@Date: July 14, 1988 To: X3T9.2 Membership From: John B. Lohmeyer, X3T9.2 Chairman Subject: July 1988 Working Group Meeting Minutes Dave McIntyre of Hewlett Packard hosted a working group meeting July 11-13, 1988 in Boise, ID at the (Compri) Hotel. I would like to thank Dave and Hewlett Packard for hosting a very productive meeting in a beautiful setting. The final agenda was: Agenda Items for SCSI-2 1. [] bracket removal 2. Review of S2R4 3. WORM Device Specific MODE SELECT Page [88-44] (P. Boulay) 4. Pin assignments for high density connector B [88-77] (J. Morse) 5. ACCESS LOG proposals [88-006R1] (McGrath, Spence) 6. RECOVER BUFFERED DATA command for sequential access devices [88-022] (Appleyard) 7. ASC's for Processor Devices [88-063] (Stai) 8. Processor Device Model [88-064] (Stai) 9. Wide Data Transfer Request (Goldman) 10. AEN Revision [88-065R1] (Boulay) 11. ASC Qualifier Codes [88-070] (Anderson) 12. CD-ROM Minute-Second-Frame debate (Boulay/Lamers) 13. Should we specify a Power-on to Selection time? (Lohmeyer) 14. Comments on S2R4 - (Snively) [88-79] 15. SCSI Device Connector Layout (Marazas) [88-80] 16. Sense Key - ASC - ASCQ Hierarchy (Goldman) [88-82] Unassigned Agenda Items 40. DEC proposal for added terminator power lines [88-048R1] (Morse) 41. IBM proposal on SCSI Physical Architecture [88-069] (Marazas) {Monday afternoon} Agenda Items for SCSI-3 50. SEND DIAGNOSTIC Pages [87-186] (Spence) 51. LOAD SKIP MASK command proposal [87-203] [87-217] (Floryance, McIntyre) 52. More that 8 devices on wide SCSI [87-206] (Harms) 53. Search Command modifications [88-002] (Stai) 54. Expanded RelAdr Bit [88-007] (Boulay) The following people attended the meeting: Name Status Organization ------------------------------ ------ ------------------------------ Mr. Robert N. Snively P Adaptec, Inc. Mr. Charles Brill P AMP, Inc. Mr. Bob Whiteman A AMP, Inc. Mr. Ken Post A Archive Corp. Mr. James J. Semenak P AT&T Mr. Gregory Fry O Ballard Synergy, Inc. Mr. Dan Davies P Cipher Data Products, Inc. Mr. Gene Milligan P Control Data Corp. Mr. Gerald Houlder A Control Data Corp. Mr. Gerald Maurer P Emulex Corp. Mr. Joel Urban O Fujitsu Mr. Robert Liu P Fujitsu Components of America Mr. Dave McIntyre A Hewlett Packard Co. Mr. Bob Pentecost O Hewlett Packard Co. Mr. Kevin C. Hess V Hewlett Packard Co. Mr. George Penokie P IBM Corp. Mr. Gerald Marazas A IBM Corp. Mr. David A. Buesing O IBM Corp. Mr. William Homans P LMSI-CPI Mr. Robert Johnston S LMSI-CPI Mr. Paul Boulay A LMSI-OSI Mr. Kim Ryal A Miniscribe Corp. Mr. James Schuessler P National Semiconductor Mr. John Lohmeyer P NCR Corp. Mr. Dennis P. Trupski O Olivetti Mr. Jim Harper O Optotech, Inc. Mr. Paul Nitza P OTL Engineering Mr. James McGrath P Quantum Corp. Mr. Dexter Anderson P Scientific Micro Systems Mr. Lawrence J. Lamers P Sony Corp. of America Mr. D. W. Spence P Texas Instruments Mr. Jeff Stai P Western Digital Status Key: P - Principal A - Alternate O - Observer S - Special Interest V - Visitor The following new documents were distributed at the meeting: Document Doc Date Author Description of Document ------------- -------- --------------- --------------------------------------- X3T9.2/88-6 7/11/88 J. McGrath LOG SELECT and LOG SENSE commands Rev 2 X3T9.2/88-65 7/11/88 P. Boulay Asynchronous Event Notification Rev 1 Revision X3T9.2/88-79 7/7/88 B. Snively Review of SCSI-2 Rev 4 X3T9.2/88-80 7/7/88 G. Marazas SCSI Device Connector Layout X3T9.2/88-81 7/12/88 B. Spence LOG SELECT/SENSE Pages X3T9.2/88-82 7/12/88 S. Goldman Sense Key - ASC - ASCQ Hierarchy Agenda Items for SCSI-2 1. [] bracket removal 2. Review of S2R4 These two agenda items were covered together beginning on Wednesday. Most of the details were recorded by marking-up the editor's copies of S2R4 and will be incorporated in S2R5, which should be available at the Colorado Springs meeting. Two changes are worth noting here: Bit(4) of the IDENTIFY message has been moved to Bit(5). Targets should treat unsupported target processes the same as invalid logical units. A new page for all device types will be generated by Jim Semenak based on Paul Boulay's proposal and incorporate the ECA enable bit. [During the subsequent editing session, the editors merged several other pages together for control of queuing, AEN, and ECA.] 3. WORM Device Specific MODE SELECT Page [88-44] (P. Boulay) This proposal was accepted for S2R5. There was some discussion of how the update block bit (RUBR bit) affects the device model. The default state and whether or not data is transferred before the error is reported needs to be resolved. Jeff, Paul, and Larry will work on it. 4. Pin assignments for high density connector B [88-77] (J. Morse) The DEC proposal is to plan for more terminator power lines being added to the interface in the future (i.e., SCSI-3) by defining some of the current GROUND lines as OPEN lines. 