diff --git a/inc/hf/mm.h b/inc/hf/mm.h
index 613e6e5..ecd205f 100644
--- a/inc/hf/mm.h
+++ b/inc/hf/mm.h
@@ -60,8 +60,8 @@
  *  - !V  O  X : Owner of memory lent to a VM that has exclusive access.
  *
  *  - !V  O !X : Unused. Owner of shared memory always has access.
+ *  - !V !O  X : Unused. Next entry is used for invalid memory.
  *
- *  - !V !O  X : Invalid memory. Memory is unrelated to the VM.
  *  - !V !O !X : Invalid memory. Memory is unrelated to the VM.
  *
  *  Modes are selected so that owner of exclusive memory is the default.
@@ -70,6 +70,9 @@
 #define MM_MODE_UNOWNED UINT32_C(0x0020)
 #define MM_MODE_SHARED  UINT32_C(0x0040)
 
+/* The mask for a mode that is considered unmapped. */
+#define MM_MODE_UNMAPPED_MASK (MM_MODE_INVALID | MM_MODE_UNOWNED)
+
 #define MM_FLAG_COMMIT  0x01
 #define MM_FLAG_UNMAP   0x02
 #define MM_FLAG_STAGE1  0x04
diff --git a/src/mm.c b/src/mm.c
index 918359f..fbc91b4 100644
--- a/src/mm.c
+++ b/src/mm.c
@@ -801,6 +801,19 @@
 }
 
 /**
+ * Selects flags to pass to the page table manipulation operation based on the
+ * mapping mode.
+ */
+static int mm_mode_to_flags(uint32_t mode)
+{
+	if ((mode & MM_MODE_UNMAPPED_MASK) == MM_MODE_UNMAPPED_MASK) {
+		return MM_FLAG_UNMAP;
+	}
+
+	return 0;
+}
+
+/**
  * Updates a VM's page table such that the given physical address range is
  * mapped in the address space at the corresponding address range in the
  * architecture-agnostic mode provided.
@@ -808,7 +821,7 @@
 bool mm_vm_identity_map(struct mm_ptable *t, paddr_t begin, paddr_t end,
 			uint32_t mode, ipaddr_t *ipa, struct mpool *ppool)
 {
-	int flags = 0;
+	int flags = mm_mode_to_flags(mode);
 	bool success = mm_ptable_identity_update(
 		t, begin, end, arch_mm_mode_to_stage2_attrs(mode), flags,
 		ppool);
@@ -827,11 +840,9 @@
 bool mm_vm_unmap(struct mm_ptable *t, paddr_t begin, paddr_t end,
 		 struct mpool *ppool)
 {
-	return mm_ptable_identity_update(
-		t, begin, end,
-		arch_mm_mode_to_stage2_attrs(MM_MODE_UNOWNED | MM_MODE_INVALID |
-					     MM_MODE_SHARED),
-		MM_FLAG_UNMAP, ppool);
+	uint32_t mode = MM_MODE_UNMAPPED_MASK;
+
+	return mm_vm_identity_map(t, begin, end, mode, NULL, ppool);
 }
 
 /**
@@ -908,9 +919,11 @@
 void *mm_identity_map(struct mm_stage1_locked stage1_locked, paddr_t begin,
 		      paddr_t end, uint32_t mode, struct mpool *ppool)
 {
+	int flags = MM_FLAG_STAGE1 | mm_mode_to_flags(mode);
+
 	if (mm_ptable_identity_update(stage1_locked.ptable, begin, end,
-				      arch_mm_mode_to_stage1_attrs(mode),
-				      MM_FLAG_STAGE1, ppool)) {
+				      arch_mm_mode_to_stage1_attrs(mode), flags,
+				      ppool)) {
 		return ptr_from_va(va_from_pa(begin));
 	}
 
@@ -924,11 +937,9 @@
 bool mm_unmap(struct mm_stage1_locked stage1_locked, paddr_t begin, paddr_t end,
 	      struct mpool *ppool)
 {
-	return mm_ptable_identity_update(
-		stage1_locked.ptable, begin, end,
-		arch_mm_mode_to_stage1_attrs(MM_MODE_UNOWNED | MM_MODE_INVALID |
-					     MM_MODE_SHARED),
-		MM_FLAG_STAGE1 | MM_FLAG_UNMAP, ppool);
+	uint32_t mode = MM_MODE_UNMAPPED_MASK;
+
+	return mm_identity_map(stage1_locked, begin, end, mode, ppool);
 }
 
 /**
diff --git a/src/mm_test.cc b/src/mm_test.cc
index ae13df7..05723b2 100644
--- a/src/mm_test.cc
+++ b/src/mm_test.cc
@@ -486,6 +486,31 @@
 }
 
 /**
+ * Mapping with a mode that indicates unmapping results in the addresses being
+ * unmapped with absent entries.
+ */
+TEST_F(mm, map_to_unmap)
+{
+	constexpr uint32_t mode = 0;
+	const paddr_t l0_begin = pa_init(uintpaddr_t(524421) * PAGE_SIZE);
+	const paddr_t l0_end = pa_add(l0_begin, 17 * PAGE_SIZE);
+	const paddr_t l1_begin = pa_init(3 * mm_entry_size(1));
+	const paddr_t l1_end = pa_add(l1_begin, 5 * mm_entry_size(1));
+	struct mm_ptable ptable;
+	ASSERT_TRUE(mm_vm_init(&ptable, &ppool));
+	ASSERT_TRUE(mm_vm_identity_map(&ptable, l0_begin, l0_end, mode, nullptr,
+				       &ppool));
+	ASSERT_TRUE(mm_vm_identity_map(&ptable, l1_begin, l1_end, mode, nullptr,
+				       &ppool));
+	EXPECT_TRUE(mm_vm_identity_map(&ptable, pa_init(0), VM_MEM_END,
+				       MM_MODE_UNMAPPED_MASK, nullptr, &ppool));
+	EXPECT_THAT(
+		get_ptable(ptable),
+		AllOf(SizeIs(4), Each(Each(arch_mm_absent_pte(TOP_LEVEL)))));
+	mm_vm_fini(&ptable, &ppool);
+}
+
+/**
  * If nothing is mapped, unmapping the hypervisor has no effect.
  */
 TEST_F(mm, vm_unmap_hypervisor_not_mapped)