FC-TAPE Minutes, 4 Aug 1999, Rochester MN

WYATT,STEWART (HP-Boise,ex1) stewart_wyatt at am.exch.hp.com
Fri Aug 6 11:00:58 PDT 1999


* From the T10 Reflector (t10 at t10.org), posted by:
* "WYATT,STEWART (HP-Boise,ex1)" <stewart_wyatt at am.exch.hp.com>
*
Joint T11/T10 TAPE AdHoc Meeting, August 4, 1999 - Rocehster, MN
T11/99-471v0
Stewart Wyatt, FC_TAPE Secretary, 208-396-3594, stewart_wyatt at hp.com

1.	Introductions: Group 

The facilitator, Dale LaFollette, called the meeting to order at 5 PM and,
as is customary, had the participants introduce themselves.

2.	Approval of this Agenda: T11/99-429v0 - Group

The agenda was approved with a number of additions and a few changes that
are included in these minutes

3.	Approval of 7/13/99 Minutes: T11/99-411v1 - Stewart Wyatt

Approved.

4.	Review Old Action Items: Stewart Wyatt

#1. HP (Steve Jerman): to continue MAM discussion and post presentation. -
Ongoing
#2. Eric Oetting - to collect issues from the SCC letter ballot comments for
the SPC-2. - Deferred until next month.
#3. Dave Peterson: - to update the SSC. - Completed
#4. Stephen Gold (Stewart Wyatt) HP: Resolve IBM letter ballot comments on
TapeAlert. - Ongoing 
#5. Stewart Wyatt: Report on FC Tape Connector resolution. Completed
#6. Bob Snively and Bob Kembel: Create a list of the specific BLS and ELS
which are allowed before Login, including explicit Login, in an annex for
later placement in the FC-FS. - Ongoing
#7. Bob Snively: Review the SRR statement, it may not be flexible enough for
the target. SRR must be properly interpreted in context and command known by
the target. "Nasty case is did the transfer ready or the response (to an
error) get hit." Deferred
#8. Bob Snively: Return clause 2.8 to previous text. - Deferred
#9. Bob Snively: Change Table 2 - Discovery of FCP capabilities entry
"Initiator performs REC", Discovery mechanism needs to be changed from
"None/Process Login" to "See References". - Deferred
#10. Dave Peterson and group: Extend current document to cover LIP discovery
process. Participants asked to communicate concerns to Dave. - Ongoing
#11. Carl Zeitler: Review need for process associators. - Due in October.

5.	SSC: T10 Working Drafts SSC-R18, LB Comment Resolution T10/99-011r3
- Dave Peterson

Will finish by next month. George has asked all SPC references be changed to
SPC-2 and all SCSI references should drop the -3 after SCSI. (Use SCSI
instead of SCSI-3)

6.	FCP-2: T10 Working Drafts FCPR02, Changes to FCP-2, T10/99-211r1 -
Bob Snively

Bob reviewed Charles Binford's "two and three star" comments from a file
called LSI.PDF. (Charles had graded his comments in order of importance by
assigning them one, two or three stars.)

#004  FCP restricts FCP_CONF to queuing. Charles felt this was unnecessarily
restricting. Bob felt that the initiator should know when it could expect
that the target could request a confirm. No change

#006 LUN reset should be added to the table 3. Bob agreed 

#007 FCP actions should only affect FCP exchanges. Bob agreed.

#008 Charles would like to reserve a device with dual ports where both ports
were included in the reservation. He thought this could be facilitated by
using the WWN for reservation. This led to a long discussion with the
problem that the address space doesn't cross domains. Charles' request is an
architectural issue in violation of SAM. Bob agreed that the concept had
merit and wanted to continue the discussion later.

#010 Confirmed Completion allowed. Charles thought the title was misleading
especially in view of comment #004.  Bob said he may review the title.

#012 page 30, 7.1, 1st paragraph, "the target shall not reject requests for
retransmission of FCP_XFER_RDY or FCP-RSP". Matthew Wakeley felt that this
was too strong of wording. Bob felt that if the device supports the error
recovery, it should always comply. It may be necessary to separately specify
FCP_RSP and FCP_XFER_RDY. The issue was not resolved.

#014 Bus Inactivity limit and #015 Disconnect time limit. Charles thought
these two parameters were specified inconsistently. George thought the
specification was taken from earlier SCSI documents. Bob will review to them
to make sure they are consistent with previous SCSI documents.

#018 Why is RR_TOV so huge?  Clause 10.3 change "must" to "shall" last
paragraph of 10.3. Add default value for RR_TOV for non REC_TOV error
recovery.

#022 page 57 11.1.2 second to last paragraph - what should a target do if it
receives an ABTS? Bob agreed to do research.

#027 REC data count is over exchange not sequence. Check wording.