88-77 modifies the DEC proposal by replacing the OPEN definition of 10K ohms with terminating these lines (at the terminator only) with either GROUND or the normal termination (i.e., 220 ohm - 330 ohm). Bob Snively commented that there is an implied unwritten agenda in this proposal: lengthening the single-ended cables, increasing driver power, lowering impedance, and smaller wire diameter. He objected to changing SCSI-2 on the assumption that we might want to make these changes in SCSI-3. He also objected on the grounds that current devices are grounding these lines and if the DEC proposal is accepted, then these devices would have to change to open. John Lohmeyer stated that accepting this proposal would serve notice to developers that changes may occur in the physical interface in SCSI-3 and may allow some degree of compatibility between SCSI-2 and SCSI-3. Whereas, not accepting this proposal would would make it very difficult to have compatibility between SCSI-2 and SCSI-3 devices, if make the assumed changes in SCSI-3. There were also some concerns regarding UL when using shared power lines: What happens if some of the shared terminator power lines get cut? Will we exceed the power limitations of CSA? The conclusion of the working group was to wait for DEC to defend their proposal at Colorado Springs. Thus it will not be included in S2R5, except for the B cable where 4 terminator power lines were included. 5. LOG SELECT/LOG SENSE proposal [88-006R2 and 88-081] (McGrath, Spence) The LOG SELECT command operates somewhat differently than the MODE SELECT command. Variable length parameter fields are allowed. Jim spent considerable time explaining the operation of these commands. Several minor modifications were proposed which Jim McGrath agreed to incorporate into a revised proposal. Bill Spence presented a rough draft of the LOG pages. His intention is to get the pages in shape for inclusion into S2R5. 6. RECOVER BUFFERED DATA command for sequential access devices [88-022] (Appleyard) Dan Davies made revised Dennis Appleyard's proposal to overcome the previous objections. He had reached agreement with several of the key people involved with sequential-access devices, except he had been unable to reach Dennis. The working group accepted the revised proposal to be incorporated in S2R5. 7. ASCs for Processor Devices [88-063] (Stai) Several ASCs were modified to use the words "logical unit" instead of "peripheral device". Two new ASCs were added for Processor Devices. It was pointed out during the discussion of terminology that the wording in Appendix A needs to be fixed to match the definitions of "target" and "peripheral device" that are in section 3. 8. Processor Device Model [88-064] (Stai) Jeff did not have his document ready to discuss. 9. Wide Data Transfer Request (Goldman) This item had been added to the working group agenda at the Irvine meeting because Dal had questioned why the processes for synchronous and wide negotiation are different. Steve had responded at St. Petersburg that the reason for the differences is that synchronous negotiates two parameters (REQ/ACK offset & transfer period) while wide only negotiates one parameter (bus width). This item will be dropped from future agenda due to lack of response from either Dal or Steve. 10. AEN Revision [88-065R1] (Boulay) This document was discussed and Paul agreed to prepare another revision to fix a few problems. 11. ASC Qualifier Codes [88-070] (Anderson) The proposal would define several generic ASCQs that apply to all ASCs. While most people agreed that we should have consistency across similar ASCQs, there was a concern that many of the generic ASCQs actually didn't apply to all of the ASCs. Also, there was a question of how to document this: two tables (one for ASCs and one for ASCQs) or expand the current table to have all of the common ASCQs under each ASC. The first case does not lend itself to easily adding unique ASCQs and can imply impossible combinations of ASCs and ASCQs are to be supported. The second case can lead to inconsistencies and can get very large if we merely put all possible ASCQs under each ASC. After much discussion, the working group asked Dexter to re-propose specific substitutions of ASCQ codes and specific additions, using the current documentation structure. We will attempt to maintain consistency of ASCQ codes using the current structure. During the discussion of this item it was noted that the St. Petersburg plenary passed a motion to include information on the sense key, ASC, and ASCQ relationship. This was included in 88-040R1, section 8.3.6. Part of the motion requested Steve Goldman to provide additional wording which we had not yet received. Bill Spence called Steve to check on his progress on this action item. Steve sent Bill a FAX (88-082R0) to complete his action item (see agenda item 16). Page 7-68 of S2R4 needs to define which ASC and ASCQ codes are vendor unique. 