Re: iSCSI: ExpStatRN

[julian_satran@il.ibm.com]

Josh,

You have probably seen it by yourself by now... but anyhow.

Statuses are numbered as issued (no relation to command numbering) and are
counted per connection.
When they disapear (digest error, dropped connection) the initiator will be
able to tell and the target will be able to resend.

Julo

Joshua Tseng <jtseng@NishanSystems.com> on 22/02/2001 02:32:21

Please respond to Joshua Tseng <jtseng@NishanSystems.com>

To:   ips@ece.cmu.edu, Julian Satran/Haifa/IBM@IBMIL
cc:
Subject:  iSCSI:  ExpStatRN




Julian,

I am confused by section 2.3.4 where it says

"Command responses up to ExpStatRN-1 (mod 2**32) have been received
(acknowledges status) on the connection."

First of all, can't a target execute the commands in a different order from
the order in which the commands were received?  If so, then won't the
StatSN
coming back NOT be in sequential order?  The initiator needs to scoreboard
responses from the target, and a single cumulative acknowledgement counter
won't do the job.

Who chooses the StatSN value for each response?  Section 2.4.7 says
the target generates this value on a per-connection basis.  But the name
"ExpStatSN", which is sent by the initiator, implies that the initiator
generates this value.

If the initiator generates this value, then how can multiple commands be
in flight if the value ExpStatSN-1 has already been received by the
initiator?

If the target generates this value, then "ExpStatSN" is an inapproporiate
and misleading name for this field.  It implies that the StatSN for the
corresponding response should be the same as the ExpStatSN generated
by the initiator for the original outgoing command.

Josh