From 8623260fb0949d368376a128bee2189ec0a67ae5 Mon Sep 17 00:00:00 2001
From: nkrapp <nico.krapp@suse.com>
Date: Mon, 22 Jul 2024 09:43:08 +0200
Subject: [PATCH] [PATCH] [4.2.x] Fixed CVE-2024-38875 -- Mitigated potential
 DoS in urlize and urlizetrunc template filters.

---
 django/utils/html.py           | 72 +++++++++++++++++++++++++++++-----
 tests/utils_tests/test_html.py | 21 ++++++----
 2 files changed, 75 insertions(+), 17 deletions(-)

diff --git a/django/utils/html.py b/django/utils/html.py
index 7a33d5f68d..1dbe39ccd1 100644
--- a/django/utils/html.py
+++ b/django/utils/html.py
@@ -1,5 +1,6 @@
 """HTML utilities suitable for global use."""
 
+import html
 import json
 import re
 from html.parser import HTMLParser
@@ -235,6 +235,16 @@ def smart_urlquote(url):
     return urlunsplit((scheme, netloc, path, query, fragment))
 
 
+class CountsDict(dict):
+    def __init__(self, *args, word, **kwargs):
+        super().__init__(*args, *kwargs)
+        self.word = word
+
+    def __missing__(self, key):
+        self[key] = self.word.count(key)
+        return self[key]
+
+
 @keep_lazy_text
 def urlize(text, trim_url_limit=None, nofollow=False, autoescape=False):
     """
@@ -259,6 +269,15 @@ def urlize(text, trim_url_limit=None, nofollow=False, autoescape=False):
             return x
         return '%s…' % x[:max(0, limit - 1)]
 
+    def wrapping_punctuation_openings():
+        return "".join(dict(WRAPPING_PUNCTUATION).keys())
+
+    def trailing_punctuation_chars_no_semicolon():
+        return TRAILING_PUNCTUATION_CHARS.replace(";", "")
+
+    def trailing_punctuation_chars_has_semicolon():
+        return ";" in TRAILING_PUNCTUATION_CHARS
+
     def unescape(text):
         """
         If input URL is HTML-escaped, unescape it so that it can be safely fed
@@ -273,21 +292,53 @@ def urlize(text, trim_url_limit=None, nofollow=False, autoescape=False):
         Trim trailing and wrapping punctuation from `middle`. Return the items
         of the new state.
         """
+        # Strip all opening wrapping punctuation.
+        middle = word.lstrip(wrapping_punctuation_openings())
+        lead = word[: len(word) - len(middle)]
+        trail = ""
+
         # Continue trimming until middle remains unchanged.
         trimmed_something = True
-        while trimmed_something:
+        counts = CountsDict(word=middle)
+        while trimmed_something and middle:
             trimmed_something = False
             # Trim wrapping punctuation.
             for opening, closing in WRAPPING_PUNCTUATION:
-                if middle.startswith(opening):
-                    middle = middle[len(opening):]
-                    lead += opening
-                    trimmed_something = True
-                # Keep parentheses at the end only if they're balanced.
-                if (middle.endswith(closing) and
-                        middle.count(closing) == middle.count(opening) + 1):
-                    middle = middle[:-len(closing)]
-                    trail = closing + trail
+                if counts[opening] < counts[closing]:
+                    rstripped = middle.rstrip(closing)
+                    if rstripped != middle:
+                        strip = counts[closing] - counts[opening]
+                        trail = middle[-strip:]
+                        middle = middle[:-strip]
+                        trimmed_something = True
+                        counts[closing] -= strip
+
+            rstripped = middle.rstrip(trailing_punctuation_chars_no_semicolon())
+            if rstripped != middle:
+                trail = middle[len(rstripped) :] + trail
+                middle = rstripped
+                trimmed_something = True
+
+            if trailing_punctuation_chars_has_semicolon() and middle.endswith(";"):
+                # Only strip if not part of an HTML entity.
+                amp = middle.rfind("&")
+                if amp == -1:
+                    can_strip = True
+                else:
+                    potential_entity = middle[amp:]
+                    escaped = html.unescape(potential_entity)
+                    can_strip = (escaped == potential_entity) or escaped.endswith(";")
+
+                if can_strip:
+                    rstripped = middle.rstrip(";")
+                    amount_stripped = len(middle) - len(rstripped)
+                    if amp > -1 and amount_stripped > 1:
+                        # Leave a trailing semicolon as might be an entity.
+                        trail = middle[len(rstripped) + 1 :] + trail
+                        middle = rstripped + ";"
+                    else:
+                        trail = middle[len(rstripped) :] + trail
+                        middle = rstripped
                     trimmed_something = True
             # Trim trailing punctuation (after trimming wrapping punctuation,
             # as encoded entities contain ';'). Unescape entites to avoid
diff --git a/tests/utils_tests/test_html.py b/tests/utils_tests/test_html.py
index 5cc2d9b95d..dc89009b63 100644
--- a/tests/utils_tests/test_html.py
+++ b/tests/utils_tests/test_html.py
@@ -260,13 +260,20 @@ class TestUtilsHtml(SimpleTestCase):
 
     def test_urlize_unchanged_inputs(self):
         tests = (
-            ('a' + '@a' * 50000) + 'a',  # simple_email_re catastrophic test
-            ('a' + '.' * 1000000) + 'a',  # trailing_punctuation catastrophic test
-            'foo@',
-            '@foo.com',
-            'foo@.example.com',
-            'foo@localhost',
-            'foo@localhost.',
+            ("a" + "@a" * 50000) + "a",  # simple_email_re catastrophic test
+            ("a" + "." * 1000000) + "a",  # trailing_punctuation catastrophic test
+            "foo@",
+            "@foo.com",
+            "foo@.example.com",
+            "foo@localhost",
+            "foo@localhost.",
+            # trim_punctuation catastrophic tests
+            "(" * 100_000 + ":" + ")" * 100_000,
+            "(" * 100_000 + "&:" + ")" * 100_000,
+            "([" * 100_000 + ":" + "])" * 100_000,
+            "[(" * 100_000 + ":" + ")]" * 100_000,
+            "([[" * 100_000 + ":" + "]])" * 100_000,
+            "&:" + ";" * 100_000,
         )
         for value in tests:
             with self.subTest(value=value):
-- 
2.45.2