Add a serialization check after parsing
[gpg-migrate.git] / gpg-migrate.py
index 5a3c8f0ef16a96ca0c4ac3ea8790ed2383485fb4..8d475f76e86f75436dde8c69221adbbed6e82942 100755 (executable)
@@ -1,10 +1,17 @@
 #!/usr/bin/python
 
+import getpass as _getpass
 import hashlib as _hashlib
+import math as _math
 import re as _re
 import subprocess as _subprocess
 import struct as _struct
 
+import Crypto.Cipher.AES as _crypto_cipher_aes
+import Crypto.Cipher.Blowfish as _crypto_cipher_blowfish
+import Crypto.Cipher.CAST as _crypto_cipher_cast
+import Crypto.Cipher.DES3 as _crypto_cipher_des3
+
 
 def _get_stdout(args, stdin=None):
     stdin_pipe = None
@@ -105,6 +112,23 @@ class PGPPacket (dict):
         'aes with 192-bit key': 128,
         'aes with 256-bit key': 128,
         'cast5': 64,
+        'twofish': 128,
+        }
+
+    _crypto_module = {
+        'aes with 128-bit key': _crypto_cipher_aes,
+        'aes with 192-bit key': _crypto_cipher_aes,
+        'aes with 256-bit key': _crypto_cipher_aes,
+        'blowfish': _crypto_cipher_blowfish,
+        'cast5': _crypto_cipher_cast,
+        'tripledes': _crypto_cipher_des3,
+        }
+
+    _key_size = {  # in bits
+        'aes with 128-bit key': 128,
+        'aes with 192-bit key': 192,
+        'aes with 256-bit key': 256,
+        'cast5': 128,
         }
 
     _compression_algorithms = {
@@ -150,6 +174,16 @@ class PGPPacket (dict):
         110: 'private',
         }
 
+    _hashlib_name = {  # map OpenPGP-based names to hashlib names
+        'md5': 'md5',
+        'sha-1': 'sha1',
+        'ripe-md/160': 'ripemd160',
+        'sha256': 'sha256',
+        'sha384': 'sha384',
+        'sha512': 'sha512',
+        'sha224': 'sha224',
+        }
+
     _string_to_key_types = {
         0: 'simple',
         1: 'salted',
@@ -168,6 +202,8 @@ class PGPPacket (dict):
         110: 'private',
         }
 
+    _string_to_key_expbias = 6
+
     _signature_types = {
         0x00: 'binary document',
         0x01: 'canonical text document',
@@ -235,6 +271,10 @@ class PGPPacket (dict):
 
     _clean_type_regex = _re.compile('\W+')
 
+    def __init__(self, key=None):
+        super(PGPPacket, self).__init__()
+        self.key = key
+
     def _clean_type(self, type=None):
         if type is None:
             type = self['type']
@@ -263,14 +303,31 @@ class PGPPacket (dict):
     def _str_public_subkey_packet(self):
         return self._str_generic_key_packet()
 
+    def _str_generic_key_packet(self):
+        return self['fingerprint'][-8:].upper()
+
     def _str_secret_key_packet(self):
-        return self._str_generic_key_packet()
+        return self._str_generic_secret_key_packet()
 
     def _str_secret_subkey_packet(self):
-        return self._str_generic_key_packet()
-
-    def _str_generic_key_packet(self):
-        return self['fingerprint'][-8:].upper()
+        return self._str_generic_secret_key_packet()
+
+    def _str_generic_secret_key_packet(self):
+        lines = [self._str_generic_key_packet()]
+        for label, key in [
+                ('symmetric encryption',
+                 'symmetric-encryption-algorithm'),
+                ('s2k hash', 'string-to-key-hash-algorithm'),
+                ('s2k count', 'string-to-key-count'),
+                ('s2k salt', 'string-to-key-salt'),
+                ('IV', 'initial-vector'),
+                ]:
+            if key in self:
+                value = self[key]
+                if isinstance(value, bytes):
+                    value = ' '.join('{:02x}'.format(byte) for byte in value)
+                lines.append('  {}: {}'.format(label, value))
+        return '\n'.join(lines)
 
     def _str_signature_packet(self):
         lines = [self['signature-type']]
@@ -354,6 +411,7 @@ class PGPPacket (dict):
             raise NotImplementedError(
                 'cannot parse packet type {!r}'.format(self['type']))
         method(data=packet)
+        self['raw'] = data[:offset]
         return offset
 
     def _parse_header(self, data):
@@ -399,6 +457,15 @@ class PGPPacket (dict):
         offset += length
         return (offset, value)
 
