Merge pull request #2 from abadger/feature/ssl-match-hostname-17997

There's apparently another security issue in the python3 match_hostname code. No CVE has been issued for it yet:

http://bugs.python.org/issue17997#msg194950

This merge includes two commits. The first updates the included match_hostname code to reflect what's in the python-3.3.3 and python-3.4 stdlib (with a minor change to preserve python2 compat). The second commit adds a check for the backports.ssl_match_hostname module from pypi: https://pypi.python.org/pypi/backports.ssl_match_hostname

If the stdlib doesn't have ssl_match_hostname but backports.ssl_match_hostname exists it uses that code. If neither one are present, then it uses the code included in setuptools.

Using backports.ssl_match_hostname helps system packagers and system admins to have a single place to maintain SSL support rather than every package that's copying the match_hostname code. On the other hand, it means that users won't get any fixes before they go into the backports.ssl_match_hostname module. Brandon Rhodes is the owner of that module and Toshio has done the last several releases to make sure that module is current with the match_hostname security issues.

setuptools/ssl_support.py

@@ -85,33 +85,74 @@ except ImportError:
 try:
     from ssl import CertificateError, match_hostname
 except ImportError:
+    try:
+        from backports.ssl_match_hostname import CertificateError
+        from backports.ssl_match_hostname import match_hostname
+    except ImportError:
+        CertificateError = None
+        match_hostname = None
+
+if not CertificateError:
     class CertificateError(ValueError):
         pass
 
-    def _dnsname_to_pat(dn, max_wildcards=1):
+if not match_hostname:
+    def _dnsname_match(dn, hostname, max_wildcards=1):
+        """Matching according to RFC 6125, section 6.4.3
+
+        http://tools.ietf.org/html/rfc6125#section-6.4.3
+        """
         pats = []
-        for frag in dn.split(r'.'):
-            if frag.count('*') > max_wildcards:
-                # Issue #17980: avoid denials of service by refusing more
-                # than one wildcard per fragment.  A survery of established
-                # policy among SSL implementations showed it to be a
-                # reasonable choice.
-                raise CertificateError(
-                    "too many wildcards in certificate DNS name: " + repr(dn))
-            if frag == '*':
-                # When '*' is a fragment by itself, it matches a non-empty dotless
-                # fragment.
-                pats.append('[^.]+')
-            else:
-                # Otherwise, '*' matches any dotless fragment.
-                frag = re.escape(frag)
-                pats.append(frag.replace(r'\*', '[^.]*'))
-        return re.compile(r'\A' + r'\.'.join(pats) + r'\Z', re.IGNORECASE)
+        if not dn:
+            return False
+
+        # Ported from python3-syntax:
+        # leftmost, *remainder = dn.split(r'.')
+        parts = dn.split(r'.')
+        leftmost = parts[0]
+        remainder = parts[1:]
+
+        wildcards = leftmost.count('*')
+        if wildcards > max_wildcards:
+            # Issue #17980: avoid denials of service by refusing more
+            # than one wildcard per fragment.  A survey of established
+            # policy among SSL implementations showed it to be a
+            # reasonable choice.
+            raise CertificateError(
+                "too many wildcards in certificate DNS name: " + repr(dn))
+
+        # speed up common case w/o wildcards
+        if not wildcards:
+            return dn.lower() == hostname.lower()
+
+        # RFC 6125, section 6.4.3, subitem 1.
+        # The client SHOULD NOT attempt to match a presented identifier in which
+        # the wildcard character comprises a label other than the left-most label.
+        if leftmost == '*':
+            # When '*' is a fragment by itself, it matches a non-empty dotless
+            # fragment.
+            pats.append('[^.]+')
+        elif leftmost.startswith('xn--') or hostname.startswith('xn--'):
+            # RFC 6125, section 6.4.3, subitem 3.
+            # The client SHOULD NOT attempt to match a presented identifier
+            # where the wildcard character is embedded within an A-label or
+            # U-label of an internationalized domain name.
+            pats.append(re.escape(leftmost))
+        else:
+            # Otherwise, '*' matches any dotless string, e.g. www*
+            pats.append(re.escape(leftmost).replace(r'\*', '[^.]*'))
+
+        # add the remaining fragments, ignore any wildcards
+        for frag in remainder:
+            pats.append(re.escape(frag))
+
+        pat = re.compile(r'\A' + r'\.'.join(pats) + r'\Z', re.IGNORECASE)
+        return pat.match(hostname)
 
     def match_hostname(cert, hostname):
         """Verify that *cert* (in decoded format as returned by
-        SSLSocket.getpeercert()) matches the *hostname*.  RFC 2818 rules
-        are mostly followed, but IP addresses are not accepted for *hostname*.
+        SSLSocket.getpeercert()) matches the *hostname*.  RFC 2818 and RFC 6125
+        rules are followed, but IP addresses are not accepted for *hostname*.
 
         CertificateError is raised on failure. On success, the function
         returns nothing.
@@ -122,7 +163,7 @@ except ImportError:
         san = cert.get('subjectAltName', ())
         for key, value in san:
             if key == 'DNS':
-                if _dnsname_to_pat(value).match(hostname):
+                if _dnsname_match(value, hostname):
                     return
                 dnsnames.append(value)
         if not dnsnames:
@@ -133,7 +174,7 @@ except ImportError:
                     # XXX according to RFC 2818, the most specific Common Name
                     # must be used.
                     if key == 'commonName':
-                        if _dnsname_to_pat(value).match(hostname):
+                        if _dnsname_match(value, hostname):
                             return
                         dnsnames.append(value)
         if len(dnsnames) > 1: