[CVE-2023-27043] gh-102988: Reject malformed addresses in email.parseaddr() (#111116)

Detect email address parsing errors and return empty tuple to
indicate the parsing error (old API). Add an optional 'strict'
parameter to getaddresses() and parseaddr() functions. Patch by
Thomas Dwyer.

Co-Authored-By: Thomas Dwyer <github@tomd.tel>

5 files changed, 357 insertions, 21 deletions

diff --git a/Doc/library/email.utils.rst b/Doc/library/email.utils.rst
index 345b640..d693a9b 100644
--- a/Doc/library/email.utils.rst
+++ b/Doc/library/email.utils.rst
@@ -58,13 +58,18 @@ of the new API.
    begins with angle brackets, they are stripped off.
 
 
-.. function:: parseaddr(address)
+.. function:: parseaddr(address, *, strict=True)
 
    Parse address -- which should be the value of some address-containing field such
    as :mailheader:`To` or :mailheader:`Cc` -- into its constituent *realname* and
    *email address* parts.  Returns a tuple of that information, unless the parse
    fails, in which case a 2-tuple of ``('', '')`` is returned.
 
+   If *strict* is true, use a strict parser which rejects malformed inputs.
+
+   .. versionchanged:: 3.13
+      Add *strict* optional parameter and reject malformed inputs by default.
+
 
 .. function:: formataddr(pair, charset='utf-8')
 
@@ -82,12 +87,15 @@ of the new API.
       Added the *charset* option.
 
 
-.. function:: getaddresses(fieldvalues)
+.. function:: getaddresses(fieldvalues, *, strict=True)
 
    This method returns a list of 2-tuples of the form returned by ``parseaddr()``.
    *fieldvalues* is a sequence of header field values as might be returned by
-   :meth:`Message.get_all <email.message.Message.get_all>`.  Here's a simple
-   example that gets all the recipients of a message::
+   :meth:`Message.get_all <email.message.Message.get_all>`.
+
+   If *strict* is true, use a strict parser which rejects malformed inputs.
+
+   Here's a simple example that gets all the recipients of a message::
 
       from email.utils import getaddresses
 
@@ -97,6 +105,9 @@ of the new API.
       resent_ccs = msg.get_all('resent-cc', [])
       all_recipients = getaddresses(tos + ccs + resent_tos + resent_ccs)
 
+   .. versionchanged:: 3.13
+      Add *strict* optional parameter and reject malformed inputs by default.
+
 
 .. function:: parsedate(date)
 
diff --git a/Doc/whatsnew/3.13.rst b/Doc/whatsnew/3.13.rst
index e222578..4f96439 100644
--- a/Doc/whatsnew/3.13.rst
+++ b/Doc/whatsnew/3.13.rst
@@ -199,6 +199,19 @@ doctest
   :attr:`doctest.TestResults.skipped` attributes.
   (Contributed by Victor Stinner in :gh:`108794`.)
 
+email
+-----
+
+* :func:`email.utils.getaddresses` and :func:`email.utils.parseaddr` now return
+  ``('', '')`` 2-tuples in more situations where invalid email addresses are
+  encountered instead of potentially inaccurate values. Add optional *strict*
+  parameter to these two functions: use ``strict=False`` to get the old
+  behavior, accept malformed inputs.
+  ``getattr(email.utils, 'supports_strict_parsing', False)`` can be use to
+  check if the *strict* paramater is available.
+  (Contributed by Thomas Dwyer and Victor Stinner for :gh:`102988` to improve
+  the CVE-2023-27043 fix.)
+
 glob
 ----
 
diff --git a/Lib/email/utils.py b/Lib/email/utils.py
index 9175f2f..103cef6 100644
--- a/Lib/email/utils.py
+++ b/Lib/email/utils.py
@@ -43,6 +43,7 @@ TICK = "'"
 specialsre = re.compile(r'[][\\()<>@,:;".]')
 escapesre = re.compile(r'[\\"]')
 
