commit 9e77ec82c40ab59846f9447b7c483e7b8e368b16 Author: Petr Viktorin Date: Thu Mar 4 13:59:56 2021 +0100 CVE-2021-23336: Add `separator` argument to parse_qs; warn with default Partially backports https://bugs.python.org/issue42967 : [security] Address a web cache-poisoning issue reported in urllib.parse.parse_qsl(). However, this solution is different than the upstream solution in Python 3.6.13. An optional argument seperator is added to specify the separator. It is recommended to set it to '&' or ';' to match the application or proxy in use. The default can be set with an env variable of a config file. If neither the argument, env var or config file specifies a separator, "&" is used but a warning is raised if parse_qs is used on input that contains ';'. Co-authors of the upstream change (who do not necessarily agree with this): Co-authored-by: Adam Goldschmidt Co-authored-by: Ken Jin <28750310+Fidget-Spinner@users.noreply.github.com> Co-authored-by: Éric Araujo diff --git a/Doc/library/cgi.rst b/Doc/library/cgi.rst index 41219eeaaba..ddecc0af23a 100644 --- a/Doc/library/cgi.rst +++ b/Doc/library/cgi.rst @@ -277,13 +277,12 @@ These are useful if you want more control, or if you want to employ some of the algorithms implemented in this module in other circumstances. -.. function:: parse(fp=None, environ=os.environ, keep_blank_values=False, strict_parsing=False) +.. function:: parse(fp=None, environ=os.environ, keep_blank_values=False, strict_parsing=False, separator=None) Parse a query in the environment or from a file (the file defaults to - ``sys.stdin``). The *keep_blank_values* and *strict_parsing* parameters are + ``sys.stdin``). The *keep_blank_values*, *strict_parsing* and *separator* parameters are passed to :func:`urllib.parse.parse_qs` unchanged. - .. function:: parse_qs(qs, keep_blank_values=False, strict_parsing=False) This function is deprecated in this module. Use :func:`urllib.parse.parse_qs` @@ -308,7 +307,6 @@ algorithms implemented in this module in other circumstances. Note that this does not parse nested multipart parts --- use :class:`FieldStorage` for that. - .. function:: parse_header(string) Parse a MIME header (such as :mailheader:`Content-Type`) into a main value and a diff --git a/Doc/library/urllib.parse.rst b/Doc/library/urllib.parse.rst index 647af613a31..bcab7c142bc 100644 --- a/Doc/library/urllib.parse.rst +++ b/Doc/library/urllib.parse.rst @@ -143,7 +143,7 @@ or on combining URL components into a URL string. now raise :exc:`ValueError`. -.. function:: parse_qs(qs, keep_blank_values=False, strict_parsing=False, encoding='utf-8', errors='replace', max_num_fields=None) +.. function:: parse_qs(qs, keep_blank_values=False, strict_parsing=False, encoding='utf-8', errors='replace', max_num_fields=None, separator=None) Parse a query string given as a string argument (data of type :mimetype:`application/x-www-form-urlencoded`). Data are returned as a @@ -168,6 +168,15 @@ or on combining URL components into a URL string. read. If set, then throws a :exc:`ValueError` if there are more than *max_num_fields* fields read. + The optional argument *separator* is the symbol to use for separating the + query arguments. It is recommended to set it to ``'&'`` or ``';'``. + It defaults to ``'&'``; a warning is raised if this default is used. + This default may be changed with the following environment variable settings: + + - ``PYTHON_URLLIB_QS_SEPARATOR='&'``: use only ``&`` as separator, without warning (as in Python 3.6.13+ or 3.10) + - ``PYTHON_URLLIB_QS_SEPARATOR=';'``: use only ``;`` as separator + - ``PYTHON_URLLIB_QS_SEPARATOR=legacy``: use both ``&`` and ``;`` (as in previous versions of Python) + Use the :func:`urllib.parse.urlencode` function (with the ``doseq`` parameter set to ``True``) to convert such dictionaries into query strings. @@ -204,6 +213,9 @@ or on combining URL components into a URL string. read. If set, then throws a :exc:`ValueError` if there are more than *max_num_fields* fields read. + The optional argument *separator* is the symbol to use for separating the + query arguments. It works as in :py:func:`parse_qs`. + Use the :func:`urllib.parse.urlencode` function to convert such lists of pairs into query strings. @@ -213,7 +225,6 @@ or on combining URL components into a URL string. .. versionchanged:: 3.6.8 Added *max_num_fields* parameter. - .. function:: urlunparse(parts) Construct a URL from a tuple as returned by ``urlparse()``. The *parts* diff --git a/Lib/cgi.py b/Lib/cgi.py index 56f243e09f0..5ab2a5d6af6 100755 --- a/Lib/cgi.py +++ b/Lib/cgi.py @@ -117,7 +117,8 @@ log = initlog # The current logging function # 0 ==> unlimited input maxlen = 0 -def parse(fp=None, environ=os.environ, keep_blank_values=0, strict_parsing=0): +def parse(fp=None, environ=os.environ, keep_blank_values=0, + strict_parsing=0, separator=None): """Parse a query in the environment or from a file (default stdin) Arguments, all optional: @@ -136,6 +137,8 @@ def parse(fp=None, environ=os.environ, keep_blank_values=0, strict_parsing=0): strict_parsing: flag indicating what to do with parsing errors. If false (the default), errors are silently ignored. If true, errors raise a ValueError exception. + + separator: str. The symbol to use for separating the query arguments. """ if fp is None: fp = sys.stdin @@ -156,7 +159,7 @@ def parse(fp=None, environ=os.environ, keep_blank_values=0, strict_parsing=0): if environ['REQUEST_METHOD'] == 'POST': ctype, pdict = parse_header(environ['CONTENT_TYPE']) if ctype == 'multipart/form-data': - return parse_multipart(fp, pdict) + return parse_multipart(fp, pdict, separator=separator) elif ctype == 'application/x-www-form-urlencoded': clength = int(environ['CONTENT_LENGTH']) if maxlen and clength > maxlen: @@ -182,21 +185,21 @@ def parse(fp=None, environ=os.environ, keep_blank_values=0, strict_parsing=0): return urllib.parse.parse_qs(qs, keep_blank_values, strict_parsing, encoding=encoding) - # parse query string function called from urlparse, # this is done in order to maintain backward compatibility. - -def parse_qs(qs, keep_blank_values=0, strict_parsing=0): +def parse_qs(qs, keep_blank_values=0, strict_parsing=0, separator=None): """Parse a query given as a string argument.""" warn("cgi.parse_qs is deprecated, use urllib.parse.parse_qs instead", DeprecationWarning, 2) - return urllib.parse.parse_qs(qs, keep_blank_values, strict_parsing) + return urllib.parse.parse_qs(qs, keep_blank_values, strict_parsing, + separator=separator) -def parse_qsl(qs, keep_blank_values=0, strict_parsing=0): +def parse_qsl(qs, keep_blank_values=0, strict_parsing=0, separator=None): """Parse a query given as a string argument.""" warn("cgi.parse_qsl is deprecated, use urllib.parse.parse_qsl instead", DeprecationWarning, 2) - return