12. CD-ROM Minute-Second-Frame debate (Boulay/Lamers) Paul Boulay had a limited number of copies of an interim revision of section 13 available for review and comment. Paul's revised section 13 was accepted for S2R5. 13. Should we specify a Power-on to Selection time? (Lohmeyer) SCSI devices in some cases do not honor selections following power-up or reset for several seconds. A new school of thought is to enable the protocol chip early in the power-up sequence and respond with assertion of the BSY signal (but not necessarily proceeding with the I/O until power-up processing is complete). Appendix H suggests that SCSI devices should respond to selections within 250 ms following application of power. The question was, do we really want to define a minimum time to be ready for selections? If so, it should go into section 4 -- not an appendix. A question was also raised on whether there should be a minimum time following a reset condition that all devices should stay off of the bus. John Lohmeyer stated that he had recommended that devices stay off at least 100 ms. This allows time for any transients to go away following a device powering off. However, many devices do not reset the bus as their power fails. No real agreement could be reached on either of these issues because different applications have different requirements. Appendix H should be fixed to reflect that SCSI does not have a requirement on honoring selections following reset. It should be noted that autoconfiguration such as proposed by IBM in 88-069 may require some times be defined. 14. Comments on S2R4 (Snively) [88-79] Item 1) Accepted, add to section 1 and 2. Item 2) 'Logical thread' wording is not acceptable. It will be replaced with I_T pairing, I_T_L pairing, I_T_P pairing, I_T_L_Q pairing, and I_T_P_Q pairing. (I = Initiator, T = Target, L = Logical unit, P = Target Process, and Q = Queue Tag.) We need to add back last deleted sentence as modified. Item 3) We need to develop model for section 6 and include the queuing examples. A new ASC (2F 00) was defined for 'TAGGED COMMANDS CLEARED BY ANOTHER INITIATOR'. 15. SCSI Device Connector Layout (Marazas) [88-80] This document illustrates preferred positioning and orientation for the power and A cable connectors on peripheral devices. The preferred position for the B cable is not shown. The preferred vertical orientation (left or right) is not addressed. Jerry said that the intent of the proposal was to solicit comments and to develop a proposed appendix which would be advisory only. There was concern over whether we could get a consensus over the correct orientation. The working group also recommended that Jerry neuter the power connector as several different power connectors are currently in use and this is not within the scope of the SCSI standard. [At one point during the discussion of this issue, Jeff Stai went spastic, getting queued up, however he was reset by the pink mascot.] Bob Snively asked the working group to recommend to the plenary that we incorporate the proposal into SCSI-2 as an appendix. A straw poll resulted in 6 in favor, 13 opposed. Gene Milligan asked that we recommended a "SCSI-2 or never" decision be made and not put off consideration to SCSI-3. A Straw poll resulted in 18 in favor, 1 opposed. Jerry plans to revise the proposal and bring it back to Colorado Springs for further consideration. 16. Sense Key - ASC - ASCQ Hierarchy (Goldman) [88-82] Steve Goldman sent a FAX to Bill Spence at the working group meeting containing wording describing the original intent of the sense keys, ASCs, and ASCQs. The editors accepted an action item to incorporate wording in S2R5 describing the current usage of these fields. Unassigned Agenda Items 40. DEC proposal for added terminator power lines [88-048R1] (Morse) (See agenda item 4.) 