+
 def _has_surrogates(s):
     """Return True if s may contain surrogate-escaped binary data."""
     # This check is based on the fact that unless there are surrogates, utf8
@@ -103,12 +104,127 @@ def formataddr(pair, charset='utf-8'):
     return address
 
 
+def _iter_escaped_chars(addr):
+    pos = 0
+    escape = False
+    for pos, ch in enumerate(addr):
+        if escape:
+            yield (pos, '\\' + ch)
+            escape = False
+        elif ch == '\\':
+            escape = True
+        else:
+            yield (pos, ch)
+    if escape:
+        yield (pos, '\\')
+
+
+def _strip_quoted_realnames(addr):
+    """Strip real names between quotes."""
+    if '"' not in addr:
+        # Fast path
+        return addr
+
+    start = 0
+    open_pos = None
+    result = []
+    for pos, ch in _iter_escaped_chars(addr):
+        if ch == '"':
+            if open_pos is None:
+                open_pos = pos
+            else:
+                if start != open_pos:
+                    result.append(addr[start:open_pos])
+                start = pos + 1
+                open_pos = None
+
+    if start < len(addr):
+        result.append(addr[start:])
+
+    return ''.join(result)
 
-def getaddresses(fieldvalues):
-    """Return a list of (REALNAME, EMAIL) for each fieldvalue."""
-    all = COMMASPACE.join(str(v) for v in fieldvalues)
-    a = _AddressList(all)
-    return a.addresslist
+
+supports_strict_parsing = True
+
+def getaddresses(fieldvalues, *, strict=True):
+    """Return a list of (REALNAME, EMAIL) or ('','') for each fieldvalue.
+
+    When parsing fails for a fieldvalue, a 2-tuple of ('', '') is returned in
+    its place.
+
+    If strict is true, use a strict parser which rejects malformed inputs.
+    """
+
+    # If strict is true, if the resulting list of parsed addresses is greater
+    # than the number of fieldvalues in the input list, a parsing error has
+    # occurred and consequently a list containing a single empty 2-tuple [('',
+    # '')] is returned in its place. This is done to avoid invalid output.
+    #
+    # Malformed input: getaddresses(['alice@example.com <bob@example.com>'])
+    # Invalid output: [('', 'alice@example.com'), ('', 'bob@example.com')]
+    # Safe output: [('', '')]
+
+    if not strict:
+        all = COMMASPACE.join(str(v) for v in fieldvalues)
+        a = _AddressList(all)
+        return a.addresslist
+
+    fieldvalues = [str(v) for v in fieldvalues]
+    fieldvalues = _pre_parse_validation(fieldvalues)
+    addr = COMMASPACE.join(fieldvalues)
+    a = _AddressList(addr)
+    result = _post_parse_validation(a.addresslist)
+
+    # Treat output as invalid if the number of addresses is not equal to the
+    # expected number of addresses.
+    n = 0
+    for v in fieldvalues:
+        # When a comma is used in the Real Name part it is not a deliminator.
+        # So strip those out before counting the commas.
+        v = _strip_quoted_realnames(v)
+        # Expected number of addresses: 1 + number of commas
+        n += 1 + v.count(',')
+    if len(result) != n:
+        return [('', '')]
+
+    return result
+
+
+def _check_parenthesis(addr):
+    # Ignore parenthesis in quoted real names.
+    addr = _strip_quoted_realnames(addr)
+
+    opens = 0
+    for pos, ch in _iter_escaped_chars(addr):
+        if ch == '(':
+            opens += 1
+        elif ch == ')':
+            opens -= 1
+            if opens < 0:
+                return False
+    return (opens == 0)
+
+
+def _pre_parse_validation(email_header_fields):
+    accepted_values = []
+    for v in email_header_fields:
+        if not _check_parenthesis(v):
+            v = "('', '')"
+        accepted_values.append(v)
+
+    return accepted_values
+
+
+def _post_parse_validation(parsed_email_header_tuples):
+    accepted_values = []
+    # The parser would have parsed a correctly formatted domain-literal
+    # The existence of an [ after parsing indicates a parsing failure
+    for v in parsed_email_header_tuples:
+        if '[' in v[1]:
+            v = ('', '')
+        accepted_values.append(v)
+
+    return accepted_values
 
 
 def _format_timetuple_and_zone(timetuple, zone):
@@ -207,16 +323,33 @@ def parsedate_to_datetime(data):
             tzinfo=datetime.timezone(datetime.timedelta(seconds=tz)))
 
 
-def parseaddr(addr):
+def parseaddr(addr, *, strict=True):
     """
     Parse addr into its constituent realname and email address parts.
 
