In-Reply-To: <2A8D600C919D2D41A3DAA2D2FEC7171437D59B@homer.qlogic.org>
Subject: Re: Immediate Commands
To: "Saikrishnan Krishnamurthy" <saikrishnan.krishnamurthy@qlogic.com>
Cc: ips@ece.cmu.edu, owner-ips@ece.cmu.edu
Message-ID: <OF50F548D1.FA7DB1EE-ONC2256D2D.006AF09E-C2256D2D.006C08D1@telaviv.ibm.com>
From: "Julian Satran" <Julian_Satran@il.ibm.com>
Date: Wed, 21 May 2003 22:40:01 +0300
MIME-Version: 1.0
Content-type: text/plain; charset=us-ascii
Sender: owner-ips@ece.cmu.edu
Precedence: bulk



owner-ips@ece.cmu.edu wrote on 21/05/2003 19:33:04:

> Hi all,
>         The iSCSI standard specifies immediate command handling
> (sec. 3.2.2.1) at the target and mandates the target to support
> atleast one Immediate Task Mgmnt. and one Immediate non-Task Mgmnt.
> command "per connection". This implies that initiators can have
> multiple (more than two) outstanding Immediate commands to the same
> target on other connections (in multiple connection scenario). Does
> the standard mandate any restriction on the initiators in handling
> or delivering multiple immediate commands ??.
>
> Sai.

iSCSI does not impose explicit restriction on initiators. However since
targets may drop immediate commands for which they don't have resources a
prudent initiator designer will build his code considering the target
limitations. And there is no real incentive for the target builder to
provide more than the minimum (in most case) since a prudent initiator will
never assume more. Black humor aside - a good (and expensive) target is
supposed the act quickly on immediate commands.

Regards,
Julo