+    @classmethod
+    def _decode_string_to_key_count(cls, data):
+        r"""Decode RFC 4880's string-to-key count
+
+        >>> PGPPacket._decode_string_to_key_count(b'\x97'[0])
+        753664
+        """
+        return (16 + (data & 15)) << ((data >> 4) + cls._string_to_key_expbias)
+
     def _parse_string_to_key_specifier(self, data):
         self['string-to-key-type'] = self._string_to_key_types[data[0]]
         offset = 1
@@ -418,7 +485,8 @@ class PGPPacket (dict):
             offset += 1
             self['string-to-key-salt'] = data[offset: offset + 8]
             offset += 8
-            self['string-to-key-coded-count'] = data[offset]
+            self['string-to-key-count'] = self._decode_string_to_key_count(
+                data=data[offset])
             offset += 1
         else:
             raise NotImplementedError(
@@ -517,13 +585,63 @@ class PGPPacket (dict):
                         self['symmetric-encryption-algorithm']))
             self['initial-vector'] = data[offset: offset + block_size]
             offset += block_size
+            ciphertext = data[offset:]
+            offset += len(ciphertext)
+            decrypted_data = self.decrypt_symmetric_encryption(data=ciphertext)
+        else:
+            decrypted_data = data[offset:key_end]
         if string_to_key_usage in [0, 255]:
             key_end = -2
+        elif string_to_key_usage == 254:
+            key_end = -20
         else:
             key_end = 0
-        self['secret-key'] = data[offset:key_end]
+        secret_key = decrypted_data[:key_end]
+        secret_offset = 0
         if key_end:
-            self['secret-key-checksum'] = data[key_end:]
+            secret_key_checksum = decrypted_data[key_end:]
+            if key_end == -2:
+                calculated_checksum = sum(secret_key) % 65536
+            else:
+                checksum_hash = _hashlib.sha1()
+                checksum_hash.update(secret_key)
+                calculated_checksum = checksum_hash.digest()
+            if secret_key_checksum != calculated_checksum:
+                raise ValueError(
+                    'corrupt secret key (checksum {} != expected {})'.format(
+                        secret_key_checksum, calculated_checksum))
+        if self['public-key-algorithm'].startswith('rsa '):
+            o, self['secret-exponent'] = self._parse_multiprecision_integer(
+                secret_key[secret_offset:])
+            secret_offset += o
+            o, self['secret-prime-p'] = self._parse_multiprecision_integer(
+                secret_key[secret_offset:])
+            secret_offset += o
+            o, self['secret-prime-q'] = self._parse_multiprecision_integer(
+                secret_key[secret_offset:])
+            secret_offset += o
+            o, self['secret-inverse-of-p-mod-q'] = (
+                self._parse_multiprecision_integer(
+                    secret_key[secret_offset:]))
+            secret_offset += o
+        elif self['public-key-algorithm'].startswith('dsa '):
+            o, self['secret-exponent'] = self._parse_multiprecision_integer(
+                secret_key[secret_offset:])
+            secret_offset += o
+        elif self['public-key-algorithm'].startswith('elgamal '):
+            o, self['secret-exponent'] = self._parse_multiprecision_integer(
+                secret_key[secret_offset:])
+            secret_offset += o
+        else:
+            raise NotImplementedError(
+                'algorithm-specific key fields for {}'.format(
+                    self['public-key-algorithm']))
+        if secret_offset != len(secret_key):
+            raise ValueError(
+                ('parsed {} out of {} bytes of algorithm-specific key fields '
+                 'for {}').format(
+                     secret_offset, len(secret_key),
+                     self['public-key-algorithm']))
 
