Date: May 16, 1991 To: X3T9.2 Membership From: John B. Lohmeyer, X3T9.2 Chairman Subject: May 14-15, 1991 X3T9.2 Working Group Meeting John Lohmeyer called the meeting to order at 9:00 am, Tuesday, May 14, 1991. He thanked Chuck Brill of AMP for hosting the meeting in Harrisburg, PA. John noted the last time he visited Harrisburg was June 1981 -- his first standards meeting. He hoped he would not have to wait another decade to return. As is customary, the people attending introduced themselves. A copy of the of the X3T9.2 membership list was circulated for attendance and corrections. Copies of the draft agenda and the recent document register were made available to those attending. Information on X3T9.2 and Mailing Subscription Forms were made available. The final agenda was as follows: 1. IEEE/PCMCIA/SFF report [McGrath] 2. Flat File System (91-056) [McGrath] 3. Statement on Queuing [McGrath] 4. Control of SCSI Device Power (91-014R3) [Penokie] 5. DEC issues on SCSI-2 implementation (91-053) [Hagerman] 6. Issues on CAM XPT/SIM document (91-046 R1) [Gallant] {Tuesday p.m.} 7. Diagnostic Command Set (90-103R3) [Pickford] {Wednesday a.m.} 8. ATA issues - cable scheme (91-052R1) [McGrath] 9. ATA - 8-bit Compatibility [Hanan] 10. ATA - Set Features command [all] 11. SCSI Device Identifier (91-49R1, -63) [Lohmeyer] 12. Serial SCSI {Wednesday p.m.} 13. SCSI-3 Packetized Protocol (SPP) (91-013R0) [Stephens] {Wednesday p.m.} 14. Recommended change to SCSI-2 (91-60) [Luttig] 15. Outbound disconnect and Disconnect Time Limit (91-62) [Wicklund] 16. Floptical Command Set for SCSI-3 [Wilhelm] 17. Questions about Contingent Allegiance [Lohmeyer] 18. Fast Single Ended Specification (91-64) [Steele] 19. Removable Media AT Attachment Devices [Manley] The following people attended the meeting: Name Status Organization ------------------------------ ------ ------------------------------ Mr. Al Wilhelm P Adaptec, Inc. Mr. Charles Brill P AMP, Inc. Mr. Ken Schreder V AMP, Inc. Mr. Dennis Pak O Apple Computer Mr. Ed Young P Archive Corp. Mr. Joe Chen A Cirrus Logic Inc. Mr. Stephen R. Cornaby P Conner Peripherals Mr. Bob Reago O Dataram Dr. William Ham A Digital Equipment Corp. Mr. John A. Gallant A Digital Equipment Corp. Mr. Douglas Hagerman A Digital Equipment Corp. Mr. Peter K. Townsend V DuPont Connector Systems Mr. I. Dal Allan P ENDL Mr. Dave O'Shea A Future Domain Mr. Kurt Chan P Hewlett Packard Co. Mr. Mike Peper A Hewlett Packard Co. Mr. Jeffrey L. Williams A Hewlett Packard Co. Mr. George Penokie P IBM Corp. Mr. Gary R. Stephens A IBM Corp. Mr. Gerald Marazas A IBM Corp. Mr. Giles Frazier V IBM Corp. Mr. William D'Andrea V J.S.T. Corporation Mr. Joel W. Heberlig V JST Corporation Mr. Chuck Grant A Madison Cable Corp. Mr. Bill Kutsche V Murata Erie N.A. Mr. John Lohmeyer P NCR Corp. Mr. David Steele S NCR Corp. Mr. James McGrath P Quantum Corp. Mr. D. Michael Robinson P Seagate Technology Mr. Gene Milligan A Seagate Technology Mr. Ram Battu V Seagate Technology, Inc. Mr. Robert L. Simpson P Sony Corp. of America Mr. William Galloway P Summus Computer Systems Mr. Dexter Anderson A Sun Microsystems, Inc. Mr. Patrick Manley V SyQuest Technology Mr. Tom Hanan A Western Digital Mr. Doug Pickford A Western Digital Mr. Erik Jessen O Western Digital 38 People Present Status Key: P Principal A Alternate O Observer S Special Interest (frequent visitor) V Visitor The following new documents were distributed at the meeting: Document Doc Date Author Description of Document ------------- -------- --------------- --------------------------------------- X3T9.2/90-103 5/1/91 D. Pickford Proposed Extensions for SCSI-3, Rev 3 Standard Physical Layer Access X3T9.2/91-14 5/3/91 G. Penokie Control of SCSI Device Power Rev 3 Consumption X3T9.2/91-46 4/16/91 J. Gallant Proposed Changes and Additions for Rev 1 Common Access Method X3T9.2/91-49 4/25/91 J. Lohmeyer SCSI Bus Type Identification Rev 1 X3T9.2/91-52 5/14/91 J. McGrath Cable Select Feature Rev 1 X3T9.2/91-56 4/22/91 J. McGrath New SCSI Commands for FFS (Flat File System) X3T9.2/91-60 4/24/91 E. Luttig Recommended change to SCSI-2 X3T9.2/91-62 5/3/91 T. Wicklund Outbound disconnect and Disconnect Time Limit X3T9.2/91-63 5/8/91 B. Medlinski SCSI