Werkzeug may allow high resource usage when parsing multipart form data with many fields

@@ -21,6 +21,10 @@ Unreleased
     the requested size in one ``read`` call. :issue:`2558`
 -   A cookie header that starts with ``=`` is treated as an empty key and discarded,
     rather than stripping the leading ``==``.
+-   Specify a maximum number of multipart parts, default 1000, after which a
+    ``RequestEntityTooLarge`` exception is raised on parsing. This mitigates a DoS
+    attack where a larger number of form/file parts would result in disproportionate
+    resource use.
 
 
 Version 2.2.2

@@ -73,23 +73,26 @@ read the stream *or* call :meth:`~Request.get_data`.
 Limiting Request Data
 ---------------------
 
-To avoid being the victim of a DDOS attack you can set the maximum
-accepted content length and request field sizes.  The :class:`Request`
-class has two attributes for that: :attr:`~Request.max_content_length`
-and :attr:`~Request.max_form_memory_size`.
-
-The first one can be used to limit the total content length.  For example
-by setting it to ``1024 * 1024 * 16`` the request won't accept more than
-16MB of transmitted data.
-
-Because certain data can't be moved to the hard disk (regular post data)
-whereas temporary files can, there is a second limit you can set.  The
-:attr:`~Request.max_form_memory_size` limits the size of `POST`
-transmitted form data.  By setting it to ``1024 * 1024 * 2`` you can make
-sure that all in memory-stored fields are not more than 2MB in size.
-
-This however does *not* affect in-memory stored files if the
-`stream_factory` used returns a in-memory file.
+The :class:`Request` class provides a few attributes to control how much data is
+processed from the request body. This can help mitigate DoS attacks that craft the
+request in such a way that the server uses too many resources to handle it. Each of
+these limits will raise a :exc:`~werkzeug.exceptions.RequestEntityTooLarge` if they are
+exceeded.
+
+-   :attr:`~Request.max_content_length` Stop reading request data after this number
+    of bytes. It's better to configure this in the WSGI server or HTTP server, rather
+    than the WSGI application.
+-   :attr:`~Request.max_form_memory_size` Stop reading request data if any form part is
+    larger than this number of bytes. While file parts can be moved to disk, regular
+    form field data is stored in memory only.
+-   :attr:`~Request.max_form_parts` Stop reading request data if more than this number
+    of parts are sent in multipart form data. This is useful to stop a very large number
+    of very small parts, especially file parts. The default is 1000.
+
+Using Werkzeug to set these limits is only one layer of protection. WSGI servers
+and HTTPS servers should set their own limits on size and timeouts. The operating system
+or container manager should set limits on memory and processing time for server
+processes.
 
 
 How to extend Parsing?

@@ -179,6 +179,8 @@ class FormDataParser:
     :param cls: an optional dict class to use.  If this is not specified
                        or `None` the default :class:`MultiDict` is used.
     :param silent: If set to False parsing errors will not be caught.
+    :param max_form_parts: The maximum number of parts to be parsed. If this is
+        exceeded, a :exc:`~exceptions.RequestEntityTooLarge` exception is raised.
     """
 
     def __init__(
@@ -190,6 +192,8 @@ def __init__(
         max_content_length: t.Optional[int] = None,
         cls: t.Optional[t.Type[MultiDict]] = None,
         silent: bool = True,
+        *,
+        max_form_parts: t.Optional[int] = None,
     ) -> None:
         if stream_factory is None:
             stream_factory = default_stream_factory
@@ -199,6 +203,7 @@ def __init__(
         self.errors = errors
         self.max_form_memory_size = max_form_memory_size
         self.max_content_length = max_content_length
+        self.max_form_parts = max_form_parts
 
         if cls is None:
             cls = MultiDict
@@ -281,6 +286,7 @@ def _parse_multipart(
             self.errors,
             max_form_memory_size=self.max_form_memory_size,
             cls=self.cls,
+            max_form_parts=self.max_form_parts,
         )
         boundary = options.get("boundary", "").encode("ascii")
 
@@ -346,10 +352,12 @@ def __init__(
         max_form_memory_size: t.Optional[int] = None,
         cls: t.Optional[t.Type[MultiDict]] = None,
         buffer_size: int = 64 * 1024,
+        max_form_parts: t.Optional[int] = None,
     ) -> None:
         self.charset = charset
         self.errors = errors
         self.max_form_memory_size = max_form_memory_size
+        self.max_form_parts = max_form_parts
 
         if stream_factory is None:
             stream_factory = default_stream_factory
@@ -409,7 +417,9 @@ def parse(
             [None],
         )
 
-        parser = MultipartDecoder(boundary, self.max_form_memory_size)
+        parser = MultipartDecoder(
+            boundary, self.max_form_memory_size, max_parts=self.max_form_parts
+        )
 
         fields = []
         files = []

@@ -87,10 +87,13 @@ def __init__(
         self,
         boundary: bytes,
         max_form_memory_size: Optional[int] = None,
+        *,
+        max_parts: Optional[int] = None,
     ) -> None:
         self.buffer = bytearray()
         self.complete = False
         self.max_form_memory_size = max_form_memory_size
+        self.max_parts = max_parts
         self.state = State.PREAMBLE
         self.boundary = boundary
 
@@ -118,6 +121,7 @@ def __init__(
             re.MULTILINE,
         )
         self._search_position = 0
+        self._parts_decoded = 0
 
     def last_newline(self) -> int:
         try:
@@ -191,6 +195,10 @@ def next_event(self) -> Event:
                     )
                 self.state = State.DATA
                 self._search_position = 0
+                self._parts_decoded += 1
+
+                if self.max_parts is not None and self._parts_decoded > self.max_parts:
+                    raise RequestEntityTooLarge()
             else:
                 # Update the search start position to be equal to the
                 # current buffer length (already searched) minus a

@@ -83,6 +83,13 @@ class Request(_SansIORequest):
     #: .. versionadded:: 0.5
     max_form_memory_size: t.Optional[int] = None
 
+    #: The maximum number of multipart parts to parse, passed to
+    #: :attr:`form_data_parser_class`. Parsing form data with more than this
+    #: many parts will raise :exc:`~.RequestEntityTooLarge`.
+    #:
+    #: .. versionadded:: 2.2.3
+    max_form_parts = 1000
+
     #: The form data parser that should be used.  Can be replaced to customize
     #: the form date parsing.
     form_data_parser_class: t.Type[FormDataParser] = FormDataParser
@@ -246,6 +253,7 @@ def make_form_data_parser(self) -> FormDataParser:
             self.max_form_memory_size,
             self.max_content_length,
             self.parameter_storage_class,
+            max_form_parts=self.max_form_parts,
         )
 
     def _load_form_data(self) -> None:

@@ -127,6 +127,15 @@ def test_limiting(self):
         req.max_form_memory_size = 400
         assert req.form["foo"] == "Hello World"
 
+        req = Request.from_values(
+            input_stream=io.BytesIO(data),
+            content_length=len(data),
+            content_type="multipart/form-data; boundary=foo",
+            method="POST",
+        )
+        req.max_form_parts = 1
+        pytest.raises(RequestEntityTooLarge, lambda: req.form["foo"])
+
     def test_missing_multipart_boundary(self):
         data = (
             b"--foo\r\nContent-Disposition: form-field; name=foo\r\n\r\n"