     def _parse_signature_subpackets(self, data):
         offset = 0
@@ -646,7 +764,7 @@ class PGPPacket (dict):
             subpacket['features'].add('modification detection')
 
     def _parse_embedded_signature_signature_subpacket(self, data, subpacket):
-        subpacket['embedded'] = PGPPacket()
+        subpacket['embedded'] = PGPPacket(key=self.key)
         subpacket['embedded']._parse_signature_packet(data=data)
 
     def _parse_user_id_packet(self, data):
@@ -684,13 +802,203 @@ class PGPPacket (dict):
             length_data,
             ])
 
+    @staticmethod
+    def _serialize_multiprecision_integer(integer):
+        r"""Serialize RFC 4880's multipricision integers
 
-def packets_from_bytes(data):
-    offset = 0
-    while offset < len(data):
-        packet = PGPPacket()
-        offset += packet.from_bytes(data=data[offset:])
-        yield packet
+        >>> PGPPacket._serialize_multiprecision_integer(1)
+        b'\x00\x01\x01'
+        >>> PGPPacket._serialize_multiprecision_integer(511)
+        b'\x00\t\x01\xff'
+        """
+        bit_length = int(_math.log(integer, 2)) + 1
+        chunks = [
+            _struct.pack('>H', bit_length),
+            ]
+        while integer > 0:
+            chunks.insert(1, bytes([integer & 0xff]))
+            integer = integer >> 8
+        return b''.join(chunks)
+
+    @classmethod
+    def _encode_string_to_key_count(cls, count):
+        r"""Encode RFC 4880's string-to-key count
+
+        >>> PGPPacket._encode_string_to_key_count(753664)
+        b'\x97'
+        """
+        coded_count = 0
+        count = count >> cls._string_to_key_expbias
+        while not count & 1:
+            count = count >> 1
+            coded_count += 1 << 4
+        coded_count += count & 15
+        return bytes([coded_count])
+
+    def _serialize_string_to_key_specifier(self):
+        string_to_key_type = bytes([
+            self._reverse(
+                self._string_to_key_types, self['string-to-key-type']),
+            ])
+        chunks = [string_to_key_type]
+        if self['string-to-key-type'] == 'simple':
+            chunks.append(bytes([self._reverse(
+                self._hash_algorithms, self['string-to-key-hash-algorithm'])]))
+        elif self['string-to-key-type'] == 'salted':
+            chunks.append(bytes([self._reverse(
+                self._hash_algorithms, self['string-to-key-hash-algorithm'])]))
+            chunks.append(self['string-to-key-salt'])
+        elif self['string-to-key-type'] == 'iterated and salted':
+            chunks.append(bytes([self._reverse(
+                self._hash_algorithms, self['string-to-key-hash-algorithm'])]))
+            chunks.append(self['string-to-key-salt'])
+            chunks.append(self._encode_string_to_key_count(
+                count=self['string-to-key-count']))
+        else:
+            raise NotImplementedError(
+                'string-to-key type {}'.format(self['string-to-key-type']))
+        return offset
+        return b''.join(chunks)
+
+    def _serialize_public_key_packet(self):
+        return self._serialize_generic_public_key_packet()
+
+    def _serialize_public_subkey_packet(self):
+        return self._serialize_generic_public_key_packet()
+
+    def _serialize_generic_public_key_packet(self):
+        key_version = bytes([self['key-version']])
+        chunks = [key_version]
+        if self['key-version'] != 4:
+            raise NotImplementedError(
+                'public (sub)key packet version {}'.format(
+                    self['key-version']))
+        chunks.append(_struct.pack('>I', self['creation-time']))
+        chunks.append(bytes([self._reverse(
+            self._public_key_algorithms, self['public-key-algorithm'])]))
+        if self['public-key-algorithm'].startswith('rsa '):
+            chunks.append(self._serialize_multiprecision_integer(
+                self['public-modulus']))
+            chunks.append(self._serialize_multiprecision_integer(
+                self['public-exponent']))
+        elif self['public-key-algorithm'].startswith('dsa '):
+            chunks.append(self._serialize_multiprecision_integer(
+                self['prime']))
+            chunks.append(self._serialize_multiprecision_integer(
+                self['group-order']))
+            chunks.append(self._serialize_multiprecision_integer(
+                self['group-generator']))
+            chunks.append(self._serialize_multiprecision_integer(
+                self['public-key']))
+        elif self['public-key-algorithm'].startswith('elgamal '):
+            chunks.append(self._serialize_multiprecision_integer(
+                self['prime']))
+            chunks.append(self._serialize_multiprecision_integer(
