Date: October 31, 1990 To: X3T9.2 Membership From: Lawrence J. Lamers, X3T9.2 Secretary John B. Lohmeyer, X3T9.2 Chairman Subject: October 30-31, 1990 X3T9.2 Working Group Meeting John Lohmeyer called the meeting to order at 9:00 a.m., Tuesday, October 30, 1990. He thanked Bill Spence of Texas Instruments 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. Caching Proposal (90-021R3) [Milligan] 2. Diagnostic Command Set (90-103R1) [Pickford] {Wednesday a.m. tentative--Doug may not be able to attend.} 3. Data Compression Mode Page Proposal (90-119) [Krupa] 4. Packetized SCSI (90-132) [Stephens] 5. Parallel SCSI (90-133R3) [Stephens] 6. Extensions for dual port SCSI (90-136R1) [Houlder] 7. Long BUSY conditions [Milligan] 8. Proposal for MDP message control in mode page [Milligan] 9. S/E Cable Test Report No. 3 (90-157) [Spence] 10. REQ/ACK signal quality and Fast Single-ended (90-159) [Chan] 11. SCSI-3 Fiber Channel Project Proposal (90-146) [Allan] 12. SCSI-3 Projects Roadmap 13. Concurrent execution of more than one I/O Process [Pepper] 14. CAM 15. ATA 16. ACTION ITEMS 17. General Working Group Schedule for 1990 The following people attended the meeting: Name Status Organization ------------------------------ ------ ------------------------------ Mr. Robert C. Herron A 3M Company Mr. Robert Otis A Apple Computer Mr. Ed Young P Archive Corp. Mr. David Hess P Berk-Tek, Inc. Mr. Wayne Sanderson P Control Data Corp. Mr. Chuck Micalizzi A Emulex Corp. Mr. I. Dal Allan P ENDL Mr. Robert Liu P Fujitsu America, Inc. Mr. Kurt Chan P Hewlett Packard Co. Mr. Mike Peper A Hewlett Packard Co. Mr. Howard Wang O Hitachi Mr. George Penokie P IBM Corp. Mr. Paul Anderson A IBM Corp. Mr. David A. Buesing O IBM Corp. Mr. Gary R. Stephens S IBM Corp. Mr. Ronald E. Rowland V IBM Corp. Mr. Lawrence J. Lamers P Maxtor Corp. Mr. Kamran Khadavi V Micropolis Corp. Mr. John Lohmeyer P NCR Corp. Mr. James McGrath O Quantum Corp. Mr. Gene Milligan A Seagate Technology Mr. Robert L. Simpson P Sony Corp. of America Mr. Roger Cummings A Storage Technology Corp. Mr. D. W. Spence P Texas Instruments Mr. Edward R. Schurig O Texas Instruments Mr. Arlan P. Stone A UNISYS Mr. Erik Jessen O Western Digital 27 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 ------------- -------- --------------- --------------------------------------- X3T9.2/90-21 10/29/90 G. Milligan Additional SCSI Caching Control Rev 3 X3T9.2/90-166 10/30/90 G. Milligan Initiator Support of Modify Data Pointers Message X3T9.2/90-167 10/30/90 B. Spence S/E Reliability Enhancements X3T9.2/90-169 10/30/90 G. Penokie SCSI-2 Rules for Multiple Active I/O Processes X3T9.2/90-173 10/30/90 M. Peper Multiple I/O Processes in SCSI ------------------------------------------------------------------------------ RESULTS OF MEETING ------------------------------------------------------------------------------ 1. Caching Proposal (90-021R3) [Milligan] Gene Milligan presented revision 3 of his proposal. Several suggestions were made to eliminate confusion with the LOG SELECT and LOG SENSE commands. Gene prepared a revision 4 at the meeting to incorporate these suggestions. The Working Group recommends that 90-021R4 be accepted for inclusion in the appropriate SCSI-3 document. 2. Diagnostic Command Set (90-103R1) [Pickford] Gene Milligan questioned the model that was included in revision 1. He expressed a desire to review the model in detail. Chuck Micalizzi summarized the current status, saying that the commands have been reviewed and agreed to, but that the model was new and not yet reviewed by the working group. Gene requested an electronic copy to facilitate his review. Erik Jessen accepted an action item to contact Doug Pickford and relay the request. Mike Peper expressed a desire to put the DCS in a stand-alone document and not be restricted within a standard. The technology moves too fast for a standard to be effective. John Lohmeyer suggested that this might be addressed by issuing a Technical Report. TRs have due process, but are not published by ANSI. CBEMA publishes TRs. He planned to investigate whether a TR is appropriate for this subject. 