Date: March 21, 1991 To: X3T9.2 Membership From: Lawrence J. Lamers, X3T9.2 Secretary John B. Lohmeyer, X3T9.2 Chairman Subject: March 19-20, 1991 X3T9.2 Working Group Meeting John Lohmeyer called the meeting to order at 9:00 a.m., Tuesday, March 19, 1991. He thanked Dal Allan of ENDL for hosting and arranging the meeting. 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. SCSI Message cleanup (91-8R1, 91-32) [Stephens, Houlder] 2. SCSI-3 Packetized Protocol (SPP) (90-132R3) [Stephens] 3. Data Compression Mode Page Proposal (90-119R1) [Krupa] 4. ATA issues - cable scheme 5. Diagnostic Command Set (90-103R2) [Pickford] 6. Concurrent execution of more than one I/O Process [Peper, Penokie] (90- 173, 91-34) 7. MODE SELECT defaults (91-019) [Hanmann] 8. Media Changer problems 9. Control of SCSI Device Power (91-014 R1) [Penokie] 10. IEEE/PCMCIA/SFF report [McGrath] 11. New Additional Sense Code 5Dh (91-27) [Penokie] 12. Alternate SE Termination Method [Trung Le] (91-037 & 91-038) 13. SE vs. Diff. Identifier Symbols? [Lohmeyer] 14. Should IGNORE WIDE RESIDUE only be permitted for the last transfer of a command? [Lamers/Lohmeyer] 15. November Working Group Meeting -- Yea or Nay 16. SCSI-3 (SPI & SIP) Draft Documents 17. Message Parity Error (91-035) 18. Editor's issues in SCSI-2 19. Block Descriptors in MODE SENSE The following new documents were distributed at the meeting: Document Doc Date Author Description of Document ------------- -------- --------------- --------------------------------------- X3T9.2/91-14 2/22/91 G. Penokie Control of SCSI Device Power Rev 1 Consumption X3T9.2/91-27 2/28/91 G. Penokie New Additional Sense Code (5Dh) X3T9.2/91-32 3/12/91 G. Houlder Response to ATN during RESELECTION X3T9.2/91-34 3/15/91 G. Penokie Response to X3T9.2/90-173 Document X3T9.2/91-35 3/18/91 B. Galloway Clean up of 5.6.13 MESSAGE PARITY ERROR X3T9.2/91-36 3/18/91 B. Galloway SCSI-2 Medium Changer Devices field descriptions X3T9.2/91-37 8/3/90 T. Le FPT - Forced Perfect Termination X3T9.2/91-38 8/3/90 T. Le FPT - Forced Perfect Termination (Foils) X3T9.2/91-39 P. Boulay Media Changer Device Capabilities Page X3T9.2/91-40 3/15/91 D. Hagerman Request for Interpretation of SCSI-2 on the use of MODE SENSE data The following people attended the meeting: Name Status Organization ------------------------------ ------ ------------------------------ Mr. Al Wilhelm P Adaptec, Inc. Mr. Gary W. Arakaki A Advanced Micro Devices Mr. Sassan Teymouri V Advanced Micro Devices Mr. Charles Brill P AMP, Inc. Mr. Bob Whiteman A AMP, Inc. Mr. Scott Smyers O Apple Computer Mr. Dennis Pak O Apple Computer Mr. Edward Hrvatin O Burndy Corp. Mr. Douglas Hagerman A Digital Equipment Corp. Mr. Paul Hanmann P Emulex Corp. Mr. I. Dal Allan P ENDL Mr. Robert Liu P Fujitsu America, Inc. Mr. Kenneth Post P Future Domain Mr. Tim Greaves V GCA Electronics Mr. Kurt Chan P Hewlett Packard Co. Mr. Steve Krupa V Hewlett Packard Co. Mr. Howard Wang O Hitachi Mr. Paul Boulay O Hitachi Computer Prod Amer Mr. George Penokie P IBM Corp. Mr. Paul Anderson A IBM Corp. Mr. Gary R. Stephens A IBM Corp. Mr. David A. Buesing O IBM Corp. Mr. Trung Le V IBM Corp. Mr. Lawrence J. Lamers P Maxtor Corp. Mr. Gary Murdock A National Semiconductor Mr. John Lohmeyer P NCR Corp. Mr. James McGrath P Quantum Corp. Mr. Gene Milligan A Seagate Technology Mr. Brian Johnson O Seagate Technology Mr. Dennis L. Mattock V Silicon Detector Corp. Mr. Robert L. Simpson P Sony Corp. of America Mr. Dennis Appleyard P Storage Technology Corp. Mr. William Galloway O Summus Computer Systems Mr. Robert N. Snively P Sun Microsystems, Inc. Mr. D. W. Spence P Texas Instruments Mr. David Griffith O Wang Laboratories Mr. Doug Pickford A Western Digital Mr. Erik Jessen O Western Digital 38 People Present Status Key: P - Principal A - Alternate O - Observer S,V - Visitor RESULTS OF MEETING 1. SCSI Message cleanup (91-8R1, 91-32) [Stephens, Houlder] Gene Milligan brought copies of Gerry Houlder's 91-32 document which objects to the proposed change in Gary Stephens' 91-8R1 document to require the target to honor the attention condition prior to transmitting the IDENTIFY message. Gary Stephens pointed out that it is architecturally better to honor an attention condition from RESELECTION phase immediately after the RESELECTION phase. This is consistent with all other phases. Also, as rev 10c is written, the initiator must