urllib.parse.parse_qsl(qs, keep_blank_values, strict_parsing) + return urllib.parse.parse_qsl(qs, keep_blank_values, strict_parsing, + separator=separator) def parse_multipart(fp, pdict): """Parse multipart input. @@ -297,7 +300,6 @@ def parse_multipart(fp, pdict): return partdict - def _parseparam(s): while s[:1] == ';': s = s[1:] @@ -405,7 +407,7 @@ class FieldStorage: def __init__(self, fp=None, headers=None, outerboundary=b'', environ=os.environ, keep_blank_values=0, strict_parsing=0, limit=None, encoding='utf-8', errors='replace', - max_num_fields=None): + max_num_fields=None, separator=None): """Constructor. Read multipart/* until last part. Arguments, all optional: @@ -453,6 +455,7 @@ class FieldStorage: self.keep_blank_values = keep_blank_values self.strict_parsing = strict_parsing self.max_num_fields = max_num_fields + self.separator = separator if 'REQUEST_METHOD' in environ: method = environ['REQUEST_METHOD'].upper() self.qs_on_post = None @@ -678,7 +681,7 @@ class FieldStorage: query = urllib.parse.parse_qsl( qs, self.keep_blank_values, self.strict_parsing, encoding=self.encoding, errors=self.errors, - max_num_fields=self.max_num_fields) + max_num_fields=self.max_num_fields, separator=self.separator) self.list = [MiniFieldStorage(key, value) for key, value in query] self.skip_lines() @@ -694,7 +697,7 @@ class FieldStorage: query = urllib.parse.parse_qsl( self.qs_on_post, self.keep_blank_values, self.strict_parsing, encoding=self.encoding, errors=self.errors, - max_num_fields=self.max_num_fields) + max_num_fields=self.max_num_fields, separator=self.separator) self.list.extend(MiniFieldStorage(key, value) for key, value in query) klass = self.FieldStorageClass or self.__class__ @@ -736,7 +739,8 @@ class FieldStorage: part = klass(self.fp, headers, ib, environ, keep_blank_values, strict_parsing,self.limit-self.bytes_read, - self.encoding, self.errors, max_num_fields) + self.encoding, self.errors, max_num_fields, + separator=self.separator) if max_num_fields is not None: max_num_fields -= 1 diff --git a/Lib/test/test_cgi.py b/Lib/test/test_cgi.py index b3e2d4cce8e..5ae3e085e1e 100644 --- a/Lib/test/test_cgi.py +++ b/Lib/test/test_cgi.py @@ -55,12 +55,9 @@ parse_strict_test_cases = [ ("", ValueError("bad query field: ''")), ("&", ValueError("bad query field: ''")), ("&&", ValueError("bad query field: ''")), - (";", ValueError("bad query field: ''")), - (";&;", ValueError("bad query field: ''")), # Should the next few really be valid? ("=", {}), ("=&=", {}), - ("=;=", {}), # This rest seem to make sense ("=a", {'': ['a']}), ("&=a", ValueError("bad query field: ''")), @@ -75,8 +72,6 @@ parse_strict_test_cases = [ ("a=a+b&b=b+c", {'a': ['a b'], 'b': ['b c']}), ("a=a+b&a=b+a", {'a': ['a b', 'b a']}), ("x=1&y=2.0&z=2-3.%2b0", {'x': ['1'], 'y': ['2.0'], 'z': ['2-3.+0']}), - ("x=1;y=2.0&z=2-3.%2b0", {'x': ['1'], 'y': ['2.0'], 'z': ['2-3.+0']}), - ("x=1;y=2.0;z=2-3.%2b0", {'x': ['1'], 'y': ['2.0'], 'z': ['2-3.