Identifier Symbols X3T9.2/91-64 5/9/91 D. Steele Proposal for Fast Single Ended Specification X3T9.2/91-65 4/22/91 J. McGrath Tagged Queuing Restrictions X3T9.2/91-66 5/8/91 J. McGrath Using a PCMCIA Interface on 1.8" Winchester Disk Drives X3T9.2/91-67 5/2/91 A. Wilhelm Request for new SCSI Peripheral Device Type "Floptical" X3T9.2/91-68 5/13/91 A. Wilhelm Proposal for addition of two new Medium Types to Table 8-45 X3T9.2/91-69 5/13/91 B. Snively Conformance of SPP and Fiber Channel Results of meeting 1. IEEE/PCMCIA/SFF report [McGrath] Jim McGrath reported that PCMCIA had last met on Monday and Tuesday, May 7- 8, 1991. Their Board of Directors has agreed to include magnetic disks in the list of PCMCIA supported devices (Please refer to 91-66 for more information). The next PCMCIA meeting is in Seattle June 24-25, 1991. Jim is proposing that X3T9.2 define electrical interfaces for ATA and SCSI using the PCMCIA physical interface. PCMCIA has two models which may be appropriate for attaching magnetic disks. One model is the traditional I/O channel model and the second is a memory device model. It may be possible to map magnetic disks to the memory device model. The Small Form Factor meeting on Monday spent most of their meeting working on connector issues. Jim McGrath volunteered to conduct a PCMCIA tutorial Tuesday afternoon June 18 following the X3T9.2 plenary meeting. 2. Flat File System (91-056) [McGrath] Jim introduced his document proposing a flat file system for SCSI-3. This would permit the device to assume several functions including data compression, file caching, and data migration (to a archival media). Jim's proposal leaves the directory structure, file access protection, and file naming as host functions. Jim plans to flesh out his proposal in revision 1. 3. Statement on Queuing [McGrath] John Lohmeyer conditionally accepted an action item to work with Larry Lamers to add Jim's one sentence to an appropriate place in 6.8.2. The condition, which was accepted, was to also allow him freedom to clarify other parts of this section. "A target with one or more I/O processes outstanding from a given initiator may reconnect to that initiator to continue any of the I/O processes in any order." The last three words were added in lieu of Jim's second paragraph. The discussion turned to the issue of tagged RESERVE commands and some potential pitfalls of poor implementations of queue algorithms. 4. Control of SCSI Device Power (91-014R3) [Penokie] George presented rev 3 of his document on SCSI device power control. This revision resulted from the comments received at the St. Petersburg Beach meeting. George received some more input and plans to do a rev 4 for the Minneapolis plenary meeting. 5. DEC issues on SCSI-2 implementation (91-053) [Hagerman] Doug Hagerman presented several questions that DEC people have about SCSI-2. Please refer to 91-53 for the questions: 1. Yes. 2. Should the BUS DEVICE RESET message clear both target and initiator I/O Processes? The standard appears to say yes, but some people think host adapters should only clear its target-mode I/O Processes. The conclusion of the working group was that it is appropriate for a BUS DEVICE RESET message to clear all I/O Processes regardless of the device's role. 3. Whatever happens first is handled first. 4. All commands are rejected (a clarification will be made on page 84). 5. Already accepted for SCSI-3. 6. Data Bytes per Physical Sector does not include ECC bytes. 7. Yes, if it is explicitly allowed or ambiguous. Where the document is clear, do what the document says. 8. Correct. The "active pointers" will be renamed the "current pointers" to make this consistent. Doug's questions prompted an extensive discussion of what is meant by suspending the queue during an ECA condition. Some of this discussion overlapped with agenda item 17. Agreement was reached on several points: 1. Either QUEUE FULL status (preferred) or MESSAGE REJECT message to the queue tag message should be returned if a tagged command is received during a contingent allegiance condition. There was some question as to whether this should be done in SCSI-2 or wait for SCSI-3. A vote on this subject will be placed on the June agenda. 