iSCSI: target resource "leak"/logout command

[Pierre Labat <pierre_labat@hp.com>]

Julian,


There is a corner case where
- a [group of] command completion[s]  will never be acked (ExpStatRN)
- the corresponding task[s] are not "retried" because they are perfectly

   completed from the initiator point of view and
- the session continues to work fine.
Thus the resource for the task[s] may never
be released on the target (the target waits for ExpStatRN acking
the completion before releasing task resources) because
on the target side the ack of the completion is not received
and on the initiator will not retry the command.

A resource (memory) "leak" happens on the target.

First i described the scenario, then the solution.

Scenario
=======
Session with 2 TCP connections 1 and 2

1)  A command completion (StatRN=10) is received by
      the initiator (over cx 1) for the command 5.

2) The initiator sends back ExpStatRN=11 with a command over  TCP cx 1

3) The TCP cx 1 drops and the command 5 (so ExpStatRN=11) will never
   make it to the target.

4) The command 5 is not retried on the connection 2 because for the
initiator
      it is perfectly completed.

5) The target never gets the completion acked and doesn't know when
    to release resources.


Solution
=======

- Add the field ExpStatRN in the logout command.
- The initiator puts in this field the last value of ExpStatRN of the
failed
    cx. (Not the ExpStatRN value of the connection used to transport the
logout)
- The target when receiving the logout, reads ExpStatRN and is able to
free up
   the resources.


Regards,

Pierre