python-idna/CVE-2026-45409.patch
2026-08-13 04:34:05 -04:00

135 lines
5.4 KiB
Diff
Raw Permalink Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

From 487f90216c00097caa6c5f10eb1f843ec2d8a4b6 Mon Sep 17 00:00:00 2001
From: Kim Davies <kim@cynosure.com.au>
Date: Sun, 10 May 2026 08:47:22 -0700
Subject: [PATCH 1/2] Merge commit from fork
---
idna/core.py | 14 ++++++++++++++
tests/test_idna.py | 13 +++++++++++++
2 files changed, 27 insertions(+)
diff --git a/idna/core.py b/idna/core.py
index 0dae61a..a549326 100644
--- a/idna/core.py
+++ b/idna/core.py
@@ -340,6 +340,15 @@ def encode(s: Union[str, bytes, bytearray], strict: bool = False, uts46: bool =
raise IDNAError('should pass a unicode string to the function rather than a byte string.')
if uts46:
s = uts46_remap(s, std3_rules, transitional)
+
+ # Reject inputs that exceed the maximum DNS domain length up-front.
+ # Each codepoint in a U-label contributes at least one octet to its
+ # A-label form, so any input longer than the domain limit cannot
+ # produce a valid A-domain. Short-circuiting here prevents per-label
+ # validation from being driven into quadratic time
+ if len(s) > 254:
+ raise IDNAError("Domain too long")
+
trailing_dot = False
result = []
if strict:
@@ -373,6 +382,11 @@ def decode(s: Union[str, bytes, bytearray], strict: bool = False, uts46: bool =
raise IDNAError('Invalid ASCII in A-label')
if uts46:
s = uts46_remap(s, std3_rules, False)
+ # See encode() for rationale; the same bound applies because every
+ # legal A-domain is at most 254 octets and every codepoint of a
+ # legal U-domain contributes at least one octet to its A-form.
+ if len(s) > 254:
+ raise IDNAError("Domain too long")
trailing_dot = False
result = []
if not strict:
diff --git a/tests/test_idna.py b/tests/test_idna.py
index 81afb32..5001b48 100755
--- a/tests/test_idna.py
+++ b/tests/test_idna.py
@@ -78,6 +78,19 @@ class IDNATests(unittest.TestCase):
self.assertFalse(idna.valid_label_length('a' * 64))
self.assertRaises(idna.IDNAError, idna.encode, 'a' * 64)
+ def test_oversized_input_rejected_promptly(self):
+ # GHSA-65pc-fj4g-8rjx: encode/decode must reject inputs that
+ # exceed the maximum DNS domain length before per-codepoint
+ # validation runs, so labels dominated by CONTEXTO codepoints
+ # cannot drive validation into quadratic time.
+ import time
+
+ for payload in ("٠" * 8000, "・" * 8000 + "漢"):
+ start = time.perf_counter()
+ self.assertRaises(idna.IDNAError, idna.encode, payload)
+ self.assertRaises(idna.IDNAError, idna.decode, payload)
+ self.assertLess(time.perf_counter() - start, 1.0)
+
def test_check_bidi(self):
l = '\u0061'
From 219662569606146771ac69fac5939d3308159064 Mon Sep 17 00:00:00 2001
From: metsw24-max <metsw24@gmail.com>
Date: Mon, 11 May 2026 20:59:30 +0530
Subject: [PATCH 2/2] Enforce early length limits in check_label
---
idna/core.py | 11 +++++++++++
tests/test_idna.py | 24 ++++++++++++++++++++++++
2 files changed, 35 insertions(+)
diff --git a/idna/core.py b/idna/core.py
index a549326..f863141 100644
--- a/idna/core.py
+++ b/idna/core.py
@@ -230,6 +230,17 @@ def check_label(label: Union[str, bytes, bytearray]) -> None:
label = label.decode('utf-8')
if len(label) == 0:
raise IDNAError('Empty Label')
+ # Reject oversized labels before per-codepoint validation runs.
+ # CONTEXTJ/CONTEXTO checks scan the whole label per codepoint, so an
+ # uncapped label drives validation into quadratic time
+ # (GHSA-65pc-fj4g-8rjx / CVE-2024-3651). encode()/decode() cap the
+ # whole-domain length; this cap protects direct callers of
+ # alabel/ulabel/check_label and the idna2008 incremental codec.
+ # Use the whole-domain bound rather than the per-label DNS bound so
+ # that UTS #46 lenient decoding of labels longer than 63 chars is
+ # preserved.
+ if not valid_string_length(label, trailing_dot=True):
+ raise IDNAError('Label too long')
check_nfc(label)
check_hyphen_ok(label)
diff --git a/tests/test_idna.py b/tests/test_idna.py
index 5001b48..2dc0892 100755
--- a/tests/test_idna.py
+++ b/tests/test_idna.py
@@ -91,6 +91,30 @@ class IDNATests(unittest.TestCase):
self.assertRaises(idna.IDNAError, idna.decode, payload)
self.assertLess(time.perf_counter() - start, 1.0)
+ def test_oversized_label_rejected_promptly(self):
+ # The whole-domain cap in encode()/decode() does not cover direct
+ # callers of alabel/ulabel/check_label, nor the idna2008
+ # incremental codec which calls alabel/ulabel per label. Without a
+ # per-label cap, a single oversized CONTEXTO-heavy label still
+ # drives validation into quadratic time.
+ import codecs
+ import time
+
+ import idna.codec # noqa: F401 (register the idna2008 codec)
+
+ payload = "・" * 8000 + "漢"
+ start = time.perf_counter()
+ self.assertRaises(idna.IDNAError, idna.check_label, payload)
+ self.assertRaises(idna.IDNAError, idna.alabel, payload)
+ self.assertRaises(idna.IDNAError, idna.ulabel, payload)
+ self.assertRaises(
+ idna.IDNAError,
+ codecs.getincrementalencoder("idna2008")().encode,
+ payload,
+ True,
+ )
+ self.assertLess(time.perf_counter() - start, 1.0)
+
def test_check_bidi(self):
l = '\u0061'