iSCSI: Check Condition and Residuals

[Black_David@emc.com]

I think Julian's point is that the use of sense data
vs. transport support to return residual counts for
CHECK CONDITION is in T10's domain to specify, and
we ought to leave it to T10, but warn implementers
that it would be wrong to assume that iSCSI residuals
will only show up when GOOD status is returned.

Thanks,
--David
---------------------------------------------------
David L. Black, Senior Technologist
EMC Corporation, 42 South St., Hopkinton, MA  01748
+1 (508) 249-6449            FAX: +1 (508) 497-8018
black_david@emc.com       Mobile: +1 (978) 394-7754
---------------------------------------------------

> -----Original Message-----
> From: Luben Tuikov [mailto:luben@splentec.com]
> Sent: Wednesday, August 07, 2002 12:19 PM
> To: Julian Satran; roweber@acm.org; ips@ece.cmu.edu
> Subject: Re: iSCSI: plugfest4 issues
> 
> 
> Let's disect:
> 
> I wrote:
> > 
> > 
> > I'd only note that when status is CHECK CONDITION
> > (respoinse = CCT), then residuals SHOULD NOT be set.
> > I.e. the target SHOULD NOT look inside the sense data
> > and retrieve them (EXTENDED COPY exception).
> > This is the responsibility of the ULP at the initiator.
> 
> The only problem I see here is that one may wrongly
> interpret target/initiator. I meant iSCSI target,
> iSCSI initiator, i.e. the transport shouldn't
> peek inside the sense data and mess with SCSI.
> It SHOULD act as a transport only. 
> 
> > As to the future, I cannot imagine a CHECK CONDITION
> > status and residuals reported by any other means
> > than the already established (inside the sense data).
> > So this condition is as strong as MUST, but should
> > be left as SHOULD NOT, and probably not mentioned at
> > all, as is the current matter.
> 
> Well, since reporting residuals with CHECK CONDITION
> is already established as the residual being reported
> IN the body of the sense data, I cannot imagine
> that in the future when status is CHECK CONDITION
> then the residuals will be anywhere else BUT in the
> sense data.
>  
> > Also, a transport (iSCSI) shouldn't be influenced by
> > another transport (FCP), but only by the unifying layer,
> > i.e. SAM-3.
> 
> Well?
> 
> -- 
> Luben
>