3. Data Compression Mode Page Proposal (90-119) [Krupa] The data compression people from HP - Bristol, England plan to be at the December plenary meeting. They may have a revised proposal at that time. The issue of data compression has some surrounding controversy. John Lohmeyer reported that two project proposals for doing data compression as parts of X3B5 projects have received "no" votes at the X3 level. The issue is whether data compression algorithms should be defined in a magnetic medium committee. John pointed out that SCSI-2 contains a field in a mode page to enable/disable compression algorithms. Bill Spence reported that the QIC committee agreed to accept the HP data compression proposal for tape drives. There was some discussion of how a removable medium device could handle the different compression schemes. 4. Packetized SCSI (90-132) [Stephens] Gary presented his Small Computer Fiber Interface (SCFI) document. He has circulated this document to several individuals and plans to revise it for the January committee mailing. His intention is to limit the number of options and to make the minimum changes from SCSI-2 to support the new functions. This document may be a candidate for the working draft of the proposed SCSI-3 Fiber Channel Protocol project. There was some concern that Gary had re-specified more of the Fiber Channel protocol than necessary. Gary had limited maximum packet sizes to 2048 bytes, but Fiber Channel has no such restriction. Fiber Channel provides a service to break large packets into smaller frames. Gary countered that the implementation is simpler if there is only one packet size. A lengthy discussion lead to the conclusion that most X3T9.2 participants could benefit from an update on the Fiber Channel project. Roger Cummings agreed to give a presentation on Fiber Channel at the December plenary meeting (tentatively scheduled for late morning on Tuesday 12/4). 5. Parallel SCSI (90-133R3) [Stephens] Gary included information in his SCFI document describing how to use packetized SCSI over the parallel copper interface using the existing ARBITRATION, SELECTION, and BUS FREE phases plus two new phases: NEXUS IN and NEXUS OUT. There was some discussion over which document should include the NEXUS IN and NEXUS OUT documentation. This issue was not resolved. 6. Extensions for dual port SCSI (90-136R1) [Houlder] Gene Milligan discussed Gerry Houlder's ideas on extending his 90-136R1 proposal to allow more than two ports. He had suggested changing it so that the RST signal would reset either only the port that received it or all ports, the BUS DEVICE RESET message would reset the target for all ports, and the CLEAR QUEUE message would clear queued commands for all ports. He would eliminate the XXX OTHER PORT messages. It was pointed out that these suggestions would effectively return the proposal to its revision 0 state and remove the changes made by the Westminster working group. An alternative method of supporting more than two ports is to change the XXX OTHER PORT messages into XXX ALL OTHER PORTS messages. Several people expressed doubt that more than two ports would be required for parallel SCSI saying that connector space would be a serious impediment to more than two ports. Dal Allan argued that systems requiring more than dual porting probably also require the full multi-porting capability that Gary Stephens has proposed. The last working group agreed to bundle multi- porting with the packetized SCSI protocol. An open item from the last working group meeting was to develop a proposal to add a field to the INQUIRY data to indicate whether a reset condition affects one port or all ports. Another open item at the last working group meeting was whether port identification was needed in the INQUIRY data. Arlan Stone stated that UNISYS requires port identification to verify system cabling. John Lohmeyer proposed that a two-bit field be defined with a reserved bit to its left (in more than two ports are supported later). The two bits would be encoded as follows: 00h Single-port device 01h Dual-port device, port A 10h Dual-port device, port B 11h Reserved It was requested that Gerry Houlder include the above fields in the next revision of his proposal. Gary Stephens raised a related question of whether messages (e.g., BUS DEVICE RESET) get through a reservation. John stated that common industry practice is to accept such messages through reservations. Furthermore, SCSI-2 requires the target to at least retrieve any messages in an initial MESSAGE OUT phase following a SELECTION phase with ATN asserted. 