+0']}), ("Hbc5161168c542333633315dee1182227:key_store_seqid=400006&cuyer=r&view=bustomer&order_id=0bb2e248638833d48cb7fed300000f1b&expire=964546263&lobale=en-US&kid=130003.300038&ss=env", {'Hbc5161168c542333633315dee1182227:key_store_seqid': ['400006'], 'cuyer': ['r'], @@ -164,6 +159,35 @@ class CgiTests(unittest.TestCase): env = {'QUERY_STRING': orig} fs = cgi.FieldStorage(environ=env) + if isinstance(expect, dict): + # test dict interface + self.assertEqual(len(expect), len(fs)) + self.assertCountEqual(expect.keys(), fs.keys()) + self.assertEqual(fs.getvalue("nonexistent field", "default"), "default") + # test individual fields + for key in expect.keys(): + expect_val = expect[key] + self.assertIn(key, fs) + if len(expect_val) > 1: + self.assertEqual(fs.getvalue(key), expect_val) + else: + self.assertEqual(fs.getvalue(key), expect_val[0]) + + def test_separator(self): + parse_semicolon = [ + ("x=1;y=2.0", {'x': ['1'], 'y': ['2.0']}), + ("x=1;y=2.0;z=2-3.%2b0", {'x': ['1'], 'y': ['2.0'], 'z': ['2-3.+0']}), + (";", ValueError("bad query field: ''")), + (";;", ValueError("bad query field: ''")), + ("=;a", ValueError("bad query field: 'a'")), + (";b=a", ValueError("bad query field: ''")), + ("b;=a", ValueError("bad query field: 'b'")), + ("a=a+b;b=b+c", {'a': ['a b'], 'b': ['b c']}), + ("a=a+b;a=b+a", {'a': ['a b', 'b a']}), + ] + for orig, expect in parse_semicolon: + env = {'QUERY_STRING': orig} + fs = cgi.FieldStorage(separator=';', environ=env) if isinstance(expect, dict): # test dict interface self.assertEqual(len(expect), len(fs)) diff --git a/Lib/test/test_urlparse.py b/Lib/test/test_urlparse.py index 68f633ca3a7..1ec86ba0fc2 100644 --- a/Lib/test/test_urlparse.py +++ b/Lib/test/test_urlparse.py @@ -2,6 +2,11 @@ import sys import unicodedata import unittest import urllib.parse +from test.support import EnvironmentVarGuard +from warnings import catch_warnings +import tempfile +import contextlib +import os.path RFC1808_BASE = "http://a/b/c/d;p?q#f" RFC2396_BASE = "http://a/b/c/d;p?q" @@ -32,6 +37,9 @@ parse_qsl_test_cases = [ (b"&a=b", [(b'a', b'b')]), (b"a=a+b&b=b+c", [(b'a', b'a b'), (b'b', b'b c')]), (b"a=1&a=2", [(b'a', b'1'), (b'a', b'2')]), +] + +parse_qsl_test_cases_semicolon = [ (";", []), (";;", []), (";a=b", [('a', 'b')]), @@ -44,6 +52,21 @@ parse_qsl_test_cases = [ (b"a=1;a=2", [(b'a', b'1'), (b'a', b'2')]), ] +parse_qsl_test_cases_legacy = [ + (b"a=1;a=2&a=3", [(b'a', b'1'), (b'a', b'2'), (b'a', b'3')]), + (b"a=1;b=2&c=3", [(b'a', b'1'), (b'b', b'2'), (b'c', b'3')]), + (b"a=1&b=2&c=3;", [(b'a', b'1'), (b'b', b'2'), (b'c', b'3')]), +] + +parse_qsl_test_cases_warn = [ + (";a=b", [(';a', 'b')]), + ("a=a+b;b=b+c", [('a', 'a b;b=b c')]), + (b";a=b", [(b';a', b'b')]), + (b"a=a+b;b=b+c", [(b'a', b'a b;b=b c')]), + ("a=1;a=2&a=3", [('a', '1;a=2'), ('a', '3')]), + (b"a=1;a=2&a=3", [(b'a', b'1;a=2'), (b'a', b'3')]), +] + # Each parse_qs testcase is a two-tuple that contains # a string with the query and a dictionary with the expected result. @@ -68,6 +91,9 @@ parse_qs_test_cases = [ (b"&a=b", {b'a': [b'b']}), (b"a=a+b&b=b+c", {b'a': [b'a b'], b'b': [b'b c']}), (b"a=1&a=2", {b'a': [b'1', b'2']}), +] + +parse_qs_test_cases_semicolon = [ (";", {}), (";;", {}), (";a=b", {'a': ['b']}), @@ -80,6 +106,24 @@ parse_qs_test_cases = [ (b"a=1;a=2", {b'a': [b'1', b'2']}), ] +parse_qs_test_cases_legacy = [ + ("a=1;a=2&a=3", {'a': ['1', '2', '3']}), + ("a=1;b=2&c=3", {'a': ['1'], 'b': ['2'], 'c': ['3']}), + ("a=1&b=2&c=3;", {'a': ['1'], 'b': ['2'], 'c': ['3']}), + (b"a=1;a=2&a=3", {b'a': [b'1', b'2', b'3']}), + (b"a=1;b=2&c=3", {b'a': [b'1'], b'b': [b'2'], b'c': [b'3']}), + (b"a=1&b=2&c=3;", {b'a': [b'1'], b'b': [b'2'], b'c': [b'3']}), +] + +parse_qs_test_cases_warn = [ + (";a=b", {';a': ['b']}), + ("a=a+b;b=b+c", {'a': ['a b;b=b c']}), + (b";a=b", {b';a': [b'b']}), + (b"a=a+b;b=b+c", {b'a':[ b'a b;b=b