.. _ccache_file_format:

Credential cache file format
============================

There are four versions of the file format used by the FILE credential
cache type.  The first byte of the file always has the value 5, and
the value of the second byte contains the version number (1 through
4).  Versions 1 and 2 of the file format use native byte order for integer
representations.  Versions 3 and 4 always use big-endian byte order.

After the two-byte version indicator, the file has three parts: the
header (in version 4 only), the default principal name, and a sequence
of credentials.


Header format
-------------

The header appears only in format version 4.  It begins with a 16-bit
integer giving the length of the entire header, followed by a sequence
of fields.  Each field consists of a 16-bit tag, a 16-bit length, and
a value of the given length.  A file format implementation should
ignore fields with unknown tags.

At this time there is only one defined header field.  Its tag value is
1, its length is always 8, and its contents are two 32-bit integers
giving the seconds and microseconds of the time offset of the KDC
relative to the client.  Adding this offset to the current time on the
client should give the current time on the KDC, if that offset has not
changed since the initial authentication.


.. _cache_principal_format:

Principal format
----------------

The default principal is marshalled using the following informal
grammar::

    principal ::=
        name type (32 bits) [omitted in version 1]
        count of components (32 bits) [includes realm in version 1]
        realm (data)
        component1 (data)
        component2 (data)
        ...

    data ::=
        length (32 bits)
        value (length bytes)

There is no external framing on the default principal, so it must be
parsed according to the above grammar in order to find the sequence of
credentials which follows.


.. _ccache_credential_format:

Credential format
-----------------

The credential format uses the following informal grammar (referencing
the ``principal`` and ``data`` types from the previous section)::

    credential ::=
        client (principal)
        server (principal)
        keyblock (keyblock)
        authtime (32 bits)
        starttime (32 bits)
        endtime (32 bits)
        renew_till (32 bits)
        is_skey (1 byte, 0 or 1)
        ticket_flags (32 bits)
        addresses (addresses)
        authdata (authdata)
        ticket (data)
        second_ticket (data)

    keyblock ::=
        enctype (16 bits) [repeated twice in version 3]
        data

    addresses ::=
        count (32 bits)
        address1
        address2
        ...

    address ::=
        addrtype (16 bits)
        data

    authdata ::=
        count (32 bits)
        authdata1
        authdata2
        ...

    authdata ::=
        ad_type (16 bits)
        data

There is no external framing on a marshalled credential, so it must be
parsed according to the above grammar in order to find the next
credential.  There is also no count of credentials or marker at the
end of the sequence of credentials; the sequence ends when the file
ends.


Credential cache configuration entries
--------------------------------------

Configuration entries are encoded as credential entries.  The client
principal of the entry is the default principal of the cache.  The
server principal has the realm ``X-CACHECONF:`` and two or three
components, the first of which is ``krb5_ccache_conf_data``.  The
server principal's second component is the configuration key.  The
third component, if it exists, is a principal to which the
configuration key is associated.  The configuration value is stored in
the ticket field of the entry.  All other entry fields are zeroed.

Programs using credential caches must be aware of configuration
entries for several reasons:

* A program which displays the contents of a cache should not
  generally display configuration entries.

* The ticket field of a configuration entry is not (usually) a valid
  encoding of a Kerberos ticket.  An implementation must not treat the
  cache file as malformed if it cannot decode the ticket field.

* Configuration entries have an endtime field of 0 and might therefore
  always be considered expired, but they should not be treated as
  unimportant as a result.  For instance, a program which copies
  credentials from one cache to another should not omit configuration
  entries because of the endtime.

The following configuration keys are currently used in MIT krb5:

fast_avail
    The presence of this key with a non-empty value indicates that the
    KDC asserted support for FAST (see :rfc:`6113`) during the initial
    authentication, using the negotiation method described in
    :rfc:`6806` section 11.  This key is not associated with any
    principal.

pa_config_data
    The value of this key contains a JSON object representation of
    parameters remembered by the preauthentication mechanism used
    during the initial authentication.  These parameters may be used
    when refreshing credentials.  This key is associated with the
    server principal of the initial authentication (usually the local
    krbtgt principal of the client realm).

pa_type
    The value of this key is the ASCII decimal representation of the
    preauth type number used during the initial authentication.  This
    key is associated with the server principal of the initial
    authentication.

proxy_impersonator
    The presence of this key indicates that the cache is a synthetic
    delegated credential for use with S4U2Proxy.  The value is the
    name of the intermediate service whose TGT can be used to make
    S4U2Proxy requests for target services.  This key is not
    associated with any principal.

refresh_time
    The presence of this key indicates that the cache was acquired by
    the GSS mechanism using a client keytab.  The value is the ASCII
    decimal representation of a timestamp at which the GSS mechanism
    should attempt to refresh the credential cache from the client
    keytab.