X3T9.2/87-80 Date: May 21, 1987 To: X3T9.2 Membership From: John Lohmeyer, Chair X3T9.2 Phone: (316) 688-8703 Subject: SCSI-2 Ad-Hoc Minutes - 05/18/87 to 05/20/87 - Minneapolis, MN An SCSI-2 working group meeting was held May 18-20 in Minneapolis, MN. I would like to thank Gerry Houlder of CDC/MPI for hosting this meeting. The excellent facilities and the PC were much appreciated. The following people attended: Meeting Attendees Name Status Organization ------------------------------ ------ ------------------------------ Mr. Anthony Kozlowski P 3M Company Mr. Robert N. Snively P Adaptec, Inc. Mr. Robert R. Williamsen A Adaptec, Inc. Mr. James J. Semenak P AT&T Information Systems Mr. Gene Milligan P CDC/MPI Mr. Gerald Houlder A CDC/MPI Mr. Dan Davies P Cipher Data Products, Inc. Mr. Paul Nitza P Emulex Corp. Mr. Randy Weber ST Emulex Corp. Mr. I. Dal Allan P ENDL Mr. Robert Liu P Fujitsu America, Inc. Mr. David Snook A Fujitsu America, Inc. Mr. Seiichi Sugaya S Fujitsu Limited Ms. Elen Hunt A Hewlett Packard Co. Mr. Michael A. Brown SO Hewlett Packard Co. Mr. Greg Floryance P IBM Corp. Mr. Steve Kappes P Interphase Corp. Mr. Skip Jones P Maxtor Corp. Mr. John Lohmeyer P NCR Corp. Mr. James McGrath P Quantum Corp. Mr. D. W. Spence A Texas Instruments Mr. Ralph Schultz P UNISYS Mr. Jeff Stai A Western Digital Status Key: P - Principal A - Alternate L - Liaison O - Observer S - Special Interest Group V - Visitor (no address info) The following documents were distributed at the meeting: Document Doc Date Author Description of Document ------------- -------- --------------- ---------------------------------------- X3T9.2/87-7 5/18/87 D. W. Spence Revised proposal for downloading Rev 1 microcode. X3T9.2/87-43 5/18/87 G. Floryance Revised Common READ LOG command Rev 1 X3T9.2/87-74 5/15/87 S. Jones Revised Media Changer Proposal from Working Group X3T9.2/87-75 5/13/87 R. Weber Miscellaneous Sequential-Access comments X3T9.2/87-76 5/16/87 P. Nitza Auto Sense and MODE SELECT/SENSE Target Option Proposal X3T9.2/87-77 5/16/87 P. Nitza INQUIRY Peripheral Device Type Proposal X3T9.2/87-78 5/8/87 R. Schultz Miscellaneous corrections to SCSI-2 Rev1 X3T9.2/87-79 3/25/87 E. Hunt Proposed Rewording of Section 9.1.8 (Page 9-15 in Rev 0) X3T9.2/87-82 5/7/87 R. Schultz Merged Tape and Disk Additional Sense Codes X3T9.2/87-83 5/18/87 A. Kozlowski Certification Pattern Page 6 of MODE SELECT/SENSE X3T9.2/87-84 5/18/87 A. Kozlowski Peripheral Device Type for Direct-Access Tape Devices All items from the agenda were discussed (not necessarily in this order) with the following results: 1. SCSI-2 Rev 1 review of: MODE SELECT and MODE SENSE page descriptions and FORMAT UNIT command. A page-by-page review of FORMAT UNIT and MODE SELECT/SENSE was conducted with the follow results: Gene Milligan questioned whether the intent of the disable bits (DPRT and DCRT) in FORMAT UNIT are to prevent time-interval interruptions -- No reassignment permitted. Does anyone have an answer? Jeff Stai accepted action item #14. Mike Brown accepted action item #15. Paul Nitza accepted action items #16 and #18. Greg Floryance accepted action item #17. This agenda item was not completed -- more work is needed at the next working group meeting. 2. MODE SELECT actions that cause UNIT ATTENTION condition (Gary Stephens). Gary was not present -- item deferred. 3. Optical disk 12-byte command set (John Hoy). [87-72] John was not present -- item deferred. 4. Version/level switching via INQUIRY DDF and RDF. [87-42] An action item was given to John Lohmeyer, Jeff Stai, Jim Semenak, and Greg Floryance to summarize the issues and the possible resolutions to this issue. (See action item #1.) {They met and Jeff Stai plans to prepare a proposal.} 5. INQUIRY device type 7Fh issue (Paul Nitza). [87-77] Accepted as amended (the qualifier field was expanded to 3 bits from 2 bits). Paul Nitza will revise this proposal. (See action item #2.) 6. Asynchronous event notification (Stephens/Snively/Boulay). [87-52, 87-61] Bob Snively's proposal is still waiting for criticism and comment. It will be acted upon at the next plenary meeting. 7. Command Queuing. [87-59] The following issues were identified by the Working Group: 1) ABORT Message Definition 2) Queue Full Status/Message? 3) Single Queue TAG Message 4) Linked Command Execution 5) Need for TAG Message to follow IDENTIFY Message 6) Data Integrity (See Action item #2) 7) Sense Data can be lost 8) Error Handling 9) Ambiguous TAG Message for REQUEST SENSE 10) Priority Override for LIFO/FIFO 11) Ordering of Commands: a) With respect to information b) With respect to execution sequence This item was remanded to the Command Queuing Task Group for further consideration (Contact: Bob Snively 408-432-8600 x442). 