+                self['group-generator']))
+            chunks.append(self._serialize_multiprecision_integer(
+                self['public-key']))
+        else:
+            raise NotImplementedError(
+                'algorithm-specific key fields for {}'.format(
+                    self['public-key-algorithm']))
+        return b''.join(chunks)
+
+    def _string_to_key(self, string, key_size):
+        if key_size % 8:
+            raise ValueError(
+                '{}-bit key is not an integer number of bytes'.format(
+                    key_size))
+        key_size_bytes = key_size // 8
+        hash_name = self._hashlib_name[
+            self['string-to-key-hash-algorithm']]
+        string_hash = _hashlib.new(hash_name)
+        hashes = _math.ceil(key_size_bytes / string_hash.digest_size)
+        key = b''
+        if self['string-to-key-type'] == 'simple':
+            update_bytes = string
+        elif self['string-to-key-type'] in [
+                'salted',
+                'iterated and salted',
+                ]:
+            update_bytes = self['string-to-key-salt'] + string
+            if self['string-to-key-type'] == 'iterated and salted':
+                count = self['string-to-key-count']
+                if count < len(update_bytes):
+                    count = len(update_bytes)
+        else:
+            raise NotImplementedError(
+                'key calculation for string-to-key type {}'.format(
+                    self['string-to-key-type']))
+        for padding in range(hashes):
+            string_hash = _hashlib.new(hash_name)
+            string_hash.update(padding * b'\x00')
+            if self['string-to-key-type'] in [
+                    'simple',
+                    'salted',
+                    ]:
+                string_hash.update(update_bytes)
+            elif self['string-to-key-type'] == 'iterated and salted':
+                remaining = count
+                while remaining > 0:
+                    string_hash.update(update_bytes[:remaining])
+                    remaining -= len(update_bytes)
+            key += string_hash.digest()
+        key = key[:key_size_bytes]
+        return key
+
+    def decrypt_symmetric_encryption(self, data):
+        """Decrypt OpenPGP's Cipher Feedback mode"""
+        algorithm = self['symmetric-encryption-algorithm']
+        module = self._crypto_module[algorithm]
+        key_size = self._key_size[algorithm]
+        segment_size_bits = self._cipher_block_size[algorithm]
+        if segment_size_bits % 8:
+            raise NotImplementedError(
+                ('{}-bit segment size for {} is not an integer number of bytes'
+                 ).format(segment_size_bits, algorithm))
+        segment_size_bytes = segment_size_bits // 8
+        padding = segment_size_bytes - len(data) % segment_size_bytes
+        if padding:
+            data += b'\x00' * padding
+        if self.key and self.key._cache_passphrase and self.key._passphrase:
+            passphrase = self.key._passphrase
+        else:
+            passphrase = _getpass.getpass(
+                'passphrase for {}: '.format(self['fingerprint'][-8:]))
+            passphrase = passphrase.encode('ascii')
+            if self.key and self.key._cache_passphrase:
+                self.key._passphrase = passphrase
+        key = self._string_to_key(string=passphrase, key_size=key_size)
+        cipher = module.new(
+            key=key,
+            mode=module.MODE_CFB,
+            IV=self['initial-vector'],
+            segment_size=segment_size_bits)
+        plaintext = cipher.decrypt(data)
+        if padding:
+            plaintext = plaintext[:-padding]
+        return plaintext
+
+    def check_roundtrip(self):
+        serialized = self.to_bytes()
+        source = self['raw']
+        if serialized != source:
+            if len(serialized) != len(source):
+                raise ValueError(
+                    ('serialized {} is {} bytes long, '
+                     'but input is {} bytes long').format(
+                         self['type'], len(serialized), len(source)))
+            chunk_size = 8
+            for i in range(0, len(source), 8):
+                in_chunk = source[i: i + chunk_size]
+                out_chunk = serialized[i: i + chunk_size]
+                if in_chunk != out_chunk:
+                    raise ValueError(
+                        ('serialized {} differs from input packet: '
+                         'at byte {}, {} != {}').format(
+                            self['type'], i,
+                            ' '.join('{:02x}'.format(byte) for byte in out_chunk),
+                            ' '.join('{:02x}'.format(byte) for byte in in_chunk)))
 
