]> git.ipfire.org Git - thirdparty/tornado.git/commitdiff
httputil: Raise errors instead of logging in multipart/form-data parsing 3497/head
authorBen Darnell <ben@bendarnell.com>
Thu, 8 May 2025 17:29:43 +0000 (13:29 -0400)
committerBen Darnell <ben@bendarnell.com>
Thu, 15 May 2025 19:33:25 +0000 (15:33 -0400)
We used to continue after logging an error, which allowed repeated
errors to spam the logs. The error raised here will still be logged,
but only once per request, consistent with other error handling in
Tornado.

tornado/httputil.py
tornado/test/httpserver_test.py
tornado/test/httputil_test.py
tornado/web.py

index 7044aca02b437e7c5fe5ef8b3cabff8eca9cf6ff..ef460985eedc8983d9c9352a2de9011b490635fd 100644 (file)
@@ -34,7 +34,6 @@ import unicodedata
 from urllib.parse import urlencode, urlparse, urlunparse, parse_qsl
 
 from tornado.escape import native_str, parse_qs_bytes, utf8, to_unicode
-from tornado.log import gen_log
 from tornado.util import ObjectDict, unicode_type
 
 
@@ -884,25 +883,22 @@ def parse_body_arguments(
     """
     if content_type.startswith("application/x-www-form-urlencoded"):
         if headers and "Content-Encoding" in headers:
-            gen_log.warning(
-                "Unsupported Content-Encoding: %s", headers["Content-Encoding"]
+            raise HTTPInputError(
+                "Unsupported Content-Encoding: %s" % headers["Content-Encoding"]
             )
-            return
         try:
             # real charset decoding will happen in RequestHandler.decode_argument()
             uri_arguments = parse_qs_bytes(body, keep_blank_values=True)
         except Exception as e:
-            gen_log.warning("Invalid x-www-form-urlencoded body: %s", e)
-            uri_arguments = {}
+            raise HTTPInputError("Invalid x-www-form-urlencoded body: %s" % e) from e
         for name, values in uri_arguments.items():
             if values:
                 arguments.setdefault(name, []).extend(values)
     elif content_type.startswith("multipart/form-data"):
         if headers and "Content-Encoding" in headers:
-            gen_log.warning(
-                "Unsupported Content-Encoding: %s", headers["Content-Encoding"]
+            raise HTTPInputError(
+                "Unsupported Content-Encoding: %s" % headers["Content-Encoding"]
             )
-            return
         try:
             fields = content_type.split(";")
             for field in fields:
@@ -911,9 +907,9 @@ def parse_body_arguments(
                     parse_multipart_form_data(utf8(v), body, arguments, files)
                     break
             else:
-                raise ValueError("multipart boundary not found")
+                raise HTTPInputError("multipart boundary not found")
         except Exception as e:
-            gen_log.warning("Invalid multipart/form-data: %s", e)
+            raise HTTPInputError("Invalid multipart/form-data: %s" % e) from e
 
 
 def parse_multipart_form_data(
@@ -942,26 +938,22 @@ def parse_multipart_form_data(
         boundary = boundary[1:-1]
     final_boundary_index = data.rfind(b"--" + boundary + b"--")
     if final_boundary_index == -1:
-        gen_log.warning("Invalid multipart/form-data: no final boundary")
-        return
+        raise HTTPInputError("Invalid multipart/form-data: no final boundary found")
     parts = data[:final_boundary_index].split(b"--" + boundary + b"\r\n")
     for part in parts:
         if not part:
             continue
         eoh = part.find(b"\r\n\r\n")
         if eoh == -1:
-            gen_log.warning("multipart/form-data missing headers")
-            continue
+            raise HTTPInputError("multipart/form-data missing headers")
         headers = HTTPHeaders.parse(part[:eoh].decode("utf-8"))
         disp_header = headers.get("Content-Disposition", "")
         disposition, disp_params = _parse_header(disp_header)
         if disposition != "form-data" or not part.endswith(b"\r\n"):
-            gen_log.warning("Invalid multipart/form-data")
-            continue
+            raise HTTPInputError("Invalid multipart/form-data")
         value = part[eoh + 4 : -2]
         if not disp_params.get("name"):
-            gen_log.warning("multipart/form-data value missing name")
-            continue
+            raise HTTPInputError("multipart/form-data missing name")
         name = disp_params["name"]
         if disp_params.get("filename"):
             ctype = headers.get("Content-Type", "application/unknown")
index 570cb64ca63c4b36c3a9024afa1a6bdd78cae705..f197cfef8466bde418b0e9544b94af93c0c13680 100644 (file)
@@ -1148,9 +1148,9 @@ class GzipUnsupportedTest(GzipBaseTest, AsyncHTTPTestCase):
         # Gzip support is opt-in; without it the server fails to parse
         # the body (but parsing form bodies is currently just a log message,
         # not a fatal error).
-        with ExpectLog(gen_log, "Unsupported Content-Encoding"):
+        with ExpectLog(gen_log, ".*Unsupported Content-Encoding"):
             response = self.post_gzip("foo=bar")
-        self.assertEqual(json_decode(response.body), {})
+        self.assertEqual(response.code, 400)
 