8. READ LOG command (Greg Floryance). [87-43] After much discussion, the working group reached the following conclusions: 1) We should accept Greg Floryance's READ LOG proposal [87-43 Rev 1] as a common command (Section 7). 2) Some features still need to be addressed: a) Save parameters in a non-volatile area b) Error logging counters c) MODE SELECT control pages d) Additional counters to differentiate between Busy bus and transfer rate data overrun/underrun counters. Greg Floryance accepted action item #8. 9. Document Restructuring (John Hoy). Samples of "pure block move" restructuring of sections 7, 8, and 15 were available at the working group. These initial drafts did not attempt to integrate the description of commands that are available in multiple CDB sizes. These plenary should review these samples. It was suggested that if we choose to restructure the document, we should do it in two steps: 1) Just use block moves with no wording changes, 2) Reword appropriate commands to consolidate the descriptions. 10. Clarification of WRITE AND VERIFY modifier bits (Greg Floryance). Ralph Schultz addresses this issue in 87-78. The central theme here is what constitutes "verification"? Does it include(exclude) retries or ECC? The working group recommends that a new MODE SELECT page be defined to control these parameters (separate from Page 01h, which controls these parameters for normal read commands). If a target does not support the new page, then the verification is vendor unique. Another issue discussed was whether the verify requires a second, separate data transfer over the SCSI bus. (See action items #5 and #6.) 11. Reservation problem in COPY command (Dan Davies). [87-70] It turns out that this issue really shouldn't occur if multi-initiator systems properly use the RESERVE and RELEASE commands. No changes to the document are needed, although implementors of multi-initiator systems perhaps should be advised to use these commands. See action item #3. 12. Power-on protocol proposal (Bob Snively). [87-53 Rev 1] This item was not reviewed. 13. Page concept for SEND DIAGNOSTIC, RECEIVE DIAGNOSTIC RESULTS, SEND, and RECEIVE commands (Daniel Loski). Daniel was not present -- item deferred. 14. Deadlock concerns when queuing with RESERVE and RELEASE commands (Jeff Stai). This item was not reviewed. 15. INQUIRY Vital Product Data proposal (Greg Floryance). [87-42] Greg will revise this proposal to remove the "soft switch" material. Another issue concerning whether the target (controller) should be separately addressable from the logical units was considered by the working group. There was considerable disagreement over whether this is desirable. The plenary should consider this issue further. 16. Broadcast Proposal (Paul Boulay). [87-61] Paul was not present -- item deferred. 17. Additional Sense Code review in Rev 1. Need codes for Saved Values not supported, Saved Values not are not available, etc. [87-82] The need for several new codes was discussed. Paul Nitza accepted action item #7. 18. Clocking on edges vs. levels in synchronous SCSI (Jeff Stai). This item was briefly discussed. Jeff feels that some wording changes may still be needed. If so, he will bring in a specific proposal. The following items were added to the agenda at the meeting: 19. Downloading microcode. (Bill Spence) [87-7 Rev 1] Revision 1 adds the concept of saving the firmware to the local disk. This function is added to the WRITE BUFFER command using the reserved mode field 11b. Some concerns were voiced over specifying the contents of the data. While it is a good idea to include error detection mechanisms, starting addresses, execution addresses, etc., the working group felt that this information should remain implementation-specific. The desirability of including an optional mechanism for uploading microcode was also discussed. In order to provide this mechanism, the Mode field in WRITE BUFFER and READ BUFFER will be expanded to 3 bits so that an upload mode can be added to the READ BUFFER command. Autoload from removable media was also discussed. No resolution was reached on exactly how autoload should be defined. We need a more detailed proposal on this subject. (See action item #8.) 20. Automatic Medium Changers proposal from the Changer Working Group. [87- 74] Skip briefly discussed 87-74. It was requested that Skip revise this document to include the Automatic Changer Device model. Further review should be conducted at the next plenary meeting. 21. Sequential-Access Issues (Randy Weber). [87-75] REQUEST SENSE command issue: The answer to this issue is in the READ LOG command. Randy will propose a revision to cover the issue. READ command issue: The new paragraph on Page 9-5 beginning with "If a filemark," has several errors. A small group met off-line to discuss the corrections and they will be included in the next revision. Additional Sense Codes: Some corrections are needed in Table 9-36. The Additional Sense Code for End-of-Data detected 00.03 should be deleted. The question was raised over what is meant by "Tape Void Detected" (2E.00) and by "Void at BOT" (2F.00). Dan Davies accepted action item #9. 