Date: May 10, 1989 To: X3T9.2 Membership From: Larry Lamers, X3T9.2 Recording Secretary John Lohmeyer, X3T9.2 Chairman Subject: May 1989 SCSI Working Group Minutes John Lohmeyer of NCR E&M Wichita hosted a working group meeting May 8 - 9, 1989 in Wichita, KS at the Hilton Hotel. The working group thanks John and NCR for hosting this short meeting. The final agenda was: SCSI-2 Items 1. Resolution of the X3T9 letter ballot comments on SCSI-2 R8 (89-055R1,-67) 2. Should we define a transition threshold voltage? (89-048) [Snively] 3. Change Vterm to 4.5 - 5.25 volts? (89-048) [Snively] 4. Resolution of the SILI/SULI issue (89-21,44,50) 5. Better definition of queuing error handling [Eneboe, Nitza] 6. MESSAGE PARITY ERROR handling (89-054R0) [Appleyard, Spence] 7. QIC-600 density code replacement. [Davies] 8. Redefine contingent allegiance (semantics) [Lohmeyer] 9. When can a target change signals following selection? [Lohmeyer] 10. Format Unit and Saving Pages [Lamers] 11. SEND DIAGNOSTIC UntOfl & DevOfl bits [Lamers] 12. Verify Operations & Reallocation [Cornaby] 13. Log Parameter Parsing [Cornaby] 14. REQUEST SENSE response to invalid logical unit (89-062R0) [Spence] 15. Physical seek function within SEND DIAGNOSTIC (89-063R0) [Cornaby] 16. Tom Wicklund's comments on SCSI-2 Rev 8 (89-053) [Wicklund] SCSI-3 Items 50. Additional SEND DIAGNOSTIC Pages (87-186R0) [Spence] 51. LOAD SKIP MASK Command Proposal (87-203 & 87-217) [Floryance, McIntyre] 52. More Than 8 Devices On Wide SCSI (87-207) [Harms] 53. Search Command Modifications (88-002) [Stai] 54. Expanded RelAdr Bit Definition (88-007) [Boulay] 55. Autoconfiguration SSWG (88-69R1, 88-092R0, & 88-100R0) [Marazas, Nitza, McGrath, Boulay] 56. Error Handling Action Codes (88-127) [Penokie] 57. Autosense [Nitza] 58. Request For Data DAT Device Type [Andrews] 59. Documentation Layering [Stephens] 60. Single-Cable 16-bit Wide SCSI [Penokie, Lohmeyer, Lamers) 61. Alternate Physical Layers (e.g., fiber optics) 62. LOGICAL UNIT RESET Message Proposal (89-058) [Lohmeyer] The following people attended the meeting: Name Stat Organization ------------------------------ ---- ------------------------------ Mr. Harlan Andrews P Apple Computer Mr. Joe Lawlor P AT&T Mr. Dan Davies P Cipher Data Products, Inc. Mr. Bharat Shah A Cipher Data Products, Inc. Mr. Paul Hanmann P Emulex Corp. Mr. Denis Brockus A Emulex Corp. Mr. I. Dal Allan P ENDL Mr. Robert Liu P Fujitsu America, Inc. Mr. Kenneth Post O Future Domain Mr. Kurt Chan P Hewlett Packard Co. Mr. Mike Peper A Hewlett Packard Co. Mr. Kiyotake Kumazawa O Hitachi America, Ltd. Mr. George Penokie P IBM Corp. Mr. Gerald Marazas A IBM Corp. Mr. David A. Buesing O IBM Corp. Mr. Paul Boulay A LMS-OSD Mr. Lawrence J. Lamers P Maxtor Corp. Mr. Stephen Cornaby P Micropolis Corp. Mr. John Lohmeyer P NCR Corp. Mr. Charles Yang P Panasonic Industrial Co. Mr. Mike Eneboe A Seagate Technology Mr. Fred Burgess V Seagate Technology Mr. Robert N. Snively P Sun Microsystems, Inc. Mr. D. W. Spence P Texas Instruments Mr. Edward R. Schurig O Texas Instruments Mr. Rob Oliver V Texas Instruments Mr. Jeff Stai P Western Digital Mr. Tom Hanan A Western Digital Mr. Doug Pickford A Western Digital 29 attendees Status Key: P - Principal A - Alternate O - Observer V - Visitor New documents received at the meeting: Document Doc Date Author Description of Document ------------- -------- --------------- --------------------------------------- X3T9.2/89-61 5/5/89 G. Penokie SCSI Bus Fairness Technique X3T9.2/89-62 5/7/89 B. Spence REQUEST SENSE Response to Selection of an Invalid LU X3T9.2/89-63 5/3/89 S. Cornaby Physical Seek Function within SEND DIAGNOSTIC X3T9.2/89-64 5/4/89 M. Eneboe Proposed changes to SCSI-2 Rev 8 regarding queuing X3T9.2/89-65 5/5/89 D. Buesing SCSI Bus Fairness through Selective Reconnection **************************** General Discussion **************************** ************************* Agenda Items for SCSI-2 ************************** 1. Resolution of the X3T9 letter ballot comments on SCSI-2 R8 (89-055) [Milligan] (89-067) [Roccati] The comments from