#029 page 58, 11.2.5 FCP_RSP Recovery. Charles questioned the statement that
a target shall retain exchange information until the queue depth is
exceeded.  Charles was concerned that the tape requirements are being
imposed on disks. Jim Coomes agreed with Charles concern. A long discussion
with many tangents followed. Clause 11 needs a statement to make sure that
optional error recovery behavior is clearly noted. The end result of the
discussion is that if a target is queuing and not using FCP_CONF, the
exchange information retained by the target can only be vendor specific or
undefined. A proposal to only keep bad status was rejected since it would
not allow differentiating a successfully completed command whose status had
been discarded from a command that was lost.

Following is a summary of the cases of where the target can clearly know
when to discard completed exchange data in regard to using FCP_CONF:

Class 2 works with and without FCP_CONF - FCP-CONF is optional.
Class 3 error retry and queuing, FCP_CONF is required, keep all FCP_RSP not
confirmed.
Class 3 no error retry, queue or not, without FCP_CONF (Bob: retry optional
and random results)

The discussion went down a significant tangent about moving the FC-TAPE
developed error recovery approach into the FCP-2.  The problem is that it
appeared that tape error recovery requirements were being imposed on disk
drives. Also what subset of the tape error requirements could a disk adopt
without breaking it. There was a concern that the FCP was making guidelines
a part of a standard. The discussion focused on the effect on disk drives
that may want to use FCP_CONF but don't want the restrictions that Bob has
required the error recovery. Bob was restricting the use of FCP_CONF to the
applications described in a model that was developed last year. This limited
FCP_CONF to bad status responses and queuing. Ed Gardner and Dal Allan
argued that the target should be able to ask for the CONF at its discretion
and that an initiator should always comply.  A straw poll vote was taken
with 14 in favor of the change and 2 in favor of retaining the current
wording. Bob asked Dal for a new model to justify the change.

#033 Target needs to provide a unique RX_ID when FCP_CONF is used. RX_ID =
FFFFh invalid, class independent.

6.1	Jim Coomes: 99-226r0

Jim's proposal was to clarify the Fibre Channel Control Mode page 19,
Disable Soft Address (DSA). The clarification limits address selection to
the hard address phase only (LIHA). Otherwise the device goes into
nonparticipating mode. The intent of this bit was to not only to preclude
the device from choosing an address during the soft address phase (LISA) but
to preclude it from selecting its address during the previously assigned
address phase (LIPA).

Matt Wakeley was concerned that this would break some existing
implementations. Jim notes that this is the way that Seagate implemented
this function. Bob Snively thought that this was the way most disks were
implemented. Jeff Williams asked for a delay to check the impact on his
design. This started a long discussion which resulted in a straw poll with
the result that 8 voted to oppose a delay and 7 voted to allow delay.  A
second straw poll was taken asking if the change should be included in the
next revision of the FCP (rev 3). The results were 7 for, 7 opposed. Another
heated discussion followed about whether this was a technical or editorial
change.  There was also a proposal to change the name to something like,
"Select only Hard Address."

6.2	Charles Binford: LSI.PDF

While the facilitator added this as a separate agenda item, it was actually
covered in item #6, FCP review.

6.3	MCM George Penokie: T11/99-206r0

George proposed adding three fields to Fibre Channel Control page 19 for
controlling MCM applications. 
Bob Snively objected to using mode pages because the applications are not
FCP functions and apply to other protocols as well. He thought they should
be controlled by the use of either an ELS or a login procedure. Jeff
Williams noted that the other bits in this page are also not necessarily
SCSI specific either and failed to see Bob's distinction. Bob Kembel raised
a concern that these behaviors will affect protocols other than SCSI. Dal
Allan spoke in favor of George's proposal by noting that there was no
precedence for using ELS or login procedures for these controlling this type
of parameters. SCSI has traditionally used mode pages to tune operations. If
other protocols want these controls let them work out a procedure. Ed
Gardner encouraged those who want to use an ELS for this purpose to make a
proposal that could be reviewed. George said he would be revising the
proposal with input from an earlier meeting where the proposal was also
presented. Bob raised a second issue about the proposal in that it extended
the length of the page. George noted that the page length is part of the
payload.  Bob wanted to make additional page length optional, as he is
concerned about compliance to FCP-2.

7.	Media Auxiliary Memory T10/99-223r1 - Sid Crighton

Last month Steve Jerman made a presentation. As a result of discussion
during and after the meeting the proposal was modified. Sid Crighton gave a
quick overview of the old proposal and described the changes that have been
made. The changes included dropping the use of Read Element Status and the
use of Log page 0Ah. The terminology is changed from "parameters" to
"attributes". The addition includes two new commands for the SPC-2, Write
Attribute and Read Attribute. The Inquiry VPD page is to be retained to
allow read access for reserved devices.

The justification of moving the documentation to the SPC is that this
capability has application to any type of removal media device.

8.	Tape Connector Update SFF8072 - Stewart Wyatt

Stewart reviewed the changes made to the proposal. The 12 Volt pins had been
reduced from 8 to 5 to improve compatibility with the SCSI connector with
the understanding that the pins were rated more than 1 Amp each. Dal Allan
reported that the pins were capable of handling 1.5 Amps currently. This
could be improved by changing the material of the connector.