 
 class StreamingChunkSizeTest(AsyncHTTPTestCase):
index 30fbec4d78e626c6bd56b87657baf76b05be84ee..7a09beaa3f316fea64d4a9b24c3e5e8a9df0fee1 100644 (file)
@@ -12,7 +12,6 @@ from tornado.httputil import (
 )
 from tornado.escape import utf8, native_str
 from tornado.log import gen_log
-from tornado.testing import ExpectLog
 from tornado.test.util import ignore_deprecation
 
 import copy
@@ -195,7 +194,9 @@ Foo
             b"\n", b"\r\n"
         )
         args, files = form_data_args()
-        with ExpectLog(gen_log, "multipart/form-data missing headers"):
+        with self.assertRaises(
+            HTTPInputError, msg="multipart/form-data missing headers"
+        ):
             parse_multipart_form_data(b"1234", data, args, files)
         self.assertEqual(files, {})
 
@@ -209,7 +210,7 @@ Foo
             b"\n", b"\r\n"
         )
         args, files = form_data_args()
-        with ExpectLog(gen_log, "Invalid multipart/form-data"):
+        with self.assertRaises(HTTPInputError, msg="Invalid multipart/form-data"):
             parse_multipart_form_data(b"1234", data, args, files)
         self.assertEqual(files, {})
 
@@ -222,7 +223,7 @@ Foo--1234--""".replace(
             b"\n", b"\r\n"
         )
         args, files = form_data_args()
-        with ExpectLog(gen_log, "Invalid multipart/form-data"):
+        with self.assertRaises(HTTPInputError, msg="Invalid multipart/form-data"):
             parse_multipart_form_data(b"1234", data, args, files)
         self.assertEqual(files, {})
 
@@ -236,7 +237,9 @@ Foo
             b"\n", b"\r\n"
         )
         args, files = form_data_args()
-        with ExpectLog(gen_log, "multipart/form-data value missing name"):
+        with self.assertRaises(
+            HTTPInputError, msg="multipart/form-data value missing name"
+        ):
             parse_multipart_form_data(b"1234", data, args, files)
         self.assertEqual(files, {})
 
index 0303f547eccacd621662432a88312426d185228d..2f702d6480b6d6d4657f7661a9bc0183c19dbf22 100644 (file)
@@ -1801,6 +1801,14 @@ class RequestHandler:
         try:
             if self.request.method not in self.SUPPORTED_METHODS:
                 raise HTTPError(405)
+
+            # If we're not in stream_request_body mode, this is the place where we parse the body.
+            if not _has_stream_request_body(self.__class__):
+                try:
+                    self.request._parse_body()
+                except httputil.HTTPInputError as e:
+                    raise HTTPError(400, "Invalid body: %s" % e) from e
+
             self.path_args = [self.decode_argument(arg) for arg in args]
             self.path_kwargs = {
                 k: self.decode_argument(v, name=k) for (k, v) in kwargs.items()
@@ -1992,7 +2000,7 @@ def _has_stream_request_body(cls: Type[RequestHandler]) -> bool:
 
 
 def removeslash(
-    method: Callable[..., Optional[Awaitable[None]]]
+    method: Callable[..., Optional[Awaitable[None]]],
 ) -> Callable[..., Optional[Awaitable[None]]]:
     """Use this decorator to remove trailing slashes from the request path.
 
@@ -2021,7 +2029,7 @@ def removeslash(
 
 
 def addslash(
-    method: Callable[..., Optional[Awaitable[None]]]
+    method: Callable[..., Optional[Awaitable[None]]],
 ) -> Callable[..., Optional[Awaitable[None]]]:
     """Use this decorator to add a missing trailing slash to the request path.
 
@@ -2445,8 +2453,9 @@ class _HandlerDelegate(httputil.HTTPMessageDelegate):
         if self.stream_request_body:
             future_set_result_unless_cancelled(self.request._body_future, None)
         else:
+            # Note that the body gets parsed in RequestHandler._execute so it can be in
+            # the right exception handler scope.
             self.request.body = b"".join(self.chunks)
-            self.request._parse_body()
             self.execute()
 
     def on_connection_close(self) -> None:
@@ -3332,7 +3341,7 @@ class GZipContentEncoding(OutputTransform):
 
 
 def authenticated(
-    method: Callable[..., Optional[Awaitable[None]]]
+    method: Callable[..., Optional[Awaitable[None]]],
 ) -> Callable[..., Optional[Awaitable[None]]]:
     """Decorate methods with this to require that the user be logged in.