restore the pointers before the nexus is fully identified. (The second half of 91-8R1, which was accepted in February, corrects this problem.) Gerry's document points out that this timing has been unchanged since 1987 and that changing it now would invalidate all SCSI-2 designs. Furthermore, with the changes in the second half of 91-8R1, nothing is broken. The consensus of the working group was that Gary's architecture is indeed better, but a better architecture is not sufficient justification to change something which is no longer broken. Gerry had identified the ABORT TAG message as being inappropriate following the IDENTIFY message, but prior to the SIMPLE QUEUE TAG message. Bob Snively pointed out that the present ABORT TAG message wording permits an ABORT TAG message at that point and the effect would be to cause a disconnect, but not to abort the I/O process which had not yet been fully identified. The following two paragraphs will be added to the ABORT TAG description: "On a reconnection, the ABORT TAG message aborts the current I/O process if it is fully identified. If the I/O process is not fully identified (i.e., an I_T_L nexus exists, but the target is reconnecting for an I_T_L_Q nexus), then the I/O process is not aborted and the target goes to the BUS FREE phase. IMPLEMENTORS NOTE: A nexus is not fully identified on a reconnection if the ATN signal is asserted during or prior to the IDENTIFY message and the target only has tagged I/O processes for that initiator on that logical unit." 2. SCSI-3 Packetized Protocol (SPP) (90-132R3) [Stephens] Due to limited time, there was no discussion on this topic. 3. Data Compression Mode Page Proposal (90-119R2) [Krupa] Steve Krupa presented the proposal developed at the Data Compression working group meeting on Monday. A few further changes were recommended. Steve plans to prepare a revision 3 of his proposal for the May mailing. 4. ATA issues - cable scheme There was no discussion on this item because Jim McGrath had to leave early. 5. Diagnostic Command Set (90-103R2, 91-22) [Pickford] Doug Pickford stated that he agreed with Jim McGrath's identification of the diagnostic markets, but not with Jim's conclusions. He then led a review his proposal which resulted in several further changes. Larry Lamers ran Doug's proposal through a computer grammar and style checking program and provided Doug with the output. Doug plans to prepare a revision 3 after the April plenary meeting. 6. Concurrent execution of more than one I/O Process [Peper, Penokie] (90- 173, 91-34) Jim McGrath provided a one-sentence explanation of multiple, concurrent I/O processes: "When multiple I/O processes are outstanding at a device for an initiator, then the device may reselect that initiator for any of these I/O processes." George Penokie's comments (see document 91-034 R0) engendered a long discussion, which resulted in a number of clarifications to be included in SCSI-2 Rev 10d. See SCSI-2 Rev 10d and/or 91-34 R1 for the details. 7. MODE SELECT defaults (91-019) [Hanmann] Paul Hanmann noted that changing the mode parameters for an SCSI device in an unknown state back to the manufacturers defaults is not always handled properly. Some targets reject a MODE SELECT command when a value in a non- changeable field has changed as a result of changes made to other (changeable) fields. For example, consider a disk drive that has a changeable bytes per sector field and a non-changeable (by the initiator) sectors per track field. Then if the disk currently has 17 sectors per track using 512-byte sectors, the initiator is required to "reflect" the 17 sectors per track value when switching to 1024-byte sectors. Some targets have rejected such MODE SELECT commands. SCSI-2 has no specific statement about such ripple effects. It was suggested that all checking of non-changeable parameter values be done prior to processing the changeable parameters. Paul Hanmann and Bill Galloway accepted an action item to develop a statement for next plenary meeting. 