data IU exception handling
Sriram Srinivasan
srirams at lsil.com
Fri Aug 18 10:32:10 PDT 2000
* From the T10 Reflector (t10 at t10.org), posted by:
* Sriram Srinivasan <srirams at lsil.com>
*
George,
The wordings that I suggested do not MANDATE that retries be made. It was
worded w/ a 'MAY' not a 'SHALL'. This gives targets the option of doing it.
Even if u do not want to put these in, I still see a problem with the wording in
para. 2 of 10.8.3.3.4. The words "before receiving another SPI L_Q information
unit" gives a feeling that this paragraph talks ONLY about doing command IUs
since this is the only time that the target "receives" a SPI L_Q. I still think
this should change.
Sriram
>
>Sriram,
>I disagree. The reason the exception condition section was written the way
>it was, was to removed target retries on IU transfers. The reason is
>because, in real life, targets do not do retries (yes I know tapes do but I
>see that changing in the future and this only applies to packetized). For
>backward compatibility reasons we cannot change the retry option on data
>group transfers but that is not an issue in packietized. So we made it
>simpler and removed (except for the conflict you found) the retry option in
>packetized.
>
>Bye for now,
>George Penokie
>
>Dept Z9V 114-2 N212
>E-Mail: gop at us.ibm.com
>Internal: 553-5208
>External: 507-253-5208 FAX: 507-253-2880
>
>
>
>Sriram Srinivasan <srirams at lsil.com> on 08/18/2000 11:25:16 AM
>
>Please respond to Sriram Srinivasan <srirams at lsil.com>
>
>To: George Penokie/Rochester/IBM at IBMUS
>cc: t10 at t10.org
>Subject: Re: data IU exception handling
>
>
>
>
>George:
>
> I'd suggest that we allow retires for the target. I base my argument on
>the
>following (from SPI-4 rev0):
>
> 1) In 10.8.3.3.7 and 10.8.3.3.6 (data group transfers) the last
>paragraphs
>allow retries. "If the target does not retry transferring the information
>transfer it or exhausts its retry limit the target shall ...". Y should
>this be
>different between data group transfers and IU transfers?
>
> 2) IU transfers already mandate an implied save data pointers (after the
>initiator receives the data IU) and restore pointers (after the initiator
>gets a
>SPI L_Q). So it'be great to retry just the IU in error by allowing the
>target
>to send the MODIFY DATA POINTER message instead of mandating that the
>target
>SHALL do the CHECK CONDITION, ABORTED COMMAND sense key etc.
>
> Based on these, I'd suggest the following change(s) be made in SPI-4:
>
> paragraph 4 of 10.8.3.3.3:
> "If the information unit that failed was not a SPI status information
>unit
>and the message received from the initiator was an INITIATOR DETECTED ERROR
>message then the target <MAY> retry the task associated with the received
>INITTIATOR DETECTED ERROR message. If the target retries the operation, it
><SHALL> send a MODIFY DATA POINTERS message then request that the SPI data
>information unit be transferred again. If the target does not retry the
>operation or it exhausts its retry limit, the target <SHALL> send a SPI
>L_Q/SPI
>status information unit pair to the initiator with a CHECK CONDITION status
>and
>a sense key set to ABORTED COMMAND and an additional sense code set to
>INITIATOR
>DETECTED ERROR MESSAGE RECEIVED for the task associated with the received
>INITTIATOR DETECTED ERROR message."
>
> paragraph 2 of 10.8.3.3.4:
>
> "If the nexus has been fully identified (i.e., an I_T_L_Q nexus has been
>established) and the target detects an iuCRC error in any SPI information
>unit
>it receives while in the DT DATA OUT phase the target <MAY> retry the task
>associated with the iuCRC error. If the target retries the operation, it
><SHALL>
>send a MODIFY DATA POINTERS message then request that the SPI data
>information
>unit be transferred again. If the target does not retry the operation or
>it
>exhausts its retry limit, the target <SHALL>, <<BEFORE RECEIVING ANOTHER
>SPI
>DATA INFORMATION UNIT OR DATA STREAM INFORMATION UNIT IN DT DATA OUT
>PHASE>>,
>switch to a DT DATA IN phase and send a SPI L_Q/SPI status information unit
>pair
>to the initiator with a CHECK CONDITION status and a sense key set to
>ABORTED
>COMMAND and an additional sense code set to iuCRC ERROR DETECTED for the
>task
>associated with the iuCRC error"
>
> The reason I added the "<<BEFORE RECEIVING ...>>" lines was because the
>original wording "before receiving another SPI L_Q information unit" can
>only
>happen if u are talking about receiving SPI L_Qs before command IUs and
>does not
>apply to write transfers where the target sends the SPI L_Qs.