2. Several places in the sixth paragraph of 6.8.2 of SCSI-2 should be changed to "from that initiator". The ordering of I/O Processes is only with respect to one initiator. If it is necessary to control the order of I/O Processes from different initiators then the initiators must use the RESERVE and RELEASE commands. (This problem appears to be a change that was not fully propagated when the committee dealt with the guaranteed data integrity issue.) 3. If the target is capable of managing separate sense data areas, then it may choose to only suspend the portion of the queue that is from the initiator that caused the contingent allegiance condition. 4. If an extended contingent allegiance condition exists all I/O Processes from all initiators for the logical unit are suspended. This is similar to an implicit device reservation. 6. Issues on CAM XPT/SIM document (91-046R1) [Gallant] {Tuesday p.m.} John Gallant reviewed his 91-46R1 document, which contained three proposals: 1. This proposal was dropped by John. 2. Accepted. 3. Accepted. John also gave Dal several minor editorial changes for Rev 2.4 of CAM XPT/SIM. Dave O'Shea also discussed several changes and clarifications most of which were accepted. Amongst Dave's comments was one to eliminate all references to OS/2 since all the document does is reference Microsoft's LADDR. With OS/2 references gone, there is no longer any need to separate the tables into OSD and Common fields, because all fields are now common. Dal will revise the XPT/SIM document and Rev 2.4 will be included in the July mailing. 7. Diagnostic Command Set (90-103R3) [Pickford] {Wednesday a.m.} Doug Pickford brought a new revision of his document which incorporated the comments on the model from Gene Milligan and Dal Allan. Once again, Jim McGrath argued that adopting a standard diagnostic command set inhibits the disk vendor's ability to innovate and evolve. John read the pertinent section from the March '90 working group minutes (90-50). Many of the same issues from that time remain. Doug has addressed the architectural issues, but the following issues (from the March '90 minutes) remain: "Some of their [disk drive manufacturers] concerns include fears about the security of their proprietary microcode, exposing the low-level functions of their drives to potential corruption from end-users, and concerns about the standard restricting future enhancements." John and Dal proposed that DCS be made a separate project from SCSI-3. This would permit the project to proceed independently of SCSI-3. It would also allow SCSI-3 disk vendors to choose to comply or not comply with DCS on its merits and not as a required consequence of complying with SCSI-3. Doug Pickford and Dal Allan will prepare a project proposal. 8. ATA issues - cable scheme (91-052R1) [McGrath] Jim's proposal was critiqued and the people present agreed to it with the change of adding a requirement for a 10k ohm pullup resistor on pin 28 (cable select/synchronize spindle). Dal will include 91-52R1 plus the pullup resistor change in ATA Rev 2.4. 9. ATA - 8-bit Compatibility [Hanan] Tom Hanan had waited on preparing a description of the additions to the Features Register to support 8-bit operation until a complete list of other uses by vendors had been created. Tom will prepare material for the Set Features command. 10. ATA - Set Features command [all] Gene Milligan agreed to provide a description of all the Seagate uses, and Jim McGrath will prepare one for a Write Cache bit. Tom Hanan will prepare his on the assumption that the valid codes are 01h to begin 8-bit data transfers and 81h as the default of 16-bit. 11. SCSI Device Identifier (91-49R1, -63) [Lohmeyer] John had revised his proposal from input received at St. Petersburg Beach. He also brought copies of a fax from Bill Medlinski (91-63) suggesting graphics also be included. Dal advocated using a circle for single-ended and a square for differential. The idea of using the same icon as used on Apple machines was considered for single ended with some other icon for differential. The idea was rejected by the working group because this icon has largely been used on nonstandard connectors and should not be associated with the SCSI-3 standard. The topic was forwarded to the plenary with no other recommendation. 