41. IBM proposal on SCSI Physical Architecture [88-069] (Marazas) {Monday afternoon} Jerry made a presentation to aid the understanding of his proposal and to clarify some misunderstandings. The proposal is made by Entry Systems Division in regard to the IBM PS/2. The proposal provides an optional extension to SCSI-2 (base SCSI in IBM terminology). The additional leads are intended to be in the A cable. It is intended that both single-ended and differential systems be supported. The skid concept used to mount 3.5-inch drives in PS/2 systems reduces accessibility of switches/jumpers. The proposal provides the ability to electronically read and over-write the SCSI device ID. Ten additional leads are required to do this. Lines 59 (Config Request In) and 60 (Config Request Pass) seem to have the most confusion associated with them. This is intended to function in a multi-initiator system subject to a private agreement of who is first among equals. An external box would need an input connector and an output connector because of the way the in/out of the signal is propagated. This method is used to reconfigure one device at a time. It is possible to re-configure a system without powering down if the bus is properly designed. One or more SCSI devices within external sub-system box(es) can be (re-)configured. However the Jerry's diagram did not show the extra ten signals internal to the PS/2. How is the system configured when the 'agency' (that is, host) that is configuring the system is in the middle? Maybe if the two signals are looked at as both in & out from the host's viewpoint and each signal is driven in turn to look at devices to the left and the right. However this does assume a master host. New devices (those that support autoconfiguration) can not be used until they are configured. New devices are placed closer to the system box and older devices (those that do not support autoconfiguration) are last in the logical chain. Including old devices defeats the external terminator sensing. Can a 50-pin solution be found and would it be acceptable to IBM? Jerry indicated that IBM might take it under consideration if it did not adversely affect existing silicon and worked with existing devices. However, Jerry seemed dubious that a 50-pin solution would actually work without extreme complexity. Several members of the committee indicated that they wanted to investigate the 50-pin solution further as it would have considerable advantages if we could avoid changes to the physical layer. John passed out a list of possible requirements related to autoconfiguration which were developed over the last two weeks on the SCSI BBS. He said that it was unlikely that any proposal would meet all of the possible requirements, but that it would help in judging the merits of proposals. The group added several requirements and rated the importance of the requirements on a scale of 1 to 10 with 10 being best. They also judged how well the IBM proposal met the requirements. "G" indicates the group's consensus on the importance and "I" indicates how well the group felt the IBM proposal met the requirements. Possible Configuration/Termination Assumptions 1. Any device may perform the configuration. G: 10 I: no 2. The configuring device may be at any position on the bus. G: 10 I: Yes, with some modifications. 3. "New" and "old" devices may be mixed on the bus. G: 9 I: Problem -- old devices do not have a 60-pin cable. 4. The order of new and old devices should not matter. G: 9 I: Problem -- they must be at the end of the cable and that defeats terminator sensing. 5. The solution works for both single-ended and differential drivers. G: 10 I: no -- IBM proposal does not work for differential (Jerry disagrees). 6. Add the minimum number of new signals -- This not only saves connector costs, but also saves pins on protocol chips. G: 8 I: Adds 10 pins. 7. New devices should have one or two identical connectors on each device (same sex and wired one-to-one). No "IN" and "OUT" connectors. G: 7 I: no 8. Time to accomplish configuration should not exceed __ seconds. G: 10 seconds I: yes 9. New devices must remember their assigned address through all but a power outage. They must request a re-configuration when they are powered-on. G: 10 Potential problems -- how to deal with hard reset? I: yes 10. The solution must be able to accommodate more than 8 devices (for SCSI-3). G: 10 (need up to 32 devices) I: Modified could do 16 devices, 32 devices: no 11. Powered down devices do not affect the process, including the device that normally does the configuration. G: 10 I: Problem -- could do with a relay. 11b. More than one device can do configuration. G: 10 I: Yes, requires private agreement. 12a. Must detect no terminators, only one terminator, or more than two terminators. G: 4.5 I: Can detect one terminator, assumes initiator is at end of cable. [Jim Schuessler proposed an analog terminator detector which would detect the number of terminators present.] 