From 6f030b1149bd8fa4ba90452e77cb3edc095ce54e Mon Sep 17 00:00:00 2001
From: Mariusz Felisiak <felisiak.mariusz@gmail.com>
Date: Tue, 22 Aug 2023 08:53:03 +0200
Subject: [PATCH] [3.2.x] Fixed CVE-2023-41164 -- Fixed potential DoS in
 django.utils.encoding.uri_to_iri().

Thanks MProgrammer (https://hackerone.com/mprogrammer) for the report.

Co-authored-by: nessita <124304+nessita@users.noreply.github.com>
---
 django/utils/encoding.py           |  6 ++++--
 docs/releases/3.2.21.txt           |  7 ++++++-
 tests/utils_tests/test_encoding.py | 21 ++++++++++++++++++++-
 3 files changed, 30 insertions(+), 4 deletions(-)

diff --git a/django/utils/encoding.py b/django/utils/encoding.py
index e1ebacef47..c5c4463b1c 100644
--- a/django/utils/encoding.py
+++ b/django/utils/encoding.py
@@ -229,6 +229,7 @@ def repercent_broken_unicode(path):
     repercent-encode any octet produced that is not part of a strictly legal
     UTF-8 octet sequence.
     """
+    changed_parts = []
     while True:
         try:
             path.decode()
@@ -236,9 +237,10 @@ def repercent_broken_unicode(path):
             # CVE-2019-14235: A recursion shouldn't be used since the exception
             # handling uses massive amounts of memory
             repercent = quote(path[e.start:e.end], safe=b"/#%[]=:;$&()+,!?*@'~")
-            path = path[:e.start] + repercent.encode() + path[e.end:]
+            changed_parts.append(path[:e.start] + repercent.encode())
+            path = path[e.end:]
         else:
-            return path
+            return b"".join(changed_parts) + path
 
 
 def filepath_to_uri(path):
diff --git a/docs/releases/3.2.21.txt b/docs/releases/3.2.21.txt
index 79efc679d1..062ac66682 100644
--- a/docs/releases/3.2.21.txt
+++ b/docs/releases/3.2.21.txt
@@ -6,4 +6,9 @@ Django 3.2.21 release notes
 
 Django 3.2.21 fixes a security issue with severity "moderate" in 3.2.20.
 
-...
+CVE-2023-41164: Potential denial of service vulnerability in ``django.utils.encoding.uri_to_iri()``
+===================================================================================================
+
+``django.utils.encoding.uri_to_iri()`` was subject to potential denial of
+service attack via certain inputs with a very large number of Unicode
+characters.
diff --git a/tests/utils_tests/test_encoding.py b/tests/utils_tests/test_encoding.py
index 36f2d8665f..42779050cb 100644
--- a/tests/utils_tests/test_encoding.py
+++ b/tests/utils_tests/test_encoding.py
@@ -1,9 +1,10 @@
 import datetime
+import inspect
 import sys
 import unittest
 from pathlib import Path
 from unittest import mock
-from urllib.parse import quote_plus
+from urllib.parse import quote, quote_plus
 
 from django.test import SimpleTestCase
 from django.utils.encoding import (
@@ -101,6 +102,24 @@ class TestEncodingUtils(SimpleTestCase):
         except RecursionError:
             self.fail('Unexpected RecursionError raised.')
 
+    def test_repercent_broken_unicode_small_fragments(self):
+        data = b"test\xfctest\xfctest\xfc"
+        decoded_paths = []
+
+        def mock_quote(*args, **kwargs):
+            # The second frame is the call to repercent_broken_unicode().
+            decoded_paths.append(inspect.currentframe().f_back.f_locals["path"])
+            return quote(*args, **kwargs)
+
+        with mock.patch("django.utils.encoding.quote", mock_quote):
+            self.assertEqual(repercent_broken_unicode(data), b"test%FCtest%FCtest%FC")
+
+        # decode() is called on smaller fragment of the path each time.
+        self.assertEqual(
+            decoded_paths,
+            [b"test\xfctest\xfctest\xfc", b"test\xfctest\xfc", b"test\xfc"],
+        )
+
 
 class TestRFC3987IEncodingUtils(unittest.TestCase):