12. Serial SCSI {Wednesday p.m.} This item was a duplicate of item 13 and/or other working group meetings scheduled for Monday afternoon and Wednesday night. 13. SCSI-3 Packetized Protocol (SPP) (91-013R0) [Stephens] {Wednesday p.m.} Gary Stephens lead a review of the SCSI-3 Packetized Protocol document (91- 13). He did not get very far before a philosophical discussion ensued. Bob Snively had sent a document (91-69) with Dexter Anderson criticizing Gary's approach for SPP. He felt that it is better to use more of the Fiber Channel protocol and not just encapsulate the SCSI protocol within Fiber Channel sequences. While encapsulation has an advantage of permitting SCSI to operate over any interface that can deliver packets, it also means that SPP must include more I/O process management. Dal, surprised to find himself in agreement with Bob, also objected to Gary's approach as the only solution. He said it would preclude hardware steering and would duplicate functions provided by FC-2 resulting in higher cost. There was considerable discussion on the pros and cons of the two techniques. Gary contended that encapsulation is necessary to insure that sequences are processed in order. Dal disagreed. John said he could see advantages with both approaches, but was confused by Dal's view of how SCSI should be mapped onto FIber Channel. He asked Dal to explain his approach. Dal gave a one-foil description of the relationship of Fiber Channel operations, exchanges, and sequences. He described how a complex fabric could deliver Class 2 frames out of order pointing out that the probability is quite low. The remainder of the meeting was spent with Dal describing how he would map SCSI onto Fiber Channel protocol. 14. Recommended change to SCSI-2 (91-60) [Luttig] The working group considered 91-60 in the absence of Ernest Luttig. The participants pointed out that SCSI-1 was completely silent on the topic of how long after a reset the target must be able to perform TEST UNIT READY, INQUIRY, or REQUEST SENSE commands. SCSI-2 added the recommendation of a maximum of 250 milliseconds Reset to Selection Time. The "shall" that is referenced in paragraph 6.1.3 of SCSI-1 relates to a different requirement. While it might be desirable to strengthen the SCSI-2 recommendation into a requirement for SCSI-3, it is not in conflict with SCSI-1. The working group recommends that this proposed change not be accepted for SCSI-2. 15. Outbound disconnect and Disconnect Time Limit (91-62) [Wicklund] Tom Wicklund's document points out that two timing parameters may both apply in the case of a target delay for a reconnection. Since the documentation is widely separate (5.6.6 and 7.3.3.2), it may be confusing. Also the interaction of the two parameters is not clear. The working group recommends that Tom's second suggestion be accepted and added to section 5.6.6 in SCSI-2 rev 10e: "... After releasing the BSY signal, the target shall not participate in another ARBITRATION phase for at least a disconnection delay or the time specified in the disconnect time limit mode parameter (see 7.3.3.2), whichever is greater." 16. Floptical Command Set for SCSI-3 [Wilhelm] Al Wilhelm brought two related proposals (91-67 and 91-68) proposing SCSI-3 additions to support Floptical (tm) devices. His proposal would create a new device type for Floptical to avoid a problem caused by boot software treating a Floptical device as a hard disk. This would result in file structure corruption or confusion. This is actually a more general problem with and removable device that uses a media interchange format including flexible disks. Jim McGrath suggested an alternative way to prevent existing boot software from working with such devices: Add a bit to the READ CAPACITY command that enables the reporting of removable disk capacity. If this bit is not one, then the Floptical (or other removable device with special media format requirements) would report zero capacity. This would prevent existing device driver software from expecting a hard-disk file structure. Al agreed to develop a proposal with Jim McGrath's assistance. After lunch, Al had second thoughts. Apparently, the READ CAPACITY approach would not work in all cases. He said his proposal still stands and he wanted it placed on the June agenda. [Chairman's note: After the meeting, Al and Jim reported that they had found yet another approach to solve the problem. They planned to document it for the June meeting.] Al's second proposal (91-68) was accepted for SCSI-3. 