diff --git a/scripts/dts/python-devicetree/src/devicetree/dtlib.py b/scripts/dts/python-devicetree/src/devicetree/dtlib.py
index 9ff3b9d94fd..68e3788ca43 100644
--- a/scripts/dts/python-devicetree/src/devicetree/dtlib.py
+++ b/scripts/dts/python-devicetree/src/devicetree/dtlib.py
@@ -17,6 +17,7 @@ import enum
 import errno
 import os
 import re
+import string
 import sys
 import textwrap
 from typing import Any, Dict, Iterable, List, \
@@ -92,6 +93,14 @@ class Node:
         self.parent = parent
         self.dt = dt
 
+        if name.count("@") > 1:
+            dt._parse_error("multiple '@' in node name")
+        if not name == "/":
+            for char in name:
+                if char not in _nodename_chars:
+                    dt._parse_error(f"{self.path}: bad character '{char}' "
+                                    "in node name")
+
         self.props: Dict[str, 'Property'] = collections.OrderedDict()
         self.nodes: Dict[str, 'Node'] = collections.OrderedDict()
         self.labels: List[str] = []
@@ -950,9 +959,6 @@ class DT:
                 if self._peek_token().val == "{":
                     # '<tok> { ...', expect node
 
-                    if tok.val.count("@") > 1:
-                        self._parse_error("multiple '@' in node name")
-
                     # Fetch the existing node if it already exists. This
                     # happens when overriding nodes.
                     child = node.nodes.get(tok.val) or \
@@ -1925,6 +1931,9 @@ _num_re = re.compile(r"(0[xX][0-9a-fA-F]+|[0-9]+)(?:ULL|UL|LL|U|L)?")
 # names that would clash with other stuff
 _propnodename_re = re.compile(r"\\?([a-zA-Z0-9,._+*#?@-]+)")
 
+# Node names are more restrictive than property names.
+_nodename_chars = set(string.ascii_letters + string.digits + ',._+-@')
+
 # Misc. tokens that are tried after a property/node name. This is important, as
 # there's overlap with the allowed characters in names.
 _misc_re = re.compile(
diff --git a/scripts/dts/python-devicetree/tests/test_dtlib.py b/scripts/dts/python-devicetree/tests/test_dtlib.py
index 13e5544da6a..60717bb8aad 100644
--- a/scripts/dts/python-devicetree/tests/test_dtlib.py
+++ b/scripts/dts/python-devicetree/tests/test_dtlib.py
@@ -100,6 +100,16 @@ def temporary_chdir(dirname):
     finally:
         os.chdir(here)
 
+def test_invalid_nodenames():
+    # Regression test that verifies node names are not matched against
+    # the more permissive set of rules used for property names.
+
+    verify_error_endswith("""
+/dts-v1/;
+/ { node? {}; };
+""",
+                          "/node?: bad character '?' in node name")
+
 def test_cell_parsing():
     '''Miscellaneous properties containing zero or more cells'''
 
@@ -2107,15 +2117,12 @@ def test_reprs():
 def test_names():
     '''Tests for node/property names.'''
 
-    # The C tools disallow '@' in property names, but otherwise accept the same
-    # characters in node and property names. Emulate that instead of the DT spec
-    # (v0.2), which gives different characters for nodes and properties.
     verify_parse(r"""
 /dts-v1/;
 
 / {
 	// A leading \ is accepted but ignored in node/propert names
-	\aA0,._+*#?- = &_, &{/aA0,._+*#?@-};
+	\aA0,._+*#?- = &_, &{/aA0,._+@-};
 
 	// Names that overlap with operators and integer literals
 
@@ -2126,7 +2133,8 @@ def test_names():
 	0 = [ 04 ];
 	0x123 = [ 05 ];
 
-	_: \aA0,._+*#?@- {
+	// Node names are more restrictive than property names.
+	_: \aA0,._+@- {
 	};
 
 	0 {
@@ -2137,14 +2145,14 @@ def test_names():
 /dts-v1/;
 
 / {
-	aA0,._+*#?- = &_, &{/aA0,._+*#?@-};
+	aA0,._+*#?- = &_, &{/aA0,._+@-};
 	+ = [ 00 ];
 	* = [ 02 ];
 	- = [ 01 ];
 	? = [ 03 ];
 	0 = [ 04 ];
 	0x123 = [ 05 ];
-	_: aA0,._+*#?@- {
+	_: aA0,._+@- {
 	};
 	0 {
 	};