apply max_form_memory_size another level up in the parser

davidism · davidism · commit 8760275afb72 · 2024-10-25T07:37:03.000-07:00
diff --git a/CHANGES.rst b/CHANGES.rst
@@ -5,6 +5,9 @@ Version 3.0.6
 
 Unreleased
 
+-   Fix how ``max_form_memory_size`` is applied when parsing large non-file
+    fields. :ghsa:`q34m-jh98-gwm2`
+
 
 Version 3.0.5
 -------------
diff --git a/src/werkzeug/formparser.py b/src/werkzeug/formparser.py
@@ -352,6 +352,7 @@ def parse(
         self, stream: t.IO[bytes], boundary: bytes, content_length: int | None
     ) -> tuple[MultiDict[str, str], MultiDict[str, FileStorage]]:
         current_part: Field | File
+        field_size: int | None = None
         container: t.IO[bytes] | list[bytes]
         _write: t.Callable[[bytes], t.Any]
 
@@ -370,13 +371,23 @@ def parse(
             while not isinstance(event, (Epilogue, NeedData)):
                 if isinstance(event, Field):
                     current_part = event
+                    field_size = 0
                     container = []
                     _write = container.append
                 elif isinstance(event, File):
                     current_part = event
+                    field_size = None
                     container = self.start_file_streaming(event, content_length)
                     _write = container.write
                 elif isinstance(event, Data):
+                    if self.max_form_memory_size is not None and field_size is not None:
+                        # Ensure that accumulated data events do not exceed limit.
+                        # Also checked within single event in MultipartDecoder.
+                        field_size += len(event.data)
+
+                        if field_size > self.max_form_memory_size:
+                            raise RequestEntityTooLarge()
+
                     _write(event.data)
                     if not event.more_data:
                         if isinstance(current_part, Field):
diff --git a/src/werkzeug/sansio/multipart.py b/src/werkzeug/sansio/multipart.py
@@ -140,6 +140,8 @@ def receive_data(self, data: bytes | None) -> None:
             self.max_form_memory_size is not None
             and len(self.buffer) + len(data) > self.max_form_memory_size
         ):
+            # Ensure that data within single event does not exceed limit.
+            # Also checked across accumulated events in MultiPartParser.
             raise RequestEntityTooLarge()
         else:
             self.buffer.extend(data)
diff --git a/tests/test_formparser.py b/tests/test_formparser.py
@@ -456,3 +456,15 @@ def test_file_rfc2231_filename_continuations(self):
         ) as request:
             assert request.files["rfc2231"].filename == "a b c d e f.txt"
             assert request.files["rfc2231"].read() == b"file contents"
+
+
+def test_multipart_max_form_memory_size() -> None:
+    """max_form_memory_size is tracked across multiple data events."""
+    data = b"--bound\r\nContent-Disposition: form-field; name=a\r\n\r\n"
+    data += b"a" * 15 + b"\r\n--bound--"
+    # The buffer size is less than the max size, so multiple data events will be
+    # returned. The field size is greater than the max.
+    parser = formparser.MultiPartParser(max_form_memory_size=10, buffer_size=5)
+
+    with pytest.raises(RequestEntityTooLarge):
+        parser.parse(io.BytesIO(data), b"bound", None)