     Return a tuple of realname and email address, unless the parse fails, in
     which case return a 2-tuple of ('', '').
+
+    If strict is True, use a strict parser which rejects malformed inputs.
     """
-    addrs = _AddressList(addr).addresslist
-    if not addrs:
-        return '', ''
+    if not strict:
+        addrs = _AddressList(addr).addresslist
+        if not addrs:
+            return ('', '')
+        return addrs[0]
+
+    if isinstance(addr, list):
+        addr = addr[0]
+
+    if not isinstance(addr, str):
+        return ('', '')
+
+    addr = _pre_parse_validation([addr])[0]
+    addrs = _post_parse_validation(_AddressList(addr).addresslist)
+
+    if not addrs or len(addrs) > 1:
+        return ('', '')
+
     return addrs[0]
 
 
diff --git a/Lib/test/test_email/test_email.py b/Lib/test/test_email/test_email.py
index 512464f..39d4ace 100644
--- a/Lib/test/test_email/test_email.py
+++ b/Lib/test/test_email/test_email.py
@@ -16,6 +16,7 @@ from unittest.mock import patch
 
 import email
 import email.policy
+import email.utils
 
 from email.charset import Charset
 from email.generator import Generator, DecodedGenerator, BytesGenerator
@@ -3337,15 +3338,137 @@ Foo
             ],
         )
 