8. Media Changer problems (91-036R0; 91-039R0 ) [Galloway, Boulay] Bill Galloway found a problem with field names in the Media Changer section. By special invitation, Paul Boulay joined the meeting briefly on Wednesday to discuss these issues. Bill's document was accepted as editorial improvements. Paul also brought 91-39 which describes a further problem. The first issue relates to interpretation of Table 16-26 and what operations are supported by setting the bits. There exist implementations that do not support all variations of the operations which are indicated to be supported. Paul suggested that a bit of one indicates all such operations are supported. Bill Galloway accepted an action item to draft wording. Kurt Chan requested operation code 0Ch be made vendor unique to match the table in Appendix I. The group recommended that Kurt's request be accepted in trade for his agreement to edit SCSI-3. The group also agreed to add an implementors note, to be drafted by Paul Boulay, to 16.3.3.3 that justifies having this page. 9. Control of SCSI Device Power (91-014 R1) [Penokie] George presented his latest proposal on power control. The power control terminology agrees with the AT Attachment document. Bob Snively raised a question about the interface signal conditions while the device is in sleep mode. In particular, Bob want a statement added to require devices to not interfere with bus operation while asleep. Sections 4.4.1.2 and 4.4.1.1 are the likely locations to receive the angelic sleep requirements. 10. IEEE/PCMCIA/SFF report [McGrath] Jim McGrath reported. The IEEE disk attach committee has a charter to look at interface definitions that attach devices directly to a CPU bus. The substance of the work is to define the mechanical, electrical, and software interface. Jurisdiction issues are unclear with related activities in X3T9, SFF, IEEE, and PCMCIA. The disk attach committee is planning to use a memory model as opposed to a channel model. The PCMCIA group also is using a memory model. The next meeting is April 24 in Sunnyvale. Meetings planned once per month, with a meeting in Sunnyvale the day after the X3T9.2 plenary meeting and at 3:00 p.m. on the same day as SFF meetings in same location as the SFF meeting. The PCMCIA committee had met earlier in the month in Hawaii for six days. Jim McGrath's presentation on using the PCMCIA interface for disk drives was not well received. Jim observed that the group was focused on using memory chips. Dal Allan reported on the Small Form Factor meeting. There was significant debate on connector styles. Some connector vendors had realized that pin- and-socket style connector is not being considered. A solution for SFF needs was proposed by those vendors. There are still many unanswered questions. Dal pointed out that the thin PCB (0.8 mm) presents unique issues of strength; connector removal forces need to be very low. A presentation was made by IBM regarding a 1.3-inch form factor device. 11. New Additional Sense Code 5Dh (91-27) [Penokie] George Penokie proposed a new ASC code to warn the system that the device is nearing failure. After some discussion, the name was revised to be FAILURE PREDICTION THRESHOLD EXCEEDED. The working group recommends that this code be included in SCSI-3. 12. Alternate SE Termination Method [Trung Le] (91-037 & 91-038) Trung Le presented the results of studies done at IBM to satisfy internal requirements for a more robust termination scheme. IBM calls this a "Forced Perfect Termination" (FPT). It uses diode clamping to prevent signal reflections at the terminator. The circuit was initially confidential, but no IBM patent exists on the circuit. A pluggable terminator using the IBM circuit is available commercially from Aironics at 512-252-2303. A similar type incorporating active regulation is available from Methode Electronics. The Aironics terminator only terminates REQ, ACK, and SEL signals with the new circuit. The other lines are terminated normally with the 220/330 ohm circuit. On Wednesday afternoon, Bill Spence presented a foil showing some of his concerns about the circuit. He was principally concerned people might assume the IBM termination will relieve them of the need for good system design. FPT will not solve problems in the middle of the cable. Cable impedance still must be about 80 ohms. Bill advocated using the FPT idea in conjunction with the regulated Boulay terminator. Bill's foil reduced to ASCII was as follows: 2.85 v 2.85 v 4.66 v 4.66 v | | | | \ \ \ \ 22.3 / 110 22.3 / 110 22.3 / 191 22.3 / 191 mA \ ohm mA \ ohm mA \ ohm mA \ ohm | | | | +---+---+---+---+---+---+ +---+---+---+---+---+---+ | | o o | | o o | | ----- 0.4 v ----- 0.4 v --- --- | | GND GND 90-123R1 I Z0 = 90 ohms Z0 = 79 ohms Z0 = 68 ohm 8/31/90 ------- ------------ ------------ ----------- 22.3 mA 2.41/2.60 v 2.16/2.45 v 1.91/2.27 v 4.66 v | \ 191 2.4 v ------+ / ohm | \ Z0 I V o | -------- --------- --------- \o--------+ V 100 ohm 20 mA 0.84 v o | 82 ohm 24.4 mA 0 v | --- 0 v ------+ \ / Diode --- | 1.8 v 13. SE vs. Diff. Identifier Symbols? [Lohmeyer] John Lohmeyer said he had a request from some NCR people to color code SCSI connectors to identify single-ended vs. differential. This engendered a wide variety of viewpoints. John managed to move on to another topic by promising to develop a proposal. 