Re: lun reset and r2t error handling

Julian Satran <[email protected]> Wed, 12 Aug 2009 06:46:41 +0300
Newsgroups gmane.ietf.ips,gmane.linux.iscsi.open-iscsi
Message-ID <OF3BD6AD19.DD34E8FC-ONC2257610.00145B55-C2257610.0014C189@il.ibm.com>
This is a multipart message in MIME format.
--===============1188894391==
Content-Type: multipart/alternative; boundary="=_alternative 0014BFBAC2257610_="

This is a multipart message in MIME format.
--=_alternative 0014BFBAC2257610_=
Content-Type: text/plain; charset="US-ASCII"

some answers in text

[email protected] wrote on 11/08/2009 23:56:10:

> From:
> 
> Mike Christie <[email protected]>
> 
> To:
> 
> [email protected], Hannes Reinecke <[email protected]>, open-iscsi <open-
> [email protected]>
> 
> Date:
> 
> 11/08/2009 23:57
> 
> Subject:
> 
> [Ips] lun reset and r2t error handling
> 
> Sent by:
> 
> [email protected]
> 
> Hi,
> 
> For a single connection session with ERL=0 and without FastAbort, if the 

> initiator has sent a lun reset task management function and the target 
> has sent a R2T, is it ok for the target to send a task management 
> response with Function Complete, before the initiator has sent the 
> data-out pdus for the R2T? It looks like the reason the target will do 
> this is due to a internal target timeout (target did not get the 
> data-outs within some timeout period).
> 
> If the target does return the task management function with Function 
> Complete, should the initiator continue to respond to the R2Ts?
> 


The initiator doesn't have to send more data but since more data may be 
still in flight
it shouldn't harm to send a "terminating" Dataout. Target should behave OK 
in both cases.

> And one other question. In section 10.5.1, we have:
> 
>     The issuing initiator SHOULD however terminate (i.e., by setting the
>     F-bit to 1) these response sequences as quickly as possible.
> 
> Does this mean if we have sent a lun reset, and the target has sent a 
> R2T, should we be setting the F-bit in the continued data-out PDU so as 
> to end the transfer, even though actual data transfer has not been 
> completed entirely?
Yes
> 
> What if we do send all the data like normal, should that still be ok?
Yes - receivers are supposed to be lenient.
> _______________________________________________
> Ips mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/ips

--=_alternative 0014BFBAC2257610_=
Content-Type: text/html; charset="US-ASCII"

<font size=2 face="sans-serif">some answers in text</font>
<br>
<br><tt><font size=2>[email protected] wrote on 11/08/2009 23:56:10:<br>
<br>
&gt; From:</font></tt>
<br><tt><font size=2>&gt; <br>
&gt; Mike Christie &lt;[email protected]&gt;</font></tt>
<br><tt><font size=2>&gt; <br>
&gt; To:</font></tt>
<br><tt><font size=2>&gt; <br>
&gt; [email protected], Hannes Reinecke &lt;[email protected]&gt;, open-iscsi &lt;open-<br>
&gt; [email protected]&gt;</font></tt>
<br><tt><font size=2>&gt; <br>
&gt; Date:</font></tt>
<br><tt><font size=2>&gt; <br>
&gt; 11/08/2009 23:57</font></tt>
<br><tt><font size=2>&gt; <br>
&gt; Subject:</font></tt>
<br><tt><font size=2>&gt; <br>
&gt; [Ips] lun reset and r2t error handling</font></tt>
<br><tt><font size=2>&gt; <br>
&gt; Sent by:</font></tt>
<br><tt><font size=2>&gt; <br>
&gt; [email protected]</font></tt>
<br><tt><font size=2>&gt; <br>
&gt; Hi,<br>
&gt; <br>
&gt; For a single connection session with ERL=0 and without FastAbort,
if the <br>
&gt; initiator has sent a lun reset task management function and the target
<br>
&gt; has sent a R2T, is it ok for the target to send a task management
<br>
&gt; response with Function Complete, before the initiator has sent the
<br>
&gt; data-out pdus for the R2T? It looks like the reason the target will
do <br>
&gt; this is due to a internal target timeout (target did not get the <br>
&gt; data-outs within some timeout period).<br>
&gt; <br>
&gt; If the target does return the task management function with Function
<br>
&gt; Complete, should the initiator continue to respond to the R2Ts?<br>
&gt; </font></tt>
<br>
<br>
<br><tt><font size=2>The initiator doesn't have to send more data but since
more data may be still in flight</font></tt>
<br><tt><font size=2>it shouldn't harm to send a &quot;terminating&quot;
Dataout. Target should behave OK in both cases.</font></tt>
<br><tt><font size=2><br>
&gt; And one other question. In section 10.5.1, we have:<br>
&gt; <br>
&gt; &nbsp; &nbsp; The issuing initiator SHOULD however terminate (i.e.,
by setting the<br>
&gt; &nbsp; &nbsp; F-bit to 1) these response sequences as quickly as possible.<br>
&gt; <br>
&gt; Does this mean if we have sent a lun reset, and the target has sent
a <br>
&gt; R2T, should we be setting the F-bit in the continued data-out PDU
so as <br>
&gt; to end the transfer, even though actual data transfer has not been
<br>
&gt; completed entirely?</font></tt>
<br><tt><font size=2>Yes<br>
&gt; <br>
&gt; What if we do send all the data like normal, should that still be
ok?</font></tt>
<br><tt><font size=2>Yes - receivers are supposed to be lenient.<br>
&gt; _______________________________________________<br>
&gt; Ips mailing list<br>
&gt; [email protected]<br>
&gt; </font></tt><a href=https://www.ietf.org/mailman/listinfo/ips><tt><font size=2>https://www.ietf.org/mailman/listinfo/ips</font></tt></a><tt><font size=2><br>
</font></tt>
--=_alternative 0014BFBAC2257610_=--

--===============1188894391==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
Ips mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ips

--===============1188894391==--