17. Questions about Contingent Allegiance [Lohmeyer] John Lohmeyer had received an email from Gerry Fredin of NCR questioning what should happen if a contingent allegiance condition is created by a target and the initiator next sends a tagged command. He could not find anything in the standard specifically addressing this case. Gary Stephens found a reference in 6.8.2: "Tagged queuing may also be temporarily disabled internal to the SCSI device during certain initialization periods or to control internal resource utilization. During these periods the target may elect to return QUEUE FULL status or it may respond to any queue tag message with a MESSAGE REJECT message." While this does not specifically address Gerry's question, it could be construed that Gerry's case falls under "controlling internal resource utilization". The last paragraph of section 6.7, ECA, contains the requirement, "If the initiator sends a queue tag message the target shall respond with QUEUE FULL status." The working group recommended that the target return QUEUE FULL status in the contingent allegiance case as well. (Also see item 1 in the second list under agenda item 1.) 18. Fast Single Ended Specification (91-64) [Steele] David Steele brought a proposal suggesting that single-ended fast data transfers be defined. SCSI-2 timings were defined assuming that differential transceivers would be employed for fast data transfer option. Several people objected to the 3 meter limitation, wanting to permit longer cables. David pointed out that operation over 3 meters may work, but is not guaranteed. Standing waves may occur and cause failures. Jim McGrath expressed concern that this proposal could result in a specification that invalidates existing silicon. John countered that the proposal is a SCSI-3 proposal and does not apply to SCSI-2 silicon. He contended that the committee should examine what is necessary to achieve reliable operation on single-ended systems without worrying about product consequences. It was pointed out that implementing all of the recommendations may not be necessary to achieve acceptable operation. The committee needs to find a way communicate this fact. The working group reviewed David's proposal point-by-point and he agreed to prepare a revision 1. The topic was deferred to the plenary meeting in order to give people time to study the proposal, especially the recommended timing parameters. 19. Removable Media AT Attachment Devices [Manley] Patrick Manley suggested several changes to ATA to support removable media devices. The Features Register, Error Register and Identify Drive command are affected. Patrick is to prepare a complete proposal for mailing to the members two weeks before the plenary so that the members can vote on whether removability should be added. Dal Allan wants to have everything included in the next revision to be mailed following the June plenary meeting, with a motion to forward in August. Action Items 1. Larry Lamers will rename the "active" pointers to the "current" pointers in SCSI-2. 2. John Lohmeyer will send mailing labels to Patrick Manley so he can mail out his ATA proposal to the X3T9.2 members prior to the June meeting. 3. John Lohmeyer will communicate several SCSI-2 Rev 10d clarifications and corrections to Larry Lamers. 4. Larry Lamers will include the disconnection delay clarification (see agenda item 15) in SCSI-2 rev 10e. 5. Dal Allan will revise the CAM XPT/SIM document in time for the July mailing. 6. Dal Allan will revise the AT Attachment document in time for the July mailing. 7. Doug Pickford and Dal Allan will prepare a project proposal on Diagnostic Command Set. 8. Dal Allan will include 91-52R1 plus the pullup resistor change in ATA. 9. Tom Hanan will prepare material for the ATA Set Features command. 10. Jim McGrath will prepare a proposal on a Write Cache bit for the ATA Set Features command.