c']}), + ("a=1;a=2&a=3", {'a': ['1;a=2', '3']}), + (b"a=1;a=2&a=3", {b'a': [b'1;a=2', b'3']}), +] + class UrlParseTestCase(unittest.TestCase): def checkRoundtrips(self, url, parsed, split): @@ -152,6 +196,40 @@ class UrlParseTestCase(unittest.TestCase): self.assertEqual(result, expect_without_blanks, "Error parsing %r" % orig) + def test_qs_default_warn(self): + for orig, expect in parse_qs_test_cases_warn: + with self.subTest(orig=orig, expect=expect): + with catch_warnings(record=True) as w: + result = urllib.parse.parse_qs(orig, keep_blank_values=True) + self.assertEqual(result, expect, "Error parsing %r" % orig) + self.assertEqual(len(w), 1) + self.assertEqual(w[0].category, urllib.parse._QueryStringSeparatorWarning) + + def test_qsl_default_warn(self): + for orig, expect in parse_qsl_test_cases_warn: + with self.subTest(orig=orig, expect=expect): + with catch_warnings(record=True) as w: + result = urllib.parse.parse_qsl(orig, keep_blank_values=True) + self.assertEqual(result, expect, "Error parsing %r" % orig) + self.assertEqual(len(w), 1) + self.assertEqual(w[0].category, urllib.parse._QueryStringSeparatorWarning) + + def test_default_qs_no_warnings(self): + for orig, expect in parse_qs_test_cases: + with self.subTest(orig=orig, expect=expect): + with catch_warnings(record=True) as w: + result = urllib.parse.parse_qs(orig, keep_blank_values=True) + self.assertEqual(result, expect, "Error parsing %r" % orig) + self.assertEqual(len(w), 0) + + def test_default_qsl_no_warnings(self): + for orig, expect in parse_qsl_test_cases: + with self.subTest(orig=orig, expect=expect): + with catch_warnings(record=True) as w: + result = urllib.parse.parse_qsl(orig, keep_blank_values=True) + self.assertEqual(result, expect, "Error parsing %r" % orig) + self.assertEqual(len(w), 0) + def test_roundtrips(self): str_cases = [ ('file:///tmp/junk.txt', @@ -885,8 +963,151 @@ class UrlParseTestCase(unittest.TestCase): with self.assertRaises(ValueError): urllib.parse.parse_qs('&'.join(['a=a']*11), max_num_fields=10) with self.assertRaises(ValueError): - urllib.parse.parse_qs(';'.join(['a=a']*11), max_num_fields=10) + urllib.parse.parse_qs(';'.join(['a=a']*11), separator=';', max_num_fields=10) + with self.assertRaises(ValueError): + urllib.parse.parse_qs('SEP'.join(['a=a']*11), separator='SEP', max_num_fields=10) urllib.parse.parse_qs('&'.join(['a=a']*10), max_num_fields=10) + urllib.parse.parse_qs(';'.join(['a=a']*10), separator=';', max_num_fields=10) + urllib.parse.parse_qs('SEP'.join(['a=a']*10), separator='SEP', max_num_fields=10) + + def test_parse_qs_separator_bytes(self): + expected = {b'a': [b'1'], b'b': [b'2']} + + result = urllib.parse.parse_qs(b'a=1;b=2', separator=b';') + self.assertEqual(result, expected) + result = urllib.parse.parse_qs(b'a=1;b=2', separator=';') + self.assertEqual(result, expected) + result = urllib.parse.parse_qs('a=1;b=2', separator=';') + self.assertEqual(result, {'a': ['1'], 'b': ['2']}) + + @contextlib.contextmanager + def _qsl_sep_config(self, sep): + """Context for the given parse_qsl default separator configured in config file""" + old_filename = urllib.parse._QS_SEPARATOR_CONFIG_FILENAME + urllib.parse._default_qs_separator = None + try: + with tempfile.TemporaryDirectory() as tmpdirname: + filename = os.path.join(tmpdirname, 'conf.cfg') + with open(filename, 'w') as file: + file.write(f'[parse_qs]\n') + file.write(f'PYTHON_URLLIB_QS_SEPARATOR = {sep}') + urllib.parse._QS_SEPARATOR_CONFIG_FILENAME = filename + yield + finally: + urllib.parse._QS_SEPARATOR_CONFIG_FILENAME = old_filename + urllib.parse._default_qs_separator = None + + def test_parse_qs_separator_semicolon(self): + for orig, expect in parse_qs_test_cases_semicolon: + with self.subTest(orig=orig, expect=expect, method='arg'): + result = urllib.parse.parse_qs(orig, separator=';') + self.assertEqual(result, expect, "Error parsing %r" % orig) + with self.subTest(orig=orig, expect=expect, method='env'): + with EnvironmentVarGuard() as environ, catch_warnings(record=True) as w: + environ['PYTHON_URLLIB_QS_SEPARATOR'] = ';' + result = urllib.parse.parse_qs(orig) + self.assertEqual(result, expect, "Error parsing %r" % orig) + self.assertEqual(len(w), 0) + with self.subTest(orig=orig, expect=expect, method='conf'): + with self._qsl_sep_config(';'), catch_warnings(record=True) as w: + result = urllib.parse.parse_qs(orig) + self.assertEqual(result, expect, "Error parsing %r" % orig) + self.assertEqual(len(w), 0) + + def test_parse_qsl_separator_semicolon(self): + for orig, expect in parse_qsl_test_cases_semicolon: + with self.subTest(orig=orig, expect=expect, method='arg'): + result = urllib.parse.parse_qsl(orig, separator=';') + self.assertEqual(result, expect, "Error parsing %r" % orig) + with self.subTest(orig=orig, expect=expect, method='env'): + with EnvironmentVarGuard() as environ, catch_warnings(record=True) as w: + environ['PYTHON_URLLIB_QS_SEPARATOR'] = ';' + result = urllib.parse.parse_qsl(orig) + self.assertEqual(result, expect, "Error parsing %r" % orig) + self.assertEqual(len(w), 0) + with self.subTest(orig=orig, expect=expect, method='conf'): + with self._qsl_sep_config(';'), catch_warnings(record=True) as w: + result = urllib.parse.parse_qsl(orig) + self.assertEqual(result, expect, "Error parsing %r" % orig) + self.assertEqual(len(w), 0) + + def test_parse_qs_separator_legacy(self): + for orig, expect in parse_qs_test_cases_legacy: + with self.subTest(orig=orig, expect=expect, method='env'): + with EnvironmentVarGuard() as environ, catch_warnings(record=True) as w: + environ['PYTHON_URLLIB_QS_SEPARATOR'] = 'legacy' + result = urllib.parse.parse_qs(orig) + self.assertEqual(result, expect, "Error parsing %r" % orig) + self.assertEqual(len(w), 0) + with self.subTest(orig=orig, expect=expect, method='conf'): + with self._qsl_sep_config('legacy'), catch_warnings(record=True) as w: + result = urllib.parse.parse_qs(orig) + self.assertEqual(result, expect, "Error parsing %r" % orig) + self.assertEqual(len(w), 0) + + def test_parse_qsl_separator_legacy(self): + for orig, expect in parse_qsl_test_cases_legacy: + with self.subTest(orig=orig, expect=expect, method='env'): + with EnvironmentVarGuard() as environ, catch_warnings(record=True) as w: + environ['PYTHON_URLLIB_QS_SEPARATOR'] = 'legacy' + result = urllib.parse.parse_qsl(orig) + self.assertEqual(result, expect, "Error parsing %r" % orig) + self.assertEqual(len(w), 0) + with self.subTest(orig=orig, expect=expect, method='conf'): + with self._qsl_sep_config('legacy'), catch_warnings(record=True) as w: + result = urllib.parse.parse_qsl(orig) + self.assertEqual(result, expect, "Error parsing %r" % orig) + self.assertEqual(len(w), 0) + + def test_parse_qs_separator_bad_value_env_or_config(self): + for bad_sep in '', 'abc', 'safe', '&;', 'SEP': + with self.subTest(bad_sep, method='env'): + with