22. Auto Sense and MODE SELECT/SENSE Target Option Proposal (Paul Nitza). [87-76] The working group favors doing auto sense by using a AUTOSENSE DATA FOLLOWS message followed by a change to DATA IN phase, which would transfer the sense data. Initiators that do not support this feature would reject this message. Also included in 87-76 was a new MODE SELECT/SENSE page. It was rejected-- reporting the ability to do autosense and command queuing should go into the INQUIRY data. The cache disable bit should be included in a separate page to control caching features. Paul Nitza accepted action item #10. 23. Miscellaneous Corrections to SCSI-2 Rev 1 (Ralph Schultz). [87-78] There was some discussion over where and how to document error reporting mechanisms. Jim Semenak and Gerry Houlder accepted action items #12 and #13. There was a considerable discussion on whether to document Additional Sense Codes as qualifiers to Sense Keys or as orthogonal to Sense Keys. The results of a straw vote on this topic were to recommend that the Sense Keys and Additional Sense Codes be orthogonal. (This is not required of the vendor unique codes.) A number of editorial suggestions were accepted. 24. Re-wording of Section 9.1.8 (Elen Hunt). [87-79] These comments were editorial -- Jim Semenak will include them in the next revision of section 9. 25. Certification Pattern Page 6 of MODE SELECT/SENSE (A. Kozlowski). [87-83] There was insufficient time to cover this item in much detail -- it will be discussed further at the next working group meeting. 26. Peripheral Device Type for Direct-Access Tape Devices (A. Kozlowski). [87-84] This proposal would create a new device type for direct-access devices that such as those that are implemented on magnetic tape. They are direct-access devices except that "seeks" can take a long time, especially for blocks that are "far away". While the device-type code would be different, all the commands would be the same as for the direct-access device command set. There was insufficient time to cover this item in much detail -- it will be discussed further at the next working group meeting. 27. Should TERMPWR usage be mandatory in SCSI-2? Some initiators cannot easily provide this power. What about mixed SCSI-1 and SCSI-2 systems? (Steve Kappes) Steve Kappes stated that VME host adapters have difficulty providing terminator power because of VME power restrictions. Interphase would like terminator power to be optional, as in SCSI-1. Steve plans to prepare a further proposal on this subject. Action Items: 1. John Lohmeyer, Jeff Stai, Jim Semenak, and Greg Floryance will summarize the issues and the possible resolutions to the Version/Revision switching issue (see item 4). 2. Paul Nitza will revise 87-77 (see item 5). 3. Working Group will address integrity issues as related to generic implementations with: a) Multiple initiators b) LUN-specific queuing mechanisms. 4. Paul Nitza will address the multiple-initiator environment including the data integrity issue. 5. Paul Nitza will create a MODE SELECT page to control the error recovery used during verification (see item 10). 6. Jeff Stai will clarify the WRITE AND VERIFY and VERIFY commands (see item 10). 7. Paul Nitza will provide a new table for Direct-Access device Additional Sense Codes including Byte 13 (see item 17). 8. Bill Spence will revise 87-7 Rev 1 to include the working group comments (see item 19). 9. Dan Davies will investigate what is meant by additional sense codes 2E.00 and 2F.00. He will provide a clarification of these terms (see item 21). 10. Paul Nitza will revise 87-76 has discussed (see item 22). 11. Greg Floryance will revise his proposal [87-43] to include the 87-81 comments and the other suggestions identified by the working group (see item 8). 12. Jim Semenak will associate Additional Sense Codes to the appropriate Sense Keys in the Sequential-Access command set. 13. Jim Semenak and Gerry Houlder will attempt to document generic error reporting methods for the Sequential-Access and Direct-Access command sets, respectively. 14. Jeff Stai will prepare revised wording for the FORMAT UNIT command to clarify a number of issues identified in the meeting. 15. Mike Brown will revise the REASSSIGN BLOCKS command wording to clarify multiple block reassignments and failure reporting. 16. Paul Nitza will revise the MODE SELECT and MODE SENSE command wording to include the changes identified by the working group. 17. Greg Floryance will revise the AWRE and ARRE descriptions in MODE SELECT and MODE SENSE. 18. Paul Nitza will revise Table 8-19. cc: Attendees who are not P, A, or L members.