** This is the quasi-official and semi-temporary T13 email list server. **
Dan,
>>On write cache off/on setting I would say it should apply to normal
>>write commands only. There is the F (flush) bit in the write stream
>>command to disable the cache on a command by command basis.
Another point you brought up, WRITE CACHE ENABLE/DISABLE. If write
cache has been disabled by Set Features, and someone issues a Stream Write
command with the Flush bit off, does that imply that Write Cache should be
enabled for that Stream command, even though it is disabled by Set Feature?
gkl
-----Original Message-----
From: Daniel J Colegrove [mailto:[EMAIL PROTECTED]]
Sent: Monday, August 06, 2001 5:52 PM
To: [EMAIL PROTECTED]
Subject: [temp t13] Thoughts on Streaming and Write Caching
** This is the quasi-official and semi-temporary T13 email list server. **
I think that the deferred write error condition is an indicator to look in
the streaming write error log for a new error from a previously
completed write stream command. I don't think that the results of
normal write commands get put in the streaming log, so they don't
cause deferred write errors.
The problem with that is that it means that a drive has to keep
track of normal write commands and streaming write commands
in the write cache, locking up the drive if a normal cached write
fails and doing a deferred write error if a streaming write fails. As Mark
Evans pointed out at the meeting that would put a big new burden
on drive designers.
Mixing IT commands and streaming commands makes no sense to me.
Either the host is managing the drive latency by assigning command
time limits or it is not. Having one IT command in the mix wrecks the
whole streaming system. To get IT data in a streaming environment
you have to use the streaming commands with RC=0. You get good
data or a time out error. If there is a time out error you retry with
HSE=1 to go further down the ERP. We have no mode switch to put
the drive in a streaming only mode so there will still be a mixed
environment
if, for example, an AV box boots up using IT commands and then switches
to streaming commands.
Maybe we could solve the problem with an explicit deferred error mode
on/off set features that changes the drive behavior. With the drive in
deferred mode it reports deferred normal write errors. With the drive
in normal mode streaming write errors lock up the drive. If we want
cache uniformity we have to set the same policy for both kinds of
write commands.
On write cache off/on setting I would say it should apply to normal
write commands only. There is the F (flush) bit in the write stream
command
to disable the cache on a command by command basis.
Best Regards,
DC
Dan Colegrove
Senior Technical Staff Member
Storage Technology Division
IBM Corporation
702-614-6119
702-614-7955 fax
Tie Line 280-9959 (IBM internal)
[EMAIL PROTECTED]
Please respond to [EMAIL PROTECTED]
Sent by: [EMAIL PROTECTED]
To: "'[EMAIL PROTECTED]'" <[EMAIL PROTECTED]>
cc:
Subject: [temp t13] Re: Minutes of the July 24-26 working group
** This is the quasi-official and semi-temporary T13 email list server.
**
on Item #15. It should also be noted that part of the issue here is that
currently, when an I/T Write command is issued and write cache is enabled,
that if an error occurs writing that data the drive will report an aborted
status on all commands after the failure occurs (of course this state is
cleared by hard reset or POR). The problem arises when someone is mixing
I/T and streaming write commands. How does the device know whether it
should set DWE for streamed write requests or go into the "abort all
commands" mode for I/T type write commands if a failure should occur. DWE
is not defined on I/T write command error status.
Also, correct me if I am wrong, but the Write Cache Enable in SET FEATURES
controls write cache for both I/T and Stream write commands? Should there
be a new command for Enable/Disable Streaming Write Cache? Doesn't
address
the problem above, just complicates it more. Some people still like to
turn
off Write Cache and this would affect Streaming Write commands. Just a
thought.
Gary Laatsch
-----Original Message-----
From: Mclean, Pete [mailto:[EMAIL PROTECTED]]
Sent: Friday, July 27, 2001 2:57 PM
To: 't13'
Subject: [temp t13] Minutes of the July 24-26 working group
** This is the quasi-official and semi-temporary T13 email list server.
**
What follows is a text version of the minutes for this week's working
group
meeting. I will upload them to the FTP site as soon as the FTP site lets
me.
During the editorial review, a number of non-editorial issues and action
items were encountered. Everyone who is planning to attend the August
plenary is encouraged to review these issues and action itens and be
prepared to address them. Also, we would like to do a change bar review of
revision 2 that will be out next week and a line-by-line of Annex C. If
folks can review these ahead of time our likelyhood of completing these
tasks improves.