Gene Milligan are denoted with a GEM-xx, the comments from Arnold Roccati are denoted with an AJR-xx. Only the comments not accepted for incorporation in SCSI-2 Rev 9 are listed below with the reason why. Some of these may be addressed at a later date. AJR-01 - Pertains to document style and will be addressed in a subsequent edit. AJR-02 - Pertains to document style and will be addressed in a subsequent edit. AJR-03 - Pertains to document style and will be addressed in a subsequent edit. AJR-05 - The connector experts assured the committee that the figures are correct and it is normal industry practice to dimension connectors this way. The socket dimensions are for the entire opening in the plastic and do not show the actual contact dimensions. A note will be added to the figures to make this clearer. (Due to pressing deadlines and a lack of appropriate software tools to modify the connector figures, the editors elected to delay incorporating the note on revision 9.) AJR-06 - Taken under advisement, but this matter of terminology and also a matter of having the software to change the drawings. AJR-07 - Taken under advisement, but this matter of terminology and also a matter of having the software to change the drawings. AJR-08 - Taken under advisement, but this matter of terminology and also a matter of having the software to change the drawings. AJR-14 - Pertains to document style and will be addressed in a subsequent edit. GEM-25 1 ampere vs. 2 amperes for A cable current limit. We split the difference and recommended 1.5 amperes. GEM-38 Gene lost this one at Florida, but should we add an implementors note recommending waiting for SEL assertion? Yes, note added. GEM-51 Is MODE SELECT a special case? No, handled by MS parameter definitions. GEM-80 Back to the future? CHANGE DEFINITION allowed when emulating SCSI-1. GEM-86 Contingent allegiance definition: time when you return busy vs. time when you preserve sense data? Second option picked. GEM-87 The contingent allegiance definition was revised. GEM-94, 95 This paragraph was fixed to reflect the new, improved handling of queuing errors. GEM-117 Alternative wording accepted. GEM-146 Not accepted GEM-149 Accepted, but still needs editorial work. GEM-153 Only the shadow knows... (Will change this to "POSITION ERROR RELATED TO TIMING"). GEM-159 "only mode 000b and 010b" GEM-168 Section 7.3.2.1, page 7-70, Table 7-55 disk revolution, 7-57 'on-the-fly' and 'by other means', new words accepted. GEM-174 Not accepted. GEM-175 Paul Boulay is to clarify this wording. The intent is that all the data transferred by a command is to be sent without interruptions. (This applies to both 01b and 10b.) Also, in 11b change "I/O process" to "command". The difference that remains is permitting disconnect between the time all data is transferred and the time of ending status. GEM-176 This table was modified to be more flexible. GEM-178 Not accepted. GEM-182 System software should deal with this. GEM-189 The unit off-line and device off-line bits could affect a reservation. GEM-194 Models can contain requirements. This particular case was changed to an example, however. GEM-195 The definition of MEDIUM ERROR is clear. A name change would cause too many phone calls to the chairman. GEM-205 The immediate bit adequately describes this. GEM-212 There was a previous hard-fought battle that selected "should". GEM-215 This line was deleted from each CDB. GEM-217 Not accepted. GEM-224 Because the requested list was not found, so another was returned instead -- this unusual treatment is a legacy from CCS. GEM-225 Larry believes the reference is correct. GEM-229 Not accepted -- would not work for ST-506. GEM-230 Not accepted. GEM-231 Not accepted -- s/b power on cycle as this may happen without going through power off - Snively GEM-232 Will use "field" instead. GEM-234 To avoid confusion over which reservation is being released. GEM-241 Not accepted. GEM-242 Not accepted. GEM-243 Not accepted. GEM-244 Not accepted -- the editors use "see" for finding references. GEM-246 Accepted to replace 2nd and 3rd sentences. GEM-247 