[3.2.x] Fixed CVE-2023-43665 -- Mitigated potential DoS in django.utils.text.Truncator when truncating HTML text.

nessita · nessita · commit ccdade1a0262 · 2023-10-04T09:41:12.000-03:00
Thanks Wenchao Li of Alibaba Group for the report.
diff --git a/django/utils/text.py b/django/utils/text.py
@@ -60,7 +60,14 @@ def _generator():
 class Truncator(SimpleLazyObject):
     """
     An object used to truncate text, either by characters or words.
+
+    When truncating HTML text (either chars or words), input will be limited to
+    at most `MAX_LENGTH_HTML` characters.
     """
+
+    # 5 million characters are approximately 4000 text pages or 3 web pages.
+    MAX_LENGTH_HTML = 5_000_000
+
     def __init__(self, text):
         super().__init__(lambda: str(text))
 
@@ -157,6 +164,11 @@ def _truncate_html(self, length, truncate, text, truncate_len, words):
         if words and length <= 0:
             return ''
 
+        size_limited = False
+        if len(text) > self.MAX_LENGTH_HTML:
+            text = text[: self.MAX_LENGTH_HTML]
+            size_limited = True
+
         html4_singlets = (
             'br', 'col', 'link', 'base', 'img',
             'param', 'area', 'hr', 'input'
@@ -206,10 +218,14 @@ def _truncate_html(self, length, truncate, text, truncate_len, words):
                 # Add it to the start of the open tags list
                 open_tags.insert(0, tagname)
 
+        truncate_text = self.add_truncation_text("", truncate)
+
         if current_len <= length:
+            if size_limited and truncate_text:
+                text += truncate_text
             return text
+
         out = text[:end_text_pos]
-        truncate_text = self.add_truncation_text('', truncate)
         if truncate_text:
             out += truncate_text
         # Close any tags still open
diff --git a/docs/ref/templates/builtins.txt b/docs/ref/templates/builtins.txt
@@ -2348,6 +2348,16 @@ If ``value`` is ``"<p>Joel is a slug</p>"``, the output will be
 
 Newlines in the HTML content will be preserved.
 
+.. admonition:: Size of input string
+
+    Processing large, potentially malformed HTML strings can be
+    resource-intensive and impact service performance. ``truncatechars_html``
+    limits input to the first five million characters.
+
+.. versionchanged:: 3.2.22
+
+    In older versions, strings over five million characters were processed.
+
 .. templatefilter:: truncatewords
 
 ``truncatewords``
@@ -2386,6 +2396,16 @@ If ``value`` is ``"<p>Joel is a slug</p>"``, the output will be
 
 Newlines in the HTML content will be preserved.
 
+.. admonition:: Size of input string
+
+    Processing large, potentially malformed HTML strings can be
+    resource-intensive and impact service performance. ``truncatewords_html``
+    limits input to the first five million characters.
+
+.. versionchanged:: 3.2.22
+
+    In older versions, strings over five million characters were processed.
+
 .. templatefilter:: unordered_list
 
 ``unordered_list``
diff --git a/docs/releases/3.2.22.txt b/docs/releases/3.2.22.txt
@@ -6,4 +6,20 @@ Django 3.2.22 release notes
 
 Django 3.2.22 fixes a security issue with severity "moderate" in 3.2.21.
 
-...
+CVE-2023-43665: Denial-of-service possibility in ``django.utils.text.Truncator``
+================================================================================
+
+Following the fix for :cve:`2019-14232`, the regular expressions used in the
+implementation of ``django.utils.text.Truncator``'s ``chars()`` and ``words()``
+methods (with ``html=True``) were revised and improved. However, these regular
+expressions still exhibited linear backtracking complexity, so when given a
+very long, potentially malformed HTML input, the evaluation would still be
+slow, leading to a potential denial of service vulnerability.
+
+The ``chars()`` and ``words()`` methods are used to implement the
+:tfilter:`truncatechars_html` and :tfilter:`truncatewords_html` template
+filters, which were thus also vulnerable.
+
+The input processed by ``Truncator``, when operating in HTML mode, has been
+limited to the first five million characters in order to avoid potential
+performance and memory issues.
diff --git a/tests/utils_tests/test_text.py b/tests/utils_tests/test_text.py
@@ -1,5 +1,6 @@
 import json
 import sys
+from unittest.mock import patch
 
 from django.core.exceptions import SuspiciousFileOperation
 from django.test import SimpleTestCase, ignore_warnings
@@ -90,11 +91,17 @@ def test_truncate_chars(self):
         # lazy strings are handled correctly
         self.assertEqual(text.Truncator(lazystr('The quick brown fox')).chars(10), 'The quick…')
 
-    def test_truncate_chars_html(self):
+    @patch("django.utils.text.Truncator.MAX_LENGTH_HTML", 10_000)
+    def test_truncate_chars_html_size_limit(self):
+        max_len = text.Truncator.MAX_LENGTH_HTML
+        bigger_len = text.Truncator.MAX_LENGTH_HTML + 1
+        valid_html = "<p>Joel is a slug</p>"  # 14 chars
         perf_test_values = [
-            (('</a' + '\t' * 50000) + '//>', None),
-            ('&' * 50000, '&' * 9 + '…'),
-            ('_X<<<<<<<<<<<>', None),
+            ("</a" + "\t" * (max_len - 6) + "//>", None),
+            ("</p" + "\t" * bigger_len + "//>", "</p" + "\t" * 6 + "…"),
+            ("&" * bigger_len, "&" * 9 + "…"),
+            ("_X<<<<<<<<<<<>", None),
+            (valid_html * bigger_len, "<p>Joel is a…</p>"),  # 10 chars
         ]
         for value, expected in perf_test_values:
             with self.subTest(value=value):
@@ -152,15 +159,25 @@ def test_truncate_html_words(self):
         truncator = text.Truncator('<p>I &lt;3 python, what about you?</p>')
         self.assertEqual('<p>I &lt;3 python,…</p>', truncator.words(3, html=True))
 
+    @patch("django.utils.text.Truncator.MAX_LENGTH_HTML", 10_000)
+    def test_truncate_words_html_size_limit(self):
+        max_len = text.Truncator.MAX_LENGTH_HTML
+        bigger_len = text.Truncator.MAX_LENGTH_HTML + 1
+        valid_html = "<p>Joel is a slug</p>"  # 4 words
         perf_test_values = [
-            ('</a' + '\t' * 50000) + '//>',
-            '&' * 50000,
-            '_X<<<<<<<<<<<>',
+            ("</a" + "\t" * (max_len - 6) + "//>", None),
+            ("</p" + "\t" * bigger_len + "//>", "</p" + "\t" * (max_len - 3) + "…"),
+            ("&" * max_len, None),  # no change
+            ("&" * bigger_len, "&" * max_len + "…"),
+            ("_X<<<<<<<<<<<>", None),
+            (valid_html * bigger_len, valid_html * 12 + "<p>Joel is…</p>"),  # 50 words
         ]
-        for value in perf_test_values:
+        for value, expected in perf_test_values:
             with self.subTest(value=value):
                 truncator = text.Truncator(value)
-                self.assertEqual(value, truncator.words(50, html=True))
+                self.assertEqual(
+                    expected if expected else value, truncator.words(50, html=True)
+                )
 
     def test_wrap(self):
         digits = '1234 67 9'