+    def test_parsing_errors(self):
+        """Test for parsing errors from CVE-2023-27043 and CVE-2019-16056"""
+        alice = 'alice@example.org'
+        bob = 'bob@example.com'
+        empty = ('', '')
+
+        # Test utils.getaddresses() and utils.parseaddr() on malformed email
+        # addresses: default behavior (strict=True) rejects malformed address,
+        # and strict=False which tolerates malformed address.
+        for invalid_separator, expected_non_strict in (
+            ('(', [(f'<{bob}>', alice)]),
+            (')', [('', alice), empty, ('', bob)]),
+            ('<', [('', alice), empty, ('', bob), empty]),
+            ('>', [('', alice), empty, ('', bob)]),
+            ('[', [('', f'{alice}[<{bob}>]')]),
+            (']', [('', alice), empty, ('', bob)]),
+            ('@', [empty, empty, ('', bob)]),
+            (';', [('', alice), empty, ('', bob)]),
+            (':', [('', alice), ('', bob)]),
+            ('.', [('', alice + '.'), ('', bob)]),
+            ('"', [('', alice), ('', f'<{bob}>')]),
+        ):
+            address = f'{alice}{invalid_separator}<{bob}>'
+            with self.subTest(address=address):
+                self.assertEqual(utils.getaddresses([address]),
+                                 [empty])
+                self.assertEqual(utils.getaddresses([address], strict=False),
+                                 expected_non_strict)
+
+                self.assertEqual(utils.parseaddr([address]),
+                                 empty)
+                self.assertEqual(utils.parseaddr([address], strict=False),
+                                 ('', address))
+
+        # Comma (',') is treated differently depending on strict parameter.
+        # Comma without quotes.
+        address = f'{alice},<{bob}>'
+        self.assertEqual(utils.getaddresses([address]),
+                         [('', alice), ('', bob)])
+        self.assertEqual(utils.getaddresses([address], strict=False),
+                         [('', alice), ('', bob)])
+        self.assertEqual(utils.parseaddr([address]),
+                         empty)
+        self.assertEqual(utils.parseaddr([address], strict=False),
+                         ('', address))
+
+        # Real name between quotes containing comma.
+        address = '"Alice, alice@example.org" <bob@example.com>'
+        expected_strict = ('Alice, alice@example.org', 'bob@example.com')
+        self.assertEqual(utils.getaddresses([address]), [expected_strict])
+        self.assertEqual(utils.getaddresses([address], strict=False), [expected_strict])
+        self.assertEqual(utils.parseaddr([address]), expected_strict)
+        self.assertEqual(utils.parseaddr([address], strict=False),
+                         ('', address))
+
+        # Valid parenthesis in comments.
+        address = 'alice@example.org (Alice)'
+        expected_strict = ('Alice', 'alice@example.org')
+        self.assertEqual(utils.getaddresses([address]), [expected_strict])
+        self.assertEqual(utils.getaddresses([address], strict=False), [expected_strict])
+        self.assertEqual(utils.parseaddr([address]), expected_strict)
+        self.assertEqual(utils.parseaddr([address], strict=False),
+                         ('', address))
+
+        # Invalid parenthesis in comments.
+        address = 'alice@example.org )Alice('
+        self.assertEqual(utils.getaddresses([address]), [empty])
+        self.assertEqual(utils.getaddresses([address], strict=False),
+                         [('', 'alice@example.org'), ('', ''), ('', 'Alice')])
+        self.assertEqual(utils.parseaddr([address]), empty)
+        self.assertEqual(utils.parseaddr([address], strict=False),
+                         ('', address))
+
+        # Two addresses with quotes separated by comma.
+        address = '"Jane Doe" <jane@example.net>, "John Doe" <john@example.net>'
+        self.assertEqual(utils.getaddresses([address]),
+                         [('Jane Doe', 'jane@example.net'),
+                          ('John Doe', 'john@example.net')])
+        self.assertEqual(utils.getaddresses([address], strict=False),
+                         [('Jane Doe', 'jane@example.net'),
+                          ('John Doe', 'john@example.net')])
+        self.assertEqual(utils.parseaddr([address]), empty)
+        self.assertEqual(utils.parseaddr([address], strict=False),
+                         ('', address))
+
+        # Test email.utils.supports_strict_parsing attribute
+        self.assertEqual(email.utils.supports_strict_parsing, True)
+
     def test_getaddresses_nasty(self):
-        eq = self.assertEqual
-        eq(utils.getaddresses(['foo: ;']), [('', '')])
-        eq(utils.getaddresses(
-           ['[]*-- =~$']),
-           [('', ''), ('', ''), ('', '*--')])
-        eq(utils.getaddresses(
-           ['foo: ;', '"Jason R. Mastaler" <jason@dom.ain>']),
-           [('', ''), ('Jason R. Mastaler', 'jason@dom.ain')])
+        for addresses, expected in (
+            (['"Sürname, Firstname" <to@example.com>'],
+             [('Sürname, Firstname', 'to@example.com')]),
+
+            (['foo: ;'],
+             [('', '')]),
+
+            (['foo: ;', '"Jason R. Mastaler" <jason@dom.ain>'],
+             [('', ''), ('Jason R. Mastaler', 'jason@dom.ain')]),
+
+            ([r'Pete(A nice \) chap) <pete(his account)@silly.test(his host)>'],
+             [('Pete (A nice ) chap his account his host)', 'pete@silly.test')]),
+
+            (['(Empty list)(start)Undisclosed recipients  :(nobody(I know))'],
+             [('', '')]),
+
+            (['Mary <@machine.tld:mary@example.net>, , jdoe@test   . example'],
+             [('Mary', 'mary@example.net'), ('', ''), ('', 'jdoe@test.example')]),
+
+            (['John Doe <jdoe@machine(comment).  example>'],
+             [('John Doe (comment)', 'jdoe@machine.example')]),
+
+            (['"Mary Smith: Personal Account" <smith@home.example>'],
+             [('Mary Smith: Personal Account', 'smith@home.example')]),
+
+            (['Undisclosed recipients:;'],
+             [('', '')]),
+
+            ([r'<boss@nil.test>, "Giant; \"Big\" Box" <bob@example.net>'],
+             [('', 'boss@nil.test'), ('Giant; "Big" Box', 'bob@example.net')]),
+        ):
+            with self.subTest(addresses=addresses):
+                self.assertEqual(utils.getaddresses(addresses),
+                                 expected)
+                self.assertEqual(utils.getaddresses(addresses, strict=False),
+                                 expected)
+
+        addresses = ['[]*-- =~$']
+        self.assertEqual(utils.getaddresses(addresses),
+                         [('', '')])
+        self.assertEqual(utils.getaddresses(addresses, strict=False),
+                         [('', ''), ('', ''), ('', '*--')])
 