Incorporated George Penokie's solution. GEM-252 Following accepted: 0h (normal priority) This value should be used whenever no distinction should be made between demand read data and other data. [Delete Imp. note and sort the table by numeric value or alphabetically.] GEM-253 Not accepted. GEM-257 Because this really is a physical sector. GEM-258 s/b "Flexible Disk Page" both places. GEM-259 "---but may be used for other devices." was accepted. GEM-261 Not accepted. GEM-262 Editorial -- resolution left to the editors. GEM-263 1st point accepted, 2nd rejected GEM-266 Not accepted. GEM-273 Larry will fix the words in this paragraph; deallocation concept will not be added. GEM-275 A reference to the verify mode page will be added. GEM-277 Not accepted GEM-278 Not accepted GEM-280-285 Not accepted GEM-290 Larry created alternative wording. GEM-294 s/b detects a change of synchronization -- Larry to investigate. GEM-301 Not accepted GEM-303 "Buffered Mode" is not the same as having a buffer. GEM-304 Accepted Bill Spence's alternative wording. GEM-305 SCSI devices do generally (always) have a buffer, but SCSI devices are not required to support "buffered mode". GEM-306 Not accepted GEM-310 Similar changes as with direct-access devices. GEM-311-315 Not accepted GEM-317 89-050r1 will be included in S2R9 GEM-319 Not accepted GEM-326 Synchronize with note on page 8-38. GEM-327 Not accepted GEM-329 Because they are needed. GEM-331 No, not accepted GEM-332 Not accepted GEM-334 Not accepted GEM-335 Alternative fix accepted. ============================================================================== Bit| 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 | ============================================================================== | WP | Buffered Mode | Speed | ============================================================================== GEM-336 Not accepted GEM-339 4 and 5 are not the same and it should be 1986. GEM-340 Note 6 s/b deleted on code 13h. GEM-342 Not accepted GEM-348 Not accepted There was insufficient time for the working group to resolve the remaining comments from Gene Milligan during the working group meeting. The editors were assigned to address these comments. Larry and Paul reviewed the CD-ROM section comments: GEM-375 Product specific data deleted. GEM-379 Not Accepted GEM-382 Paragraph will be reworded to clarify that the CHECK resulted because the resume request could not be satisfied because of the state of the device. GEM-384 Noted notes to be reworked into requirements as appropriate. GEM-385, 387 The intent is to permit approximate (within known limits) reports so that the command cam complete within a short time. The offending section will be worked on. GEM-389 The note provides a rationale for the arrangement of byte 2 of the CDB. It anticipates a possible future need for standardizing this data. It is correct. GEM-390 An additional requirement will be explicitly stated that values 13h and 14h will be returned only to the initiator of the audio operation. GEM-396 To make things cleared the continue bit will be renamed to Stop on Track Crossing. GEM-397 Not accepted. This is a reminder due to concerns for safety. GEM-398 Meaning of 0h modified to Vendor Specific. GEM-399 All references to EER deleted. GEM-402 Definition of "h" deleted as redundant and unnecessary. However the glossary stays put. Paul separately reviewed the Media Changer comments: GEM-425 The definition of address will be reworded as a requirement. GEM-427 Not accepted. That slash will be globally purged. GEM-428 Not accepted. The model is a good place to state and also justify requirements of this sort. Pointers to this section will be added to 16.2.5 and 16.2.6. GEM-430 z+1 is correct, z appears at the end of table 16-11. changing that to z-1 and this reference to z might make things clearer. GEM-431 The text will be changed to make clear that the bytes are to be left out. This was done because the volume tag is so long, because you may have two of them per piece of media and because you may have over 100 elements to report on. GEM-438 This attempts to define the default responses to be used if a shortened version of the parameter block as specified by the length field. This may be over-specifying it. The field is 2 bytes solely for compatibility with other 4 byte headers. GEM-440 Not accepted. Consistency with other device type sections. 