SAM UML and question about command priority and Task manager/device server

George Penokie george.penokie at avagotech.com
Tue Jun 10 14:17:12 PDT 2014


Formatted message: <a href="http://www.t10.org/cgi-bin/ac.pl?t=r&f=r1406103_f.htm">HTML-formatted message</a>

You guys are so far out in the weeds that there is no point in me even
saying anything.
The state machine wording proposal is going remain like it is.
That means that the device server will be the only class that determines
the processing order of commands it knows about.
The task manager will only control when the device server knows about those
commands based on the simple/ordering rules for commands currently in SAM-5.
Bye for now,
George Penokie
Avago Technologies
3033 41 St NW
Rochester , MN 55901
507-328-9017
george.penokie at avagotech.com
On Tue, Jun 10, 2014 at 3:06 PM, Kevin D Butt <kdbutt at us.ibm.com> wrote:
> Fred,
>
> If your assertion is correct (and I must admit I cannot find anything in
> SAM to disprove it), then as SAM is written today, an application client
> can send cmd1, cmd2, cmd3, abort(cmd3) and the device server can process
> those requests as abort(cmd3), cmd1, cmd2, cmd3 and the application client
> has been told cmd3 was aborted and then at a later time will get a STATUS
> for cmd3.
>
> On the device, cmd1, cd2, and cdm3 have all been processed and the medium
> potentially change by cmd3 with the application client non the wiser.
>
> Is this what you are saying?
>
> Thanks,
>
> Kevin D. Butt
> SCSI Architect, Tape Firmware, T10 Standards
> Data Protection & Retention
> MS 6TYA, 9000 S. Rita Rd., Tucson, AZ 85744
> Tel: 520-799-5280
> Fax: 520-799-2723 (T/L:321)
> Email address: kdbutt at us.ibm.com
> http://www-03.ibm.com/servers/storage/
>
>
>
> From:        "Knight, Frederick" <Frederick.Knight at netapp.com>
> To:	     Kevin D Butt/Tucson/IBM at IBMUS,
> Cc:	     "george.penokie at lsi.com" <george.penokie at lsi.com>
> Date:        06/10/2014 12:40 PM
> Subject:	  RE: SAM UML and question about command priority and Task
> manager/device server
> ------------------------------
>
>
>
> That sub-clause covers commands also.  It asserts the lack of ordering
> requirements for everything (commands and TMFs).
>
>
>
> Since a command sent before the TMF to abort it could actually arrive at
> the device after the TMF arrives, I’m not sure I agree.
>
>
>
> The task manager / device server would be trying to abort a command that
> the device doesn’t know about yet (because the TMF arrived before the
> command arrived).  And, by the time the command is received, the TMF could
> have already completed.
>
>
>
> I don’t see anything in SAM that prevents this.  I know that forcing
> ordering makes things a lot easier, but I don’t see that SAM does that;
in
> fact, it seems that is explicitly does NOT require it.
>
>
>
>		  Fred
>
>
>
> *From:* Kevin D Butt [mailto:kdbutt at us.ibm.com <kdbutt at us.ibm.com>]
> * Sent:* Monday, June 09, 2014 6:22 PM
> * To:* Knight, Frederick
> * Cc:* george.penokie at lsi.com; owner-t10 at t10.org; T10 Reflector
> * Subject:* RE: SAM UML and question about command priority and Task
> manager/device server
>
>
>
> Fred,
>
> This statement talks to the ordering of one task management request
> related to another.  My ordering comment was