FCP-2: Lost Write Data, Not Last Frame of Sequence, Unacknowledged classes.

Dave Peterson dap at cisco.com
Wed Jun 26 22:18:24 PDT 2002


* From the T10 Reflector (t10 at t10.org), posted by:
* "Dave Peterson" <dap at cisco.com>
*
Howdy Santosh,

Comments below...dap

> -----Original Message-----
> From: owner-t10 at t10.org [mailto:owner-t10 at t10.org]On Behalf Of Santosh
> Rao
> Sent: Tuesday, June 25, 2002 7:18 PM
> To: T10 Reflector
> Subject: FCP-2: Lost Write Data, Not Last Frame of Sequence,
> Unacknowledged classes.
>
>
> * From the T10 Reflector (t10 at t10.org), posted by:
> * Santosh Rao <santoshr at cup.hp.com>
> *
> Hello,
>
> We have a question regarding Annexe C Figure C.15 of FCP-2 Rev 07a. This
> figure depicts the FCP-2 SLER (sequence level error recovery) for the
> case of a lost write data iu which is not the last frame in the sequence
> for an un-acknowledged class.
>
> The figure shows the recovery as :
> "Wait REC_TOV. If FCP_DATA is returned, continue with the exchange.
> Otherwise, perform error recovery".
>
> Are the above statements referring to the actions to be taken by the
> initiator or the target ? Since this is a write operation, it is not
> clear how FCP_DATA could be returned to the initiator. (?)
>

The statements refer to the actions to be taken by the initiator.
But, I don't know how FCP_DATA could be returned for a write operation
either.
The note is not correct and should be removed. I think the intent was to
specify the behavior for out-of-order class 3.


> Since the last frame of the sequence is assumed to have made it to the
> target (lost write data iu was not the last frame of sequence), the
> target has the sequence initiative and can complete the exchange with a
> FCP_RSP indicating a CHECK CONDITION status with appropriate sk/asc/ascq
> (as per Section 9.4.1). In this case, is the initiator expected to look
> for CHECK CONDITION command completions of certain sk/asc/ascq
> combinations and initiate a SLER operation on seeing those ? What
> specific sk/asc/ascq combinations should the initiator look for while
> parsing command completions to initiate SLER ?
>
> If this is not done, a write underrun condition as described above
> cannot be recovered from, since SLER will not kick in and the
> ULP/application will see the error.
>
> Can someone clarify what is the error recovery actions taken by the
> initiator and target in this situation ?

For this case:

The target may or may not have sequence initiative
(i.e., seq=1, seq_cnt=1 may not be the last frame of the sequence).

For in-order delivery the target should have detected a Sequence error.

Per clause 12.3.5 Additional error recovery by target:
For unacknowledged classes of service, the target shall not attempt recovery
for Sequence errors. The target
shall depend on initiator time-outs for recovery.

If SLER is enabled, the target should not return check condition.

And we could use some words stating this in the doc...dap

>
> Thanks,
> Santosh Rao
>
>
> --
> The world is so fast that there are days when the person who says
> it can't be done is interrupted by the person who is doing it.
> 	~ Anon
> *
> * For T10 Reflector information, send a message with
> * 'info t10' (no quotes) in the message body to majordomo at t10.org

*
* 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