[AFS3-std] rxgk implementation notes

Benjamin Kaduk kaduk@MIT.EDU
Fri, 1 Mar 2013 15:45:29 -0500 (EST)


On Thu, 28 Feb 2013, Benjamin Kaduk wrote:

> On Thu, 28 Feb 2013, Benjamin Kaduk wrote:
>
>> On Thu, 28 Feb 2013, Simon Wilkinson wrote:
>> 
>>> 
>>> On 28 Feb 2013, at 22:05, Benjamin Kaduk wrote:
>>
>>>>  All I see looking nowis discussion of using the maxcalls information as 
>>>> a way to migrate to having more than 4 calls per connection, which 
>>>> doesn't sound familiar. We also don't mention whether call numbers in use 
>>>> are checked as part of verifying the authenticator -- I do not currently 
>>>> do so.
>>> 
>>> You should just provide the call number vector to the RX layer, which will 
>>> take care of checking the call numbers within it (OpenAFS's RX currently 
>>> just sets up the connection according to the provided authenticator, which 
>>> will cause non-matching calls to fail as they are encountered)
>> 
>> Oh, hmm, I guess rxi_SetCallNumberVector is exported.
>
> We should probably adjust the description of RXGK_Authenticator.call_numbers, 
> as my first reading of it was that the variable-length array was only to 
> include active calls, not the whole vector.  Perhaps something about "all 
> available call channels" on this connection?

Currently have:
call_numbers: The set of current rx call numbers for all available 
channels; unused channels should report a call number of zero.

I also added in the "checking the response" section "The call number 
vector should be supplied to the rx implementation."

I'm still not sure whether maxcalls is supposed to be RX_MAXCALLS (4) or 
max(call_numbers).

-Ben