty(vcpu->kvm, table_gfn);
 		ret = FNAME(cmpxchg_gpte)(vcpu->kvm, table_gfn, index, pte,
 			    pte|PT_DIRTY_MASK);
 		if (ret)
 			goto walk;
+		mark_page_dirty(vcpu->kvm, table_gfn);
 		pte |= PT_DIRTY_MASK;
 		walker->ptes[walker->level - 1] = pte;
 	}