]> git.ipfire.org Git - thirdparty/Python/cpython.git/commitdiff
bpo-44022: Fix http client infinite line reading (DoS) after a HTTP 100 Continue...
authorMiss Islington (bot) <31488909+miss-islington@users.noreply.github.com>
Wed, 5 May 2021 23:14:28 +0000 (16:14 -0700)
committerGitHub <noreply@github.com>
Wed, 5 May 2021 23:14:28 +0000 (16:14 -0700)
Fixes http.client potential denial of service where it could get stuck reading lines from a malicious server after a 100 Continue response.

Co-authored-by: Gregory P. Smith <greg@krypto.org>
(cherry picked from commit 47895e31b6f626bc6ce47d175fe9d43c1098909d)

Co-authored-by: Gen Xu <xgbarry@gmail.com>
Lib/http/client.py
Lib/test/test_httplib.py
Misc/NEWS.d/next/Security/2021-05-05-17-37-04.bpo-44022.bS3XJ9.rst [new file with mode: 0644]

index 4b1f692844474f8971da779895ccb8461fcb32b3..08cf2ed9b3716b7942dec780c24f0388ea471da1 100644 (file)
@@ -202,15 +202,11 @@ class HTTPMessage(email.message.Message):
                 lst.append(line)
         return lst
 
-def parse_headers(fp, _class=HTTPMessage):
-    """Parses only RFC2822 headers from a file pointer.
-
-    email Parser wants to see strings rather than bytes.
-    But a TextIOWrapper around self.rfile would buffer too many bytes
-    from the stream, bytes which we later need to read as bytes.
-    So we read the correct bytes here, as bytes, for email Parser
-    to parse.
+def _read_headers(fp):
+    """Reads potential header lines into a list from a file pointer.
 
+    Length of line is limited by _MAXLINE, and number of
+    headers is limited by _MAXHEADERS.
     """
     headers = []
     while True:
@@ -222,6 +218,19 @@ def parse_headers(fp, _class=HTTPMessage):
             raise HTTPException("got more than %d headers" % _MAXHEADERS)
         if line in (b'\r\n', b'\n', b''):
             break
+    return headers
+
+def parse_headers(fp, _class=HTTPMessage):
+    """Parses only RFC2822 headers from a file pointer.
+
+    email Parser wants to see strings rather than bytes.
+    But a TextIOWrapper around self.rfile would buffer too many bytes
+    from the stream, bytes which we later need to read as bytes.
+    So we read the correct bytes here, as bytes, for email Parser
+    to parse.
+
+    """
+    headers = _read_headers(fp)
     hstring = b''.join(headers).decode('iso-8859-1')
     return email.parser.Parser(_class=_class).parsestr(hstring)
 
@@ -309,15 +318,10 @@ class HTTPResponse(io.BufferedIOBase):
             if status != CONTINUE:
                 break
             # skip the header from the 100 response
-            while True:
-                skip = self.fp.readline(_MAXLINE + 1)
-                if len(skip) > _MAXLINE:
-                    raise LineTooLong("header line")
-                skip = skip.strip()
-                if not skip:
-                    break
-                if self.debuglevel > 0:
-                    print("header:", skip)
+            skipped_headers = _read_headers(self.fp)
+            if self.debuglevel > 0:
+                print("headers:", skipped_headers)
+            del skipped_headers
 
         self.code = self.status = status
         self.reason = reason.strip()
index db41e29a4b839402c0a562eb950588e7bc866065..e9272569ecc531197901ddcb017dedd3cafdde56 100644 (file)
@@ -1180,6 +1180,14 @@ class BasicTest(TestCase):
         resp = client.HTTPResponse(FakeSocket(body))
         self.assertRaises(client.LineTooLong, resp.begin)
 
+    def test_overflowing_header_limit_after_100(self):
+        body = (
+            'HTTP/1.1 100 OK\r\n'
+            'r\n' * 32768
+        )
+        resp = client.HTTPResponse(FakeSocket(body))
+        self.assertRaises(client.HTTPException, resp.begin)
+
     def test_overflowing_chunked_line(self):
         body = (
             'HTTP/1.1 200 OK\r\n'
@@ -1581,7 +1589,7 @@ class Readliner:
 class OfflineTest(TestCase):
     def test_all(self):
         # Documented objects defined in the module should be in __all__
-        expected = {"responses"}  # White-list documented dict() object
+        expected = {"responses"}  # Allowlist documented dict() object
         # HTTPMessage, parse_headers(), and the HTTP status code constants are
         # intentionally omitted for simplicity
         denylist = {"HTTPMessage", "parse_headers"}
diff --git a/Misc/NEWS.d/next/Security/2021-05-05-17-37-04.bpo-44022.bS3XJ9.rst b/Misc/NEWS.d/next/Security/2021-05-05-17-37-04.bpo-44022.bS3XJ9.rst
new file mode 100644 (file)
index 0000000..cf6b63e
--- /dev/null
@@ -0,0 +1,2 @@
+mod:`http.client` now avoids infinitely reading potential HTTP headers after a
+``100 Continue`` status response from the server.