EnvironmentVarGuard() as environ, catch_warnings(record=True) as w: + environ['PYTHON_URLLIB_QS_SEPARATOR'] = bad_sep + with self.assertRaises(ValueError): + urllib.parse.parse_qsl('a=1;b=2') + with self.subTest(bad_sep, method='conf'): + with self._qsl_sep_config('bad_sep'), catch_warnings(record=True) as w: + with self.assertRaises(ValueError): + urllib.parse.parse_qsl('a=1;b=2') + + def test_parse_qs_separator_bad_value_arg(self): + for bad_sep in True, {}, '': + with self.subTest(bad_sep): + with self.assertRaises(ValueError): + urllib.parse.parse_qsl('a=1;b=2', separator=bad_sep) + + def test_parse_qs_separator_num_fields(self): + for qs, sep in ( + ('a&b&c', '&'), + ('a;b;c', ';'), + ('a&b;c', 'legacy'), + ): + with self.subTest(qs=qs, sep=sep): + with EnvironmentVarGuard() as environ, catch_warnings(record=True) as w: + if sep != 'legacy': + with self.assertRaises(ValueError): + urllib.parse.parse_qsl(qs, separator=sep, max_num_fields=2) + if sep: + environ['PYTHON_URLLIB_QS_SEPARATOR'] = sep + with self.assertRaises(ValueError): + urllib.parse.parse_qsl(qs, max_num_fields=2) + + def test_parse_qs_separator_priority(self): + # env variable trumps config file + with self._qsl_sep_config('~'), EnvironmentVarGuard() as environ: + environ['PYTHON_URLLIB_QS_SEPARATOR'] = '!' + result = urllib.parse.parse_qs('a=1!b=2~c=3') + self.assertEqual(result, {'a': ['1'], 'b': ['2~c=3']}) + # argument trumps config file + with self._qsl_sep_config('~'): + result = urllib.parse.parse_qs('a=1$b=2~c=3', separator='$') + self.assertEqual(result, {'a': ['1'], 'b': ['2~c=3']}) + # argument trumps env variable + with EnvironmentVarGuard() as environ: + environ['PYTHON_URLLIB_QS_SEPARATOR'] = '~' + result = urllib.parse.parse_qs('a=1$b=2~c=3', separator='$') + self.assertEqual(result, {'a': ['1'], 'b': ['2~c=3']}) def test_urlencode_sequences(self): # Other tests incidentally urlencode things; test non-covered cases: diff --git a/Lib/urllib/parse.py b/Lib/urllib/parse.py index fa8827a9fa7..57b8fcf8bbd 100644 --- a/Lib/urllib/parse.py +++ b/Lib/urllib/parse.py @@ -28,6 +28,7 @@ test_urlparse.py provides a good indicator of parsing behavior. """ import re +import os import sys import collections @@ -644,7 +645,8 @@ def unquote(string, encoding='utf-8', errors='replace'): def parse_qs(qs, keep_blank_values=False, strict_parsing=False, - encoding='utf-8', errors='replace', max_num_fields=None): + encoding='utf-8', errors='replace', max_num_fields=None, + separator=None): """Parse a query given as a string argument. Arguments: @@ -673,7 +675,8 @@ def parse_qs(qs, keep_blank_values=False, strict_parsing=False, parsed_result = {} pairs = parse_qsl(qs, keep_blank_values, strict_parsing, encoding=encoding, errors=errors, - max_num_fields=max_num_fields) + max_num_fields=max_num_fields, + separator=separator) for name, value in pairs: if name in parsed_result: parsed_result[name].append(value) @@ -681,9 +684,16 @@ def parse_qs(qs, keep_blank_values=False, strict_parsing=False, parsed_result[name] = [value] return parsed_result +class _QueryStringSeparatorWarning(RuntimeWarning): + """Warning for using default `separator` in parse_qs or parse_qsl""" + +# The default "separator" for parse_qsl can be specified in a config file. +# It's cached after first read. +_QS_SEPARATOR_CONFIG_FILENAME = '/etc/python/urllib.cfg' +_default_qs_separator = None def parse_qsl(qs, keep_blank_values=False, strict_parsing=False, - encoding='utf-8', errors='replace', max_num_fields=None): + encoding='utf-8', errors='replace', max_num_fields=None, separator=None): """Parse a query given as a string argument. Arguments: @@ -710,15 +720,77 @@ def parse_qsl(qs, keep_blank_values=False, strict_parsing=False, """ qs, _coerce_result = _coerce_args(qs) + if isinstance(separator, bytes): + separator = separator.decode('ascii') + + if (not separator or (not isinstance(separator, (str, bytes)))) and separator is not None: + raise ValueError("Separator must be of type string or bytes.") + + # Used when both "&" and ";" act as separators. (Need a non-string value.) + _legacy = object() + + if separator is None: + global _default_qs_separator + separator = _default_qs_separator + envvar_name = 'PYTHON_URLLIB_QS_SEPARATOR' + if separator is None: + # Set default separator from environment variable + separator = os.environ.get(envvar_name) + config_source = 'environment variable' + if separator is None: + # Set default separator from the configuration file + try: + file = open(_QS_SEPARATOR_CONFIG_FILENAME) + except FileNotFoundError: + pass + else: + with file: + import configparser + config = configparser.ConfigParser( + interpolation=None, + comment_prefixes=('#', ), + ) + config.read_file(file) + separator = config.get('parse_qs', envvar_name, fallback=None) + _default_qs_separator = separator + config_source = _QS_SEPARATOR_CONFIG_FILENAME + if separator is None: + # The default is '&', but warn if not specified explicitly + if ';' in qs: + from warnings import warn + warn("The default separator of urllib.parse.parse_qsl and " + + "parse_qs was changed to '&' to avoid a web cache " + + "poisoning issue (CVE-2021-23336). " + + "By default, semicolons no longer act as query field " + + "separators. " + + "See https://access.redhat.com/articles/5860431 for " + + "more details.", + _QueryStringSeparatorWarning, stacklevel=2) + separator = '&' + elif separator == 'legacy': + separator = _legacy + elif len(separator) != 1: + raise ValueError( + f'{envvar_name} (from {config_source}) must contain ' + + '1 character, or "legacy". See ' + + 'https://access.redhat.com/articles/5860431 for more details.' + ) + # If max_num_fields is defined then check that the number of fields # is less than max_num_fields. This prevents a memory exhaustion DOS # attack via post bodies with many fields. if max_num_fields is not None: - num_fields = 1 + qs.count('&') + qs.count(';') + if separator is _legacy: + num_fields = 1 + qs.count('&') + qs.count(';') + else: + num_fields = 1 + qs.count(separator) if max_num_fields < num_fields: raise ValueError('Max number of fields exceeded') - pairs = [s2 for s1 in qs.split('&') for s2 in s1.split(';')] + if separator is _legacy: + pairs = [s2 for s1 in qs.split('&') for s2 in s1.split(';')] + else: + pairs = [s1 for s1 in qs.split(separator)] r = [] for name_value in pairs: if not name_value and not strict_parsing: diff --git a/Misc/NEWS.d/next/Security/2021-02-14-15-59-16.bpo-42967.YApqDS.rst b/Misc/NEWS.d/next/Security/2021-02-14-15-59-16.bpo-42967.YApqDS.rst new file mode 100644 index 00000000000..bc82c963067 --- /dev/null +++ b/Misc/NEWS.d/next/Security/2021-02-14-15-59-16.bpo-42967.YApqDS.rst @@ -0,0 +1 @@ +Make it possible to fix web cache poisoning vulnerability by allowing the user to choose a custom separator query args.