Pete
-------------------------------------------------
Document Number: T13/e01130r0
Date: 26 July 2001
Reply to: Pete McLean
To: Membership of T13
From: Mark Evans, Secretary of T13, and Pete McLean, Chair of T13
Subject: Minutes of T13 Ad Hoc Working Group at Irvine, CA
July 24 - July 26, 2001
Agenda
1 Opening remarks 2
2 Approval of agenda 2
3 Membership and attendance 2
4 Document distribution 2
5 ATA/ATAPI-6 project 1410-D, revision 1e 2
5.1 Editorial review of the document 3
6 Issues List 3
7 Review of Action Items 4
8 Call for Patents 4
9 Meeting Schedule 5
9.1 Long term meeting schedule 5
10 Adjournment 5
Results of the Meeting
1. Opening remarks
Pete McLean, the T13 Chairman, called the meeting to order at 10:00 A.M.
on
Tuesday, July 24, 2001. Pete thanked Ron Salmon of Phoenix for hosting
the
meeting.
As is customary, the people in attendance introduced themselves, and a
copy
of the attendance list was circulated. Copies of the draft agenda were
made
available to those in attendance.
The Chairman briefly reviewed the meeting rules: T13 is a Technical
Committee of NCITS and operates under NCITS rules. Anyone with interest
in
the work of the committee is free to attend and take part in discussions.
However, motions and votes are limited to member organizations, one vote
per
member organization. In addition, the committee complies with NCITS
antitrust guidelines. As a result discussions about unannounced products,
prices, volumes, vendor relations, etc., are prohibited. The Chairman
noted
that a set of Policies and Procedures for NCITS are available at the
www.ncits.org site under the standing documents category as SD-2. The
committee is operating under these revised policies and procedures
augmented
by the T13 procedures at the T13 web site www.t13.org.
2. Approval of agenda
The agenda as reflected in these minutes was approved unanimously.
3. Membership and attendance
The Chairman reviewed the rules for membership and attendance. Attendance
at a working group meeting does not affect membership status in T13.
The following members were in attendance:
Dan Colegrove IBM Gary Laatsch Phoenix
Mark Evans Maxtor Ron Salmon Phoenix
Pete McLean Maxtor Greg Elkins Qlogic
Tony Goodfellow Pacific Digital Bob Chang Sandisk
Curtis Stevens Pacific Digital Tasuku Kasebayashi Toshiba
4. Document distribution
The following documents were distributed:
Agenda
In addition diskettes were passed around with soft copies of the above
documents.
5. ATA/ATAPI-6 project 1410-D, revision 1e
1) Editorial review of the document
The editor reviewed the current revision of the draft standard.
During the review many modifications and corrections were recommended
which
the editor agreed to include in the next revision of the draft standard.
Change bars will identify text changes only.
Dan Colegrove moved and Ron Salmon seconded that the working group
recommend
to the T13 Plenary that the editorial changes made at this meeting be
incorporated into the next revision (revision 2) of the draft standard.
The motion carried unanimously.
6. Issues List
During the review of ATA/ATAPI-6 revision 1e the following issues were
identified. These issues will be resolved at the T13 plenary meeting in
August.
1. 6.6.2.3 Ultra DMA burst termination phase rules, item 2): Should
DEVICE RESET be added as something that would allow a host to abort an
Ultra
DMA transfer?
2. 6.15.2 BIOS locking SET MAX, in the description for state
SM2:Set_Max_Locked: transition SM2:SM3, "T13 intends to remove this
transition in ATA/ATAPI-6." What does this mean?
3. 6.23.1 Streaming commands: The last sentence in the seventh
paragraphs reads, "The SET FEATURES command to enable/disable caching may
not affect caching for streaming commands." What should happen to the
caching function of Streaming when non-Streaming caching is disabled or
enabled? May it have an effect or shall it have no effect?
4. 7.8.3 Device Control register, Access restrictions: Curtis
Stevens
proposed that access to the Device Control register when DMACK- is
asserted
should be allowed, though the words in the standard prohibit this access.
5. 8.17.18 IDENTIFY PACKET DEVICE, Word 49: Capabilities, bit 9: Is
defining LBA translation supported for ATAPI necessary? Is this
translation
ever used for this device type?
6. Is the Host Protected Area feature set ever used for ATAPI
devices?
Shouldn't this function be done via SCSI methods for these devices? If
so,
shouldn't use of this feature set be prohibited for use by these devices?
7. There was strong feeling at the working group for making the C/D
and
I/O bits obsolete for READ and WRITE DMA QUEUED and QUEUED EXT.
8. The entries in the Delayed LBA Log may be affected by changes in
the
Streaming Performance Parameters Log or other external environmental
factors. Should this be more clearly defined? If so, how?
9. Can sectors be removed from the log if found later to be good
based
on the above exceptions?
10. What is the response if RC is set and the transfer time limit
expires, and what happens if RC is not set and the time limit expires.
Should all of the data be transferred? How should the error be reported?