2. Should we define a transition threshold voltage? (89-048) [Snively] Many of the silicon designers did not feel that such a tight specification (1.4+_0.1 volts) is cost effective. A tighter specification on hysteresis was preferred. TTL switches around 1.4-1.5 volts; CMOS switches around 1.8-2.0 volts and thus needs a special circuit to use the TTL switching voltages. The TTL levels are desirable for new and future silicon. The threshold should be in the 1.4 volt range with more hysteresis. Bob Snively accepted an action item to draft an implementors note. 3. Change Vterm to 4.5 - 5.25 volts? (89-048) [Snively] Rather than change the terminator power specification, the new style terminator was indicated as preferred. The old terminator (220/330 ohm) was retained for SCSI-1 compatibility. Paul Boulay accepted an action item provide a generic figure of his circuit. 4. Resolution of the SILI/SULI issue (89-021, 89-044, 89-050R1) [Davies] 89-050R1 was accepted to resolve this. Subsequently, Bill Homan's spokesman, Paul Boulay, stated that a separate bit in the mode parameters is more desirable for controlling the SILI/SULI nature of the SILI bit. The representatives from IBM and HP were reluctant to expose themselves to fire once more. Shyness won. 5. Better definition of queuing error handling [Eneboe, Nitza] Bob Snively presented a proposal based on a letter (89-064) he received from Mike Eneboe on how to handle errors related to duplicate nexus and command queuing. If the continent allegiance condition does not exist at the time a new nexus is attempted and if a new nexus is attempted when there is an existing nexus for the same initiator-logical unit the action taken by the target is as follows: new current I/O Number of I/O processes I/O process in the queue when the new process queue I/O process arrives Action I_T_x_Q --> I_T_x 1 Abort all I_T_x --> I_T_x 1 Abort all I_T_x_Q --> I_T_x_Q n Abort all I_T_x --> I_T_x_Q n Abort all (except if CA or ECA) The approach outlined in the information above was accepted for incorporation into the section on redundant I/O processes. 6. 89-54 MESSAGE PARITY ERROR handling [Appleyard, Spence] An archaeological dig for the resolution of 87-213 was conducted and it turned up that 87-213 was accepted in March of 1988. There was a requirement in section 5.1.9 that the message be contained wholly within a message phase. The sentence from 89-054 as amended was added to resend the 'entire' message in section 5.6.13. John's on-line wording was accepted for revision 9. 7. QIC-600 density code replacement. (part of 89-050R1) [Davies] The table entries needed were included in 89-050R1 and were accepted. 8. Consistent Definition of Contingent Allegiance [Lohmeyer] Carried over. 9. When can a target change signals following selection? [Lohmeyer] There is some ambiguity on when the target may change signals following assertion of the BSY signal in response to be selected. John Lohmeyer suggested adding a requirement that "a target not change the REQ signal to enter an information transfer phase until SEL is false." This suggestion was accepted for incorporation into section 5.1.3. 10. Format Unit and Saving Pages [Lamers] The issue of when pages are saved is covered in one place in the document. However a reference is needed in section 7 to advise that, for direct-access devices, not all pages are saved until a FORMAT UNIT command is successfully performed. Larry was assigned to add the reference. 