Re: iSCSI : On the subject of R2T and Task Tags.

[Pierre Labat <pierre_labat@hp.com>]

Matt Wakeley wrote:

> Santosh Rao wrote:
>
> > I agree with Matt on the semantics of "Target Task Tag" for WRITE PDUs and would
> > also like to request a "Target Task Tag" per task for WRITE PDUs.
> >
> > In addition, since multiple R2Ts can be outstanding, an additional qualifier is
> > required to track the R2T (Target R2T Tag) along the lines of a (RX_ID + SEQ_ID)
> > that Fibre Channel uses.
>
> I don't think an additional qualifier is needed.  For what purpose?
>
> The target just sends out all the R2Ts, and counts all the bytes coming back
> (remember - TCP guarantees to deliver each byte only once!).  Also remember,
> the iSCSI spec mandates that the initiator send only what was requested in the
> R2T... nothing more.

Matt,

To have differents tag or (tag + subtag) one for each R2T allows the target to find
directly the buffer address. For example  using the tag as an index in a table. If
not, the target has
to go through a list of R2T control structures  and for each one compare the offset
with
with the interval recorded in the R2T structure, in order to find the right buffer.
As it doesn't hurt on the initiator side, why not to do it?

In fact an R2T and the data related are as an exchange. Hence it is natural
to have one identifier/exchange.
Of course if the target want to use the same it could.

Regards,

Pierre

>
>
> -Matt Wakeley
> Agilent Technologies
>
> >
> > Thanks,
> > Santosh