7. Long BUSY conditions [Milligan] There was no new documentation, but this issue stemmed from Gene's 90-125 proposal. The ARC bit proposed in 90-125 essentially results from the inability of some operating systems to properly deal with long BUSY conditions. Many operating systems have timers on I/O operations. The BUSY status is handled at a low-level in the driver which re-issues the I/O process without causing the timer to be reset. If the device stays BUSY for more than the timer value, the system attempts to recover with drastic actions (e.g., BUS DEVICE RESET). SCSI does not place any requirements on the maximum time a device can remain BUSY. Larry Lamers pointed out that most systems can deal successfully with CHECK CONDITION statuses so he recommends that conditions that could result in a long BUSY condition be reported as CHECK CONDITION status with NOT READY sense key. However, ECA requires the BUSY status during the ECA condition. The target has no control over how long an ECA condition lasts. It was concluded that ECA should permit CHECK CONDITION status to avoid the long BUSY condition. Gene Milligan accepted an action item to develop a proposal for adding a BUSY/CHECK CONDITION control bit for ECA to the Control Mode Page. 8. Proposal for MDP message control in mode page [Milligan] Gene Milligan presented his proposal (90-166) for initiator control of the usage of the MODIFY DATA POINTERS message. John Lohmeyer objected to the use of the ILLEGAL REQUEST sense key when the initiator has enabled MODIFY DATA POINTER messages and the initiator subsequently responds to a MODIFY DATA POINTER message with MESSAGE REJECT. There was sentiment for going to unexpected BUS FREE and the group was sympathetic with Erik Jessen's tongue-in-cheek suggestion to respond with a 24 volt spike, but settled on ABORTED COMMAND sense key with an ASCQ of INVALID MESSAGE ERROR. George Penokie asked that the EMDP bit be moved to bit 7 so that the DTDC field could be expanded in the future, if needed. Several people felt that the polarity of the EMDP bit is reversed. They preferred a Disable MDP bit. Gene stated he had selected the polarity to enable when the bit is one because it would cause the fewest problems with existing initiators. There was a lot of debate on this issue with several people torn between architectural purity and minimizing the compatibility issues. A straw pole on this issue was a virtual tie. Since the working group could not reach consensus, the issue will be advanced to the plenary group for resolution. 9. S/E Cable Test Report No. 3 (90-157) [Spence] This topic was covered at the plenary and at the cable working group. 10. REQ/ACK signal quality and Fast Single-ended (90-159) [Chan] This topic was covered at the plenary meeting. Kurt is recommending the use of totem-pole drivers on REQ and ACK and increasing receiver hysteresis from 0.2 to 0.4 volts. Kurt will continue his testing on the P-cable implementation to define its limits. Bill Spence accepted an action item to compile a list of improvements for single-ended implementations. 11. SCSI-3 Fiber Channel Project Proposal (90-146) [Allan] This item was briefly discussed under the following agenda item. 12. SCSI-3 Projects Roadmap There was considerable discussion about how to divide the SCSI-3 project. This ultimately resulted in Larry Lamers presenting a colorful foil that defines three layers: * The top layer includes the SCSI command sets. * The middle layer defines the protocol to use the bottom layer. * The bottom layer defines the physical interfaces. The working group agreed with Larry's proposed structure. (Later in the week, John, Dal, and Larry drafted a complete set of project proposals for the mailing.) The current road map is as follows: SCSI-3 Projects Road Map +---------------+ | SCSI-3 | | Command | | Set | | (90-172) | +---------------+ | | +--------+ +--------+ | | +----------------+ +---------------+ | SCSI-3 | | SCSI-3 | | Parallel | | Fiber Channel | | Protocol | | Protocol | | (90-171) | | (90-146) | +----------------+ +---------------+ | | +----------------+ +---------------+ | SCSI-3 | | Fiber | | Parallel | | Channel | | Interface | | Interface | | (90-105) | | (X3T9.3) | +----------------+ +---------------+ The document numbers shown on the bottom line of each box are the project proposals. The Fiber Channel Interface is an X3T9.3 project. 