diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index f446c36f5800..1353439691cf 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -1883,6 +1883,13 @@ static bool vma_is_valid(struct vm_area_struct *vma, bool write_fault)
 	return true;
 }
 
+static int kvm_try_get_pfn(kvm_pfn_t pfn)
+{
+	if (kvm_is_reserved_pfn(pfn))
+		return 1;
+	return get_page_unless_zero(pfn_to_page(pfn));
+}
+
 static int hva_to_pfn_remapped(struct vm_area_struct *vma,
 			       unsigned long addr, bool *async,
 			       bool write_fault, bool *writable,
@@ -1932,13 +1939,21 @@ static int hva_to_pfn_remapped(struct vm_area_struct *vma,
 	 * Whoever called remap_pfn_range is also going to call e.g.
 	 * unmap_mapping_range before the underlying pages are freed,
 	 * causing a call to our MMU notifier.
+	 *
+	 * Certain IO or PFNMAP mappings can be backed with valid
+	 * struct pages, but be allocated without refcounting e.g.,
+	 * tail pages of non-compound higher order allocations, which
+	 * would then underflow the refcount when the caller does the
+	 * required put_page. Don't allow those pages here.
 	 */ 
-	kvm_get_pfn(pfn);
+	if (!kvm_try_get_pfn(pfn))
+		r = -EFAULT;
 
 out:
 	pte_unmap_unlock(ptep, ptl);
 	*p_pfn = pfn;
-	return 0;
+
+	return r;
 }
 
 /*