I've been reading through RFC 2869 section 5.14
And RFC 5176 Section 3.4
And from how I understand it and how this is implemented by FreeRadius in 3.0.3 and 2.2.5.
Access-Request or Accounting or Status Messages:
The Message-Authenticator is calculated after the packet Authenticator has been generated using a HMAC-MD5 with the Message-Authenticator being a zero 16 byte string and then the HMAC-MD5 calculated using the shared secret as the key then replaced over the 16 zero byte field with the calculated value.
So a packet would look like this if the User-Name was "bob" with password "hello" and the shared secret of "testing123"
Request to HMAC-MD5 calculate:
0101002b14dbc57d377ba1846411553a49bf019b0105626f62501200000000000000000000000000000000
HMAC-MD5 46ecb6b4dde2c7e15a7d2a86fb05f1b2
Resulting request:
0101002b14dbc57d377ba1846411553a49bf019b0105626f62501246ecb6b4dde2c7e15a7d2a86fb05f1b2
Whereas because of this line in RFC5176 with my emphasis on the word before:
"The Message-Authenticator Attribute is calculated and inserted in the packet *before* the Request Authenticator is calculated."
So this means that the Message-Authenticator HMAC value should be calculated on the assumption the Packet Authenticator is all zero bytes so it would look something like this:
2b90002b000000000000000000000000000000000105626f62501200000000000000000000000000000000
And then the Packet Authenticator and the Message-Authenticator gets added in and you end up with a packet like this:
2b90002b9b6756059c3b56559d67f44418ae1fb70105626f6250125d68bd8fc122f6f2346e51872ba21fc3
Is this correct? As that is how it seems to be working for me. And I just wanted to make sure I was approaching this correctly. As it seems a little strange that the CoA/DM messages would prefer to have a null Authenticator message when calculating a Message-Authenticator. But it seems to be the way it is.
Cheers
Peter