 
 class PGPKey (object):
@@ -729,8 +1037,10 @@ class PGPKey (object):
     [1]: http://tools.ietf.org/search/rfc4880#section-11.1
     [2]: http://tools.ietf.org/search/rfc4880#section-11.2
     """
-    def __init__(self, fingerprint):
+    def __init__(self, fingerprint, cache_passphrase=False):
         self.fingerprint = fingerprint
+        self._cache_passphrase = cache_passphrase
+        self._passphrase = None
         self.public_packets = None
         self.secret_packets = None
 
@@ -754,7 +1064,7 @@ class PGPKey (object):
         key_export = _get_stdout(
             ['gpg', '--export', self.fingerprint])
         self.public_packets = list(
-            packets_from_bytes(data=key_export))
+            self._packets_from_bytes(data=key_export))
         if self.public_packets[0]['type'] != 'public-key packet':
             raise ValueError(
                 '{} does not start with a public-key packet'.format(
@@ -762,11 +1072,20 @@ class PGPKey (object):
         key_secret_export = _get_stdout(
             ['gpg', '--export-secret-keys', self.fingerprint])
         self.secret_packets = list(
-            packets_from_bytes(data=key_secret_export))
+            self._packets_from_bytes(data=key_secret_export))
         if self.secret_packets[0]['type'] != 'secret-key packet':
             raise ValueError(
                 '{} does not start with a secret-key packet'.format(
                     self.fingerprint))
+        for packet in self.public_packets + self.secret_packets:
+            packet.check_roundtrip()
+
+    def _packets_from_bytes(self, data):
+        offset = 0
+        while offset < len(data):
+            packet = PGPPacket(key=self)
+            offset += packet.from_bytes(data=data[offset:])
+            yield packet
 
     def export_to_gpg(self):
         raise NotImplemetedError('export to gpg')
@@ -776,16 +1095,16 @@ class PGPKey (object):
         pass
 
 
-def migrate(old_key, new_key):
+def migrate(old_key, new_key, cache_passphrase=False):
     """Add the old key and sub-keys to the new key
 
     For example, to upgrade your master key, while preserving old
     signatures you'd made.  You will lose signature *on* your old key
     though, since sub-keys can't be signed (I don't think).
     """
-    old_key = PGPKey(fingerprint=old_key)
+    old_key = PGPKey(fingerprint=old_key, cache_passphrase=cache_passphrase)
     old_key.import_from_gpg()
-    new_key = PGPKey(fingerprint=new_key)
+    new_key = PGPKey(fingerprint=new_key, cache_passphrase=cache_passphrase)
     new_key.import_from_gpg()
     new_key.import_from_key(key=old_key)
 
@@ -797,4 +1116,4 @@ if __name__ == '__main__':
     import sys as _sys
 
     old_key, new_key = _sys.argv[1:3]
-    migrate(old_key=old_key, new_key=new_key)
+    migrate(old_key=old_key, new_key=new_key, cache_passphrase=True)