     def test_getaddresses_embedded_comment(self):
         """Test proper handling of a nested comment"""
@@ -3536,6 +3659,54 @@ multipart/report
                 m = cls(*constructor, policy=email.policy.default)
                 self.assertIs(m.policy, email.policy.default)
 
+    def test_iter_escaped_chars(self):
+        self.assertEqual(list(utils._iter_escaped_chars(r'a\\b\"c\\"d')),
+                         [(0, 'a'),
+                          (2, '\\\\'),
+                          (3, 'b'),
+                          (5, '\\"'),
+                          (6, 'c'),
+                          (8, '\\\\'),
+                          (9, '"'),
+                          (10, 'd')])
+        self.assertEqual(list(utils._iter_escaped_chars('a\\')),
+                         [(0, 'a'), (1, '\\')])
+
+    def test_strip_quoted_realnames(self):
+        def check(addr, expected):
+            self.assertEqual(utils._strip_quoted_realnames(addr), expected)
+
+        check('"Jane Doe" <jane@example.net>, "John Doe" <john@example.net>',
+              ' <jane@example.net>,  <john@example.net>')
+        check(r'"Jane \"Doe\"." <jane@example.net>',
+              ' <jane@example.net>')
+
+        # special cases
+        check(r'before"name"after', 'beforeafter')
+        check(r'before"name"', 'before')
+        check(r'b"name"', 'b')  # single char
+        check(r'"name"after', 'after')
+        check(r'"name"a', 'a')  # single char
+        check(r'"name"', '')
+
+        # no change
+        for addr in (
+            'Jane Doe <jane@example.net>, John Doe <john@example.net>',
+            'lone " quote',
+        ):
+            self.assertEqual(utils._strip_quoted_realnames(addr), addr)
+
+
+    def test_check_parenthesis(self):
+        addr = 'alice@example.net'
+        self.assertTrue(utils._check_parenthesis(f'{addr} (Alice)'))
+        self.assertFalse(utils._check_parenthesis(f'{addr} )Alice('))
+        self.assertFalse(utils._check_parenthesis(f'{addr} (Alice))'))
+        self.assertFalse(utils._check_parenthesis(f'{addr} ((Alice)'))
+
+        # Ignore real name between quotes
+        self.assertTrue(utils._check_parenthesis(f'")Alice((" {addr}'))
+
 
 # Test the iterator/generators
 class TestIterators(TestEmailBase):
diff --git a/Misc/NEWS.d/next/Library/2023-10-20-15-28-08.gh-issue-102988.dStNO7.rst b/Misc/NEWS.d/next/Library/2023-10-20-15-28-08.gh-issue-102988.dStNO7.rst
new file mode 100644
index 0000000..3d0e9e4
--- /dev/null
+++ b/Misc/NEWS.d/next/Library/2023-10-20-15-28-08.gh-issue-102988.dStNO7.rst
@@ -0,0 +1,8 @@
+:func:`email.utils.getaddresses` and :func:`email.utils.parseaddr` now
+return ``('', '')`` 2-tuples in more situations where invalid email
+addresses are encountered instead of potentially inaccurate values. Add
+optional *strict* parameter to these two functions: use ``strict=False`` to
+get the old behavior, accept malformed inputs.
+``getattr(email.utils, 'supports_strict_parsing', False)`` can be use to check
+if the *strict* paramater is available. Patch by Thomas Dwyer and Victor
+Stinner to improve the CVE-2023-27043 fix.