>
> Are there any other reasons to not allow the target from doing retries
>in IU
>transfers?
>
> Sriram
>
>>
>>Sriram,
>>You are correct, the is a conflict between those two sections. The error
>>handling was added in some time after the wording about retries being
>>allowed. I suggest the sentence:
>>'If a target retries an operation it shall send a MODIFY DATA POINTERS
>>message then request that the SPI data information unit be transferred
>>again. '
>>be deleted.
>>John, This should be placed on the agenda for the SPI-4 meeting next week.
>>
>>Bye for now,
>>George Penokie
>>
>>Dept Z9V 114-2 N212
>>E-Mail: gop at us.ibm.com
>>Internal: 553-5208
>>External: 507-253-5208 FAX: 507-253-2880
>>
>>
>>
>>Sriram Srinivasan <srirams at lsil.com> on 08/17/2000 05:24:28 PM
>>
>>Please respond to Sriram Srinivasan <srirams at lsil.com>
>>
>>To: t10 at t10.org
>>cc:
>>Subject: data IU exception handling
>>
>>
>>
>>
>>* From the T10 Reflector (t10 at t10.org), posted by:
>>* Sriram Srinivasan <srirams at lsil.com>
>>*
>> I have a question on data IU exception handling:
>>
>> In 10.8.3.3.3 (SPI4-rev0) paragraph 4 it states:
>>
>> "If the information unit that failed was not a SPI status information
>>unit
>>and the message received from the initiator was an INITIATOR DETECTED
>ERROR
>>message then the target <SHALL> send a SPI L_Q/SPI status information unit
>>pair
>>to the initiator with a CHECK CONDITION status and a sense key set to
>>ABORTED
>>COMMAND and ..."
>>
>> Paragraph 6 of 14.1 states:
>>
>> "The initiator shall save the data pointers as soon as the last byte of
>>the
>>last iuCRC for a SPI information unit is transferred. The save <SHALL>
>>occur
>>even if the initiator detects an error in the SPI data information unit.
>>If a
>>target retries an operation it <SHALL> send a MODIFY DATA POINTERS message
>>then
>>request that the SPI data information unit be transferred again."
>>
>> Doesn't the first one (para 4 in 10.8.3.3.3) say that tha target cannot
>>retry
>>the command ('coz he sent the ABORTED COMMAND sense key)? In this case,
>>the
>>initiator would probably resend the whole command again. But para. 6 of
>>14.1
>>seems to imply that targets can retry the SPI data IU and mandates that
>the
>>MODIFY DATA POINTES message be sent prior to such a retry. These seem to
>>contradict each other. The same wordings are present in SPI-3 as well.
>>
>> Is there a bigger thing that I'm missing here? Please let me know.
>>
>> Thanx,
>> Sriram
>>
>>----------------------------------------------------------------------
>>
>> Sriram Srinivasan Sriram.Srinivasan at lsil.com
>> ASIC Design Engineer, LSI Logic,
>> 2001 Danfield Ct., Phone: 970-206-5847
>> Fort Collins, CO 80525 FAX : 970-206-5244
>>----------------------------------------------------------------------
>>
>>*
>>* For T10 Reflector information, send a message with
>>* 'info t10' (no quotes) in the message body to majordomo at t10.org
>>
>>
>>
>>
>
>=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
> Sriram Srinivasan, | hain aur bhi duniya me sukhanvar bahut ach-che,
> ASIC Design Engineer, | kehte hain ki 'ghalib' ka hai andaaz-e-bayaan aur
> LSI Logic, |
> 2001 Danfield Ct., | Meaning in English
>Ft. Collins, CO 80525 | ------------------
>Phone : (970) 206 5847 | There are a lot of good poets in the world,
>FAX : (970) 206 5244 | But they say that Ghalib's descriptions are
> | a class apart
>
>
>
>
>
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Sriram Srinivasan, | hain aur bhi duniya me sukhanvar bahut ach-che,
ASIC Design Engineer, | kehte hain ki 'ghalib' ka hai andaaz-e-bayaan aur
LSI Logic, |
2001 Danfield Ct., | Meaning in English
Ft. Collins, CO 80525 | ------------------
Phone : (970) 206 5847 | There are a lot of good poets in the world,
FAX : (970) 206 5244 | But they say that Ghalib's descriptions are
| a class apart
*
* 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