11. How should errors be reported for Stream commands in general? When
should the task file be used for error location reporting?
12. How should the SET FEATURES subcommand for Enable/disable
reverting
to defaults affect settings, e.g., Ultra DMA modes?
13. Didn't we make SMART SAVE ATTRIBUTE VALUES obsolete?
14. When should DWE be cleared after being set?
15. Which errors are reported for DWE? It is extremely difficult (and
a
new requirement not placed on I/T data type commands) for devices to
maintain context for all dirty cache content.
16. Which errors go into the Write Stream Log? (see above item)
7. Review of Action Items
1. Folks should think about the value reported for READ and
WRITE MULTIPLE commands: what really should be the default value? Is it
in
word 47 or 59 in IDENTIFY DEVICE data?
2. What is the value in word 59, bits (7:0) in IDENTIFY
DEVICE
data (i.e., the current multiple value) at power-up or after a hardware
reset, if bit 8 in word 59 is set to one?
3. Mark Evans is to come up with a clear definition for the
self-test failure checkpoint byte (see 8.30.8.5.3 and 8.55.6.8.4.2).
8. Call for Patents
The Chairman explained the patent policy. A document is available from
ANSI, "Procedures for the Development and Coordination of American
National
Standards", at no charge. This document is also on the WWW at
ftp://www.ncits.org/help/ansi_sdo.html. Section 1.2.11 contains the ANSI
patent policy. Amy Marasco manages patent issues for ANSI and can be
contacted at [EMAIL PROTECTED] or 212-642-4954.
The Chairman then called for patents and/or patent applications that may
be
necessary to comply with the requirements in T13 standards, draft
standards,
technical reports, or draft technical reports.
There were no responses to the call for patents.
9. Meeting Schedule
1) Long term meeting schedule
The following is the proposed schedule through CY2003. Requests for hosts
will be discussed, and the group will vote on approval of the CY2003
meeting
dates at the next T13 plenary meeting in August.
CY 2001 Dates Location Host Contact
August 21-23, 2001 Longmont, CO Maxtor Pete McLean
October 23-25, 2001 Kauai, HI Pacific Digital Tony Goodfellow
December 11-13, 2001 Irvine, CA Western Digital Darrin Bulik
CY 2002 Dates Location Host Contact
February 19-21, 2002 Placerville, CA Adaptec Ron Roberts
April 23-25, 2002 Longmont, CO Seagate Marc Noblitt
June 25-27, 2002 Irvine, CA Phoenix Curtis Stevens
August 20-22, 2002 Redmond, WA Microsoft TBD
October 22-24, 2002 Maui, HI Pacific Digital Tony Goodfellow
December 16-18, 2002 Irvine, CA Western Digital Darrin Bulik
CY 2003 Dates Location Host Contact
February 18-20, 2003
April 22-24, 2003
June 24-26, 2003
August 19-21, 2003
October 21-23, 2003
December 16-18, 2003
Note: 2) Meetings are scheduled for three days only. 3) Due to the
change in the Dell membership, the April, 2002 meeting has been changed
from
Dell hosting in Austin to Marc Noblitt of Seagate hosting this meeting in
Longmont, CO. 4) If Microsoft does not wish to host the meeting in
August, 2002, Pete McLean has volunteered to host this meeting in
Longmont,
CO.
As noted above the next plenary meeting of T13 is August 21st through 23rd
in Longmont, Colorado, hosted by Pete McLean of Maxtor. A host meeting
notice will be sent out on the reflector and will be added to the web
site.
10. Adjournment
The meeting adjourned by the Chairman, Pete McLean, upon completion of all
the agenda items at 4:30 PM. on Thursday, July 26, 2001.
--
If you have any questions or wish to unsubscribe send a
message to Hale Landis, [EMAIL PROTECTED] To post to
this list server send your message to [EMAIL PROTECTED]
For questions concerning Thistle Grove Industries or TGI's
list services please send email to [EMAIL PROTECTED]
--
If you have any questions or wish to unsubscribe send a
message to Hale Landis, [EMAIL PROTECTED] To post to
this list server send your message to [EMAIL PROTECTED]
For questions concerning Thistle Grove Industries or TGI's
list services please send email to [EMAIL PROTECTED]
--
If you have any questions or wish to unsubscribe send a
message to Hale Landis, [EMAIL PROTECTED] To post to
this list server send your message to [EMAIL PROTECTED]
For questions concerning Thistle Grove Industries or TGI's
list services please send email to [EMAIL PROTECTED]
--
If you have any questions or wish to unsubscribe send a
message to Hale Landis, [EMAIL PROTECTED] To post to
this list server send your message to [EMAIL PROTECTED]
For questions concerning Thistle Grove Industries or TGI's
list services please send email to [EMAIL PROTECTED]