13. Concurrent execution of more than one I/O Process [Pepper] Mike Peper brought a paper (90-173) describing the opinion of another HP employee that the SCSI-2 standard does not permit more than one concurrently active I/O process for the same initiator on the same LUN. This opinion is based on several occurrences of using "the" instead of "an" as the article for "I/O process". This paper also includes references that imply multiple concurrently active I/O processes are permitted. George Penokie also had a paper which cited many of the same references and reach the opposite conclusion: that multiple concurrently active I/O processes are permitted (90-169). It was perfectly clear that SCSI-2 is not. A straw poll indicated that everyone preferred permitting multiple concurrently active I/O processes. So the real issue became how to document this concept. Mike Peper, Jim McGrath, and George Penokie were assigned to draft a document on how multiple I/O processes should work. In particular, the Peper document identifies several areas of error recovery that are not described in SCSI-2 and should be for complete documentation. This document will likely become the basis for a SCSI Technical Information Bulletin. 14. CAM There is a need for additional working group meetings on SCSI CAM. The issue of when (in the working group week) these meetings should be held was not resolved. The current CAM document is in need of revision as a result of changes made on Monday. Dal could not make the changes in time for the November mailing and plans to have it ready for distribution in the January mailing. 15. ATA Those interested in the ATA project discussed the following subjects on which there will be a working group organized at the December plenary: 1. Jim McGrath: Assign a second use to Pin 26 (Synchronized Spindle) for automatic address recognition as Drive 0 (Master) or Drive 1 (Slave). Jim will prepare a proposal. 2. George Penokie: Nowhere in the ATA document is it stated that if a second command is issued while the first is active, the first "dies" and is replaced by the second command. George will prepare a proposal. 3. Gene Milligan: He was concerned that there may be a timing problem on slow processors because a well-respected BIOS fails on writing data. Larry Lamers suggested that the BIOS manufacturer be consulted, as there is known to be a problem in a specific revision of their BIOS. Gene agreed to investigate further. 16. ACTION ITEMS 1. Erik Jessen will contact Doug Pickford and request an electronic copy of the Diagnostic Command Set (90-103R1). 2. Gene Milligan will develop a proposal for adding a BUSY/CHECK CONDITION control bit for ECA to the Control Mode Page. 3. Gerry Houlder will include a reset identification bit and a port ID in the next revision of his 90-136 proposal. 4. Bill Spence will compile a list of the signal quality improvements for single-ended implementations. 17. General Working Group Schedule for 1990 The next working group meeting is being hosted by Ken Post of Future Domain at (see document 90-158): Red Lion Hotel 3050 Bristol Court Costa Mesa, CA 92626-3098 TEL: 714-540-7000 FAX: 714-540-9176 THE RESERVATION CUTOFF DATE IS DECEMBER 27, 1990. The schedule of meetings for the week is as follows: Monday 01/14/91 09:00-17:00 Small Form Factor Tuesday 01/15/91 09:00-17:00 X3T9.2 WG : HiPPI Wednesday 01/16/91 09:00-17:00 X3T9.2 WG : HiPPI Thursday 01/17/91 09:00-17:00 Fiber Channel : FC-0 : IPI-2 Friday 01/18/91 09:00-17:00 Fiber Channel : FC-0 : IPI-2 The Working Group Schedule is as follows: Date Location Host Comments ------------------- ---------------- --------------- ---------------- Jan 14-18, 1991 Costa Mesa, CA Future Domain (see 90-158) Mar 18-22, 1991 Cupertino, CA {tentative} May 13-17, 1991 Harrisburg, PA AMP {location & date have changed} Jul 15-19, 1991 Valley Forge, PA UNISYS {tentative} Sep 16-20, 1991 Minneapolis, MN {tentative} Nov 4- 8, 1991 Albuquerque, NM Los Alamos Labs