Add HTML-5 "formaction" attribute to "defs.link_attrs" (GH-316)

Resolves https://bugs.launchpad.net/lxml/+bug/1888153
See https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2021-28957

2 files changed, 17 insertions, 0 deletions

diff --git a/src/lxml/html/defs.py b/src/lxml/html/defs.py
index 1b3a75b3..2058ea33 100644
--- a/src/lxml/html/defs.py
+++ b/src/lxml/html/defs.py
@@ -23,6 +23,8 @@ link_attrs = frozenset([
     'usemap',
     # Not standard:
     'dynsrc', 'lowsrc',
+    # HTML5 formaction
+    'formaction'
     ])
 
 # Not in the HTML 4 spec:
diff --git a/src/lxml/html/tests/test_clean.py b/src/lxml/html/tests/test_clean.py
index 0e669f98..45c2e83a 100644
--- a/src/lxml/html/tests/test_clean.py
+++ b/src/lxml/html/tests/test_clean.py
@@ -123,6 +123,21 @@ class CleanerTest(unittest.TestCase):
             b'<math><style>/* deleted */</style></math>',
             lxml.html.tostring(clean_html(s)))
 
+    def test_formaction_attribute_in_button_input(self):
+        # The formaction attribute overrides the form's action and should be
+        # treated as a malicious link attribute
+        html = ('<form id="test"><input type="submit" formaction="javascript:alert(1)"></form>'
+        '<button form="test" formaction="javascript:alert(1)">X</button>')
+        expected = ('<div><form id="test"><input type="submit" formaction=""></form>'
+        '<button form="test" formaction="">X</button></div>')
+        cleaner = Cleaner(
+            forms=False,
+            safe_attrs_only=False,
+        )
+        self.assertEqual(
+            expected,
+            cleaner.clean_html(html))
+
 
 def test_suite():
     suite = unittest.TestSuite()