FCP-2: Task Retry Identification

Baldwin, David Dave.Baldwin at emulex.com
Fri Jun 28 17:35:08 PDT 2002


* From the T10 Reflector (t10 at t10.org), posted by:
* "Baldwin, David" <Dave.Baldwin at Emulex.Com>
*
Santosh,

I completely agree with you. Task Retry Identification must be mandatory or
implementations will be exposed to potential data corruption. This must be
fixed in FCP-3.

At Emulex, our implementation will not run FCP-2 SLER on a target that does
not support Task Retry Identification. We will fall back to FCP exchange
level recovery. We are aware of some FCP-2 target implementations that
support Task Retry Identification, and some that do not. As customers become
aware of these details, they stop buying the FCP-2 devices that don't
include Task Retry Identification support.

Regards,
Dave Baldwin
Emulex Corporation

-----Original Message-----
From: Santosh Rao [mailto:santoshr at cup.hp.com]
Sent: Friday, June 28, 2002 3:47 PM
To: T10 Reflector; Dave Peterson; Robert Snively
Subject: [Fwd: FCP-2: Task Retry Identification.]


* From the T10 Reflector (t10 at t10.org), posted by:
* Santosh Rao <santoshr at cup.hp.com>
*
Hello,

I am re-sending this message since I have'nt been able to get any
response on this. Clarifications from FCP-2 editors and implementors
would be appreciated.

Thanks,
Santosh


Santosh Rao wrote:
> 
> * From the T10 Reflector (t10 at t10.org), posted by:
> * Santosh Rao <santoshr at cup.hp.com>
> *
> Hello,
> 
> We have a question regarding the Task Retry Identification functionality
> described in Section 4.6.
> 
> The re-use of OX_IDs within RR_TOV of their last use exposes the
> initiator to this condition described in Section 4.6.
> 
> Since it is the initiator's OX_ID generation model which can determine
> the exposure to this problem, the Task Retry Identification
> functionality must be a "mandatory to implement, optional to use"
> feature and it should be left to the initiator's discretion on whether
> this should be enabled.
> 
> Can someone clarify why this is not a mandatory feature of FCP-2 ?
> Without this feature, there is a risk of exposure to data corruption and
> FCP-2 sequence level error recovery (SLER) cannot be used in a reliable
> manner.
> 
> Can the FCP-2 target implementors on this list comment on whether they
> support Task Retry Identification ? What is the extent of support for
> this feature among FCP-2 target implementations ?
> 
> Thanks,
> Santosh
> 
> --
*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org




More information about the T10 mailing list