8.1	Curt Ridgeway 99-410v1

Curt had a counter proposal to the one made by Stewart. Curt's proposal was
for a "fat pipe" proposal of 8 channels on the same connector to enable
improved performance by striping the data transfer across 4 dual port
channels. While Curt was successful in demonstrating that it was possible,
he agreed to rework his proposal with a 100 pin connector to provide some
spare pins and to avoid slowing down the tape connector proposal which is a
critical path activity for the tape manufacturers.

9.	T11 New Business: Group

None

10.	T10 New Business: Group

None

11.	Next Meeting Requirements: Group

Four hours minimum

12.	Review New Action Items: Stewart Wyatt

Old forwarded action items

#1. Eric Oetting - to collect issues from the SCC letter ballot comments for
the SPC-2. - Deferred until next month.
#2. Stephen Gold (Stewart Wyatt) HP: Resolve IBM letter ballot comments on
TapeAlert. - Ongoing 
#3. Bob Snively and Bob Kembel: Create a list of the specific BLS and ELS
which are allowed before Login including explicit Login in an annex for
later placement in the FC-FS. - Ongoing
#4. Bob Snively: Review the SRR statement, it may not be flexible enough for
the target. SRR must be properly interpreted in context and command known by
the target. "Nasty case is did the transfer ready or the or response (to an
error) get hit." Deferred
#5. Bob Snively: Return clause 2.8 to previous text. - Deferred
#6. Bob Snively: Change Table 2 - Discovery of FCP capabilities entry
"Initiator performs REC", Discovery mechanism needs to be changed from
"None/Process Login" to "See References". - Deferred
#7. Dave Peterson and group: Extend current document to cover LIP discovery
process. Participants asked to communicate concerns to Dave. - Ongoing
#8. Carl Zeitler: Review need for process associators. - Due in October.

New Action Items

#1. HP (Steve Jerman/Sid Crighton): Continue MAM discussion. Post
presentation made today. Contact Ralph Weber about including MAM in the
SPC-2. Make presentation to SCSI working group next month. Resolve issues
with Sony AIT per Joe Breher's concerns. Determine need, if any, for content
to be placed in the SSC.
#2. Bob Snively: (Binford FCP comment #006) LUN reset should be added to the
table 3. 
#3. Bob Snively: (Binford FCP comment #007) FCP actions should only affect
FCP exchanges. 
#4. Bob Snively: (Binford FCP comment #014 and #015) Check to see that Bus
Inactivity limit and  Disconnect time limit are specified consistently with
other SCSI standards.
#5. Bob Snively: (Binford FCP comment #022) Define what a target shall do
when it receives a ABTS. 
#6. Bob Snively: (Binford FCP comment #027) Check wording to see that REC
data is counted is over an exchange not a sequence.
#7. Bob Snively: (Binford FCP comment #029) Optional error recovery
procedures need to be clearly marked. The requirements for using error
recovery procedures in class 3 need to be clearly specified for disk drives.
A target may request an FCP_CONF at its disgression.
#8. Bob Snively: (Binford FCP comment #033) Target must use valid RX_ID (<>
FFFFh) when using FCP_CONF.
#9. Bob Snively: (Binford FCP comment #018) Add a default value for RR_TOV
for non REC_TOV error recovery.
#10. Bob Snively: Binford FCP comments #008, using WWN for reservations, and
#010, title for Confirmed Completion Allowed, appeared to be unresolved
without specific action items. In addition the proposal by Jim Coomes to
clarify the DSA field appeared unresolved as well as a proposal to rename it
more appropriately.

13.	Adjournment: Group - At 9:20

Attendance:

Dale Lafollette	Storage Tek	Horst Truestedt	True Focus
Cutis Ridgeway	LSI	Charles Binford	LSI Logic
Dave Ford	Clariion	Arlan Stone	UNISYS
Sid Crighton	HP	Joe Breher	Exabyte
Gile Frazier	IBM	Damian Bannon	SSL
Bill Lynn	Adaptec	Shinya Iguchi	Hitachi
Mark Dewilde	Pathlight	Matt Gaffney	StorageTek
Mike Fitzpatrick	Fujitsu CPA	Edward A. Gardner	Ophidian
Designs
Pak Seto	Quantum	Matt Wakeley	HP
Jim Coomes	Seagate	Bill Martin	Gadzoox Networks
Stewart Wyatt	HP	Danny Ybarra	TI
Paul Suhler	Seagate	Bob Kembel	Connectivity Solutions
Carl Zeitler	Compaq	Paul Entzel	Quantum
Dave Baldwin	Emulex	John Lohmeyer	LSI Logic
Dal Allan	ENDL	George Penokie	IBM
Jeff Williams	WD	Dennis Moore	KnowledgeTek			
*
* 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