12b. Terminators are at the end of the cable. G: 4.5 I: No 13. Cabling between devices must be identical. G: 10 I: Yes, with all new devices. 14. In mixed systems of old and new devices, re-configuration must not change any addresses that were previously assigned. Old devices could be very confused by a device that changes addresses. G: 10 I: Yes 15. All terminator assemblies must be identical. G: 10 I: Yes 16. The solution must be compatible with existing protocol chips. G: 10 I: Yes 17. Addressing must be capable of being stable over multiple configurations. (same or different configuring device? Paul Nitza) 17a. via software [to be considered later] 17b. via hardware [to be considered later] 18. Fifty pins (related to item 6) G: 10 I: No 19. Bi-directional cable G: 9 I: No 20. Must comply with ANSI patent policy. G: 10 I: Yes 21. Device can have no SCSI ID to handle more than the maximum number of devices connected to the bus. G: 2 I: Yes 22. Possible to daisy-chain with only one connector. G: 10 I:no Dave McIntyre moved and Gene Milligan seconded that the working group recommend to the plenary that the autoconfiguration proposal be remanded to a SCSI-3 Specific-Subject Working Group (SSWG) for further consideration, and if a proposal is forthcoming in a timely manner that meets the pleasure of the committee, it could be considered for incorporation in SCSI-2. 18 in favor, 4 opposed, 1 abstained. The following activities occurred during SSWG meetings which were held late Monday afternoon, Tuesday morning (8:00 to 9:00), and Wednesday morning (8:00 to 9:00). SSWG - John Lohmeyer outlined an approach using the existing 50-pin bus which may work. The approach would use an Ethernet-like technique to determine collisions on the bus and to back off. It appears to meet most of the twenty-two requirements. Paul Nitza accepted an action item to draft the approach into a proposal. SSWG continued (Tuesday 8:00am) - Jim Schuessler spoke in favor of the 50- pin solution claiming it has the following advantages: 1. Fewer transceivers needed on-chip. 2. Change to protocol chips is simple. 3. It automatically works with differential. One approach to an autoconfiguration master does following to find devices on the bus: assert RST, assert BSY, assert one ID, assert SEL, release BSY, wait for BSY asserted from the device. Problems: What to do if a device comes on the bus at a later time? Two old devices set to same ID. (No approach handles this.) Old initiators that do not respond to an unexpected reselection. Another possible approach instead of establishing an autoconfiguration master: Pseudo-random SCSI ID picking. First step is to establish a set of unique SCSI IDs. The second step is to reconfigure with the desired SCSI IDs. Possible techniques to accomplish auto-configuration: 1. config request in/config pass (IBM approach) 2. psuedo-random signal "wiggling" 3. analog voltage sensing 4. grab ID's with collision detection (possibly using 2. or 3.) Reconvened at 8:00am Wednesday Snively at the board: A) CONFIGURATION PHASE 1) Arbitration phase without an ID -if old device wins configuration phase is not entered. -if there are no IDs asserted the configuration phase is entered. 2) Set up a unique phase to cause new devices to enter the configuration process. Possibly by setting BUSY & C/D, I/O OR MSG. 3) Each new initiator picks a negated ID. 4) Bus goes BUS FREE phase. B) SCSI ID RESOLUTION PROCESS 1) All new initiators arbitrate with their chosen ID. The winner performs a test for common IDs by selection with his chosen ID to determine if any other device on the bus responds to that ID. If there is a response, the initiator invalidates his present ID and returns to the configuration phase. 2) If there was no response, there are no old devices with same ID and the test for multiple IDs is performed. -Pseudo-random testing is one possibility -Analog voltage sensing -Unique command sequence If passed, then the initiator has that ID and responds as an old device. Any new initiator that has not passed the resolution phase then re-enters the configuration phase. C) SCSI ID RE-ASSIGNMENT The SSWG will meet again Monday evening August 15 at the Colorado Springs plenary meeting. Agenda Items for SCSI-3 (There was insufficient time to consider the following agenda items.) 50. SEND DIAGNOSTIC Pages [87-186] (Spence) 51. LOAD SKIP MASK command proposal [87-203] [87-217] (Floryance, McIntyre) 52. More that 8 devices on wide SCSI [87-206] (Harms) 53. Search Command modifications [88-002] (Stai) 54. Expanded RelAdr Bit [88-007] (Boulay)