11. SEND DIAGNOSTIC UntOfl & DevOfl bits [Lamers] Last sentence of section 7.2.15 paragraph 6 needed some repairs. The solution adopted was to delete it and add in paragraph 4 of page 81 of SCSI-1. The definition of the real meaning of UntOfl and DevOfl bits was not clear. John Lohmeyer stated that the bits informed the target that the logical unit or device was off-line in the opinion of the initiator, that is, it would not issue other commands that could conflict with the requested diagnostic operation. Others stated that these bits are more active in nature and take a logical unit or target device off-line for the purpose of performing diagnostic operations. If this is the case, then what is the targets response to selections from other initiators? Should BUSY status be returned to all initiators other than the one requesting the diagnostic operation? Should a unit attention condition be required to notify the other initiators following a SEND DIAGNOSTIC command with either the UntOfl or DevOfl bits set? The consensus was that the implementors note in section 7.2.15 advising the use of reservations was sufficient. However, reservation information could be destroyed when the DevOfl bit is set. No resolution was reached on this. The standard will not require a specific implementation. 12. Verify Operations and Reallocation [Cornaby] The discussion was whether or not the ARRE and AWRE bits of the Read- Write Error Recovery Page affected verify operations. A separate page exists for control of verify operations. The consensus was that the ARRE bit is not used for verify operations; the AWRE bit applies to the WRITE AND VERIFY command; the VERIFY command shall not perform reallocations; reallocation depends on the mode parameters for the copy manager when performing the COPY and COPY AND VERIFY commands. 13. Log Parameter Parsing [Cornaby] How is a error log constructed in the real world? The document presents a structure but gives no real guidance on how to implement it. Steve Cornaby asked that people interested in this topic to discuss it with him. The working group requested that Steve develop a written proposal. 14. REQUEST SENSE response to invalid logical unit (89-062R0) [Spence] Accepted for incorporation in SCSI-2 Revision 9. 15. Physical seek function within SEND DIAGNOSTIC (89-063R0) [Cornaby] This proposal was added to the list of items for SCSI-3. 16. Tom Wicklund's comments on SCSI-2 Rev 8 (89-053) [Wicklund] This item was not addressed during the meeting due to lack of time and forgetfulness of the chairman. However, the editors reviewed the comments on Wednesday and accepted all but first, fourth, ninth, and tenth comments as editorial improvements. The first comment was not accepted because target routines do have value (although minimal). Future enhancements in SCSI-3 are possible. The fourth comment questioned the meaning of "left-most" regarding bits in fields. The editors felt that the general industry usage is clear that left-most refers to the highest numbered bit of the lowest numbered byte. The ninth comment was concerned with the READ LONG command when there is not a one-to-one relationship between physical and logical blocks. This issue has previously been addressed and the group has agreed to permit vendor specific handling of these issues. The tenth comment appeared to be confusion over the meaning of the valid bit. This bit indicates that the information field contains valid data, not that the entire sense data is valid. ************************* Agenda Items for SCSI-3 ************************** The only action taken on SCSI-3 items was to add Item 63. 50. Additional SEND DIAGNOSTIC Pages (87-186R0) [Spence] 51. LOAD SKIP MASK Command Proposal (87-203 & 87-217) [Floryance, McIntyre] 52. More Than 8 Devices On Wide SCSI (87-207) [Harms] 53. Search Command Modifications (88-002) [Stai] 54. Expanded RelAdr Bit Definition (88-007) [Boulay] 55. Autoconfiguration SSWG (88-69R1, 88-092R0, & 88-100R0) [Marazas, Nitza, McGrath, Boulay] 56. Error Handling Action Codes (88-127) [Penokie] 57. Autosense [Nitza] 58. Request For Data DAT Device Type [Andrews] 59. Documentation Layering [Stephens] 60. Single-Cable 16-bit Wide SCSI [Penokie, Lohmeyer, Lamers) 61. Alternate Physical Layers (e.g., fiber optics) 62. LOGICAL UNIT RESET Message Proposal (89-058) [Lohmeyer] 63. Physical seek function within SEND DIAGNOSTIC (89-063R0) [Cornaby]