Blame SOURCES/CVE-2021-43818.patch

62262c
diff --git a/src/lxml/html/clean.py b/src/lxml/html/clean.py
62262c
index 15298b5..ee2f0f8 100644
62262c
--- a/src/lxml/html/clean.py
62262c
+++ b/src/lxml/html/clean.py
62262c
@@ -73,18 +73,25 @@ _looks_like_tag_content = re.compile(
62262c
 
62262c
 # All kinds of schemes besides just javascript: that can cause
62262c
 # execution:
62262c
-_is_image_dataurl = re.compile(
62262c
-    r'^data:image/.+;base64', re.I).search
62262c
+_find_image_dataurls = re.compile(
62262c
+    r'^data:image/(.+);base64,', re.I).findall
62262c
 _is_possibly_malicious_scheme = re.compile(
62262c
-    r'(?:javascript|jscript|livescript|vbscript|data|about|mocha):',
62262c
-    re.I).search
62262c
+    r'(javascript|jscript|livescript|vbscript|data|about|mocha):',
62262c
+    re.I).findall
62262c
+# SVG images can contain script content
62262c
+_is_unsafe_image_type = re.compile(r"(xml|svg)", re.I).findall
62262c
+
62262c
 def _is_javascript_scheme(s):
62262c
-    if _is_image_dataurl(s):
62262c
-        return None
62262c
-    return _is_possibly_malicious_scheme(s)
62262c
+    is_image_url = False
62262c
+    for image_type in _find_image_dataurls(s):
62262c
+        is_image_url = True
62262c
+        if _is_unsafe_image_type(image_type):
62262c
+            return True
62262c
+    if is_image_url:
62262c
+        return False
62262c
+    return bool(_is_possibly_malicious_scheme(s))
62262c
 
62262c
 _substitute_whitespace = re.compile(r'[\s\x00-\x08\x0B\x0C\x0E-\x19]+').sub
62262c
-# FIXME: should data: be blocked?
62262c
 
62262c
 # FIXME: check against: http://msdn2.microsoft.com/en-us/library/ms537512.aspx
62262c
 _conditional_comment_re = re.compile(
62262c
@@ -512,6 +519,8 @@ class Cleaner(object):
62262c
             return True
62262c
         if 'expression(' in style:
62262c
             return True
62262c
+        if '@import' in style:
62262c
+            return True
62262c
         if '
62262c
             # e.g. '">'
62262c
             return True
62262c
diff --git a/src/lxml/html/tests/test_clean.py b/src/lxml/html/tests/test_clean.py
62262c
index d8df527..7021e48 100644
62262c
--- a/src/lxml/html/tests/test_clean.py
62262c
+++ b/src/lxml/html/tests/test_clean.py
62262c
@@ -1,3 +1,5 @@
62262c
+import base64
62262c
+import gzip
62262c
 import unittest
62262c
 from lxml.tests.common_imports import make_doctest
62262c
 
62262c
@@ -89,6 +91,69 @@ class CleanerTest(unittest.TestCase):
62262c
             b'<math><style>/* deleted */</style></math>',
62262c
             lxml.html.tostring(clean_html(s)))
62262c
 
62262c
+    def test_sneaky_import_in_style(self):
62262c
+        # Prevent "@@importimport" -> "@import" replacement.
62262c
+        style_codes = [
62262c
+            "@@importimport(extstyle.css)",
62262c
+            "@ @  import import(extstyle.css)",
62262c
+            "@ @ importimport(extstyle.css)",
62262c
+            "@@  import import(extstyle.css)",
62262c
+            "@ @import import(extstyle.css)",
62262c
+            "@@importimport()",
62262c
+        ]
62262c
+        for style_code in style_codes:
62262c
+            html = '<style>%s</style>' % style_code
62262c
+            s = lxml.html.fragment_fromstring(html)
62262c
+
62262c
+            cleaned = lxml.html.tostring(clean_html(s))
62262c
+            self.assertEqual(
62262c
+                b'<style>/* deleted */</style>',
62262c
+                cleaned,
62262c
+                "%s  ->  %s" % (style_code, cleaned))
62262c
+
62262c
+    def test_svg_data_links(self):
62262c
+        # Remove SVG images with potentially insecure content.
62262c
+        svg = b'<svg onload="alert(123)" />'
62262c
+        svgz = gzip.compress(svg)
62262c
+        svg_b64 = base64.b64encode(svg).decode('ASCII')
62262c
+        svgz_b64 = base64.b64encode(svgz).decode('ASCII')
62262c
+        urls = [
62262c
+            "data:image/svg+xml;base64," + svg_b64,
62262c
+            "data:image/svg+xml-compressed;base64," + svgz_b64,
62262c
+        ]
62262c
+        for url in urls:
62262c
+            html = '' % url
62262c
+            s = lxml.html.fragment_fromstring(html)
62262c
+
62262c
+            cleaned = lxml.html.tostring(clean_html(s))
62262c
+            self.assertEqual(
62262c
+                b'',
62262c
+                cleaned,
62262c
+                "%s  ->  %s" % (url, cleaned))
62262c
+
62262c
+    def test_image_data_links(self):
62262c
+        data = b'123'
62262c
+        data_b64 = base64.b64encode(data).decode('ASCII')
62262c
+        urls = [
62262c
+            "data:image/jpeg;base64," + data_b64,
62262c
+            "data:image/apng;base64," + data_b64,
62262c
+            "data:image/png;base64," + data_b64,
62262c
+            "data:image/gif;base64," + data_b64,
62262c
+            "data:image/webp;base64," + data_b64,
62262c
+            "data:image/bmp;base64," + data_b64,
62262c
+            "data:image/tiff;base64," + data_b64,
62262c
+            "data:image/x-icon;base64," + data_b64,
62262c
+        ]
62262c
+        for url in urls:
62262c
+            html = '' % url
62262c
+            s = lxml.html.fragment_fromstring(html)
62262c
+
62262c
+            cleaned = lxml.html.tostring(clean_html(s))
62262c
+            self.assertEqual(
62262c
+                html.encode("UTF-8"),
62262c
+                cleaned,
62262c
+                "%s  ->  %s" % (url, cleaned))
62262c
+
62262c
     def test_formaction_attribute_in_button_input(self):
62262c
         # The formaction attribute overrides the form's action and should be
62262c
         # treated as a malicious link attribute