14. Should IGNORE WIDE RESIDUE only be permitted for the last transfer of a command? [Lamers/Lohmeyer] The question was should IGNORE WIDE RESIDUE message be permitted once per DATA IN phase or only on the last DATA IN phase of a command. The present wording in 5.6.8 is fairly clear that IGNORE WIDE RESIDUE messages are permitted inbound only following any DATA IN phase. Al Wilhelm preferred only once per I/O process, but could live with once per DATA IN phase. He requested that the document be made explicitly clear. John had excavated the minutes of the October 1987 working group meeting where the group had agreed to permit IGNORE WIDE RESIDUE once per data phase, in or out, but only from the target to the initiator. See 87-185R0 for details. Apparently, the working group later revised this agreement to only permit IGNORE WIDE RESIDUE messages for DATA IN phases. Gary Stephens said that some applications require that the initiator be able to sent an IGNORE WIDE RESIDUE message for DATA OUT phases, especially following a TERMINATE I/O PROCESS message. Gary said this was necessary to support some older hardware and software that can shorten a write command into just the data transferred so far, not the length specified in the CDB. Gary found himself in a minority of one. No one else felt these old applications were worth the complexity of supporting outbound IGNORE WIDE RESIDUE messages. One place in 5.6.8 will be clarified from "the DATA IN phase" to "any DATA IN phase". An implementors note will also be added. 15. November Working Group Meeting -- Yea or Nay Apathy clearly won, but the straw poll indicated that a November meeting is desired. 16. SCSI-3 (SPI & SIP) Draft Documents Kurt Chan brought a limited number of copies of the SCSI-3 Parallel Interface and SCSI-3 Interlocked Protocol draft documents. Kurt said that Gary Stephens had already provided more input than he could handle, so in the interests of time the documents were not discussed. An editors meeting was tentatively scheduled for April 10-12 in Wichita. 17. Message Parity Error (91-035) Bill Galloway had a one-page document describing suggested editorial improvements to the MESSAGE PARITY ERROR message description. Bill did not know how much committee energy had previously been directed on this subject, but after the initial "you can't change what we worked so hard on" comments, Bill's proposal was accepted with a couple minor changes. 18. Editor's issues in SCSI-2 Larry Lamers had several questions about committee intentions regarding SCSI-2 changes. The cryptic conclusions were: Delete p5 of 6.5.5; note that AEN should not be used if a nexus exists. In 6.8 add "ECA and CA excepted". Table 8-56, PER bit - applies only to user data. Delete the last sentence of p1 of 5.2.2. Please see SCSI-2 Rev 10d for the full-English versions of these clarifications. Larry also discussed the SCSI-2 editing schedule. Incorporating the ANSI comments is taking much more time than initially anticipated. In order to get a review copy out sooner, he proposed that Rev 10d only include the committee changes and that 10e include the ANSI changes. 19. Block Descriptors in MODE SENSE (91-040R0) [Doug Hagerman] Doug Hagerman brought a draft Request for Interpretation concerning whether the MODE SENSE block descriptors can return a changeable mask. He had encounter two different implementations. Block descriptors existed prior to the page concept and are not mode pages. Therefore, when a MODE SENSE command is used to return the changeable values, the block descriptors (if enabled) are returned with their "current" values. Doug was satisfied with the above statement and said he would not formally submit his draft RFI. Doug's document will be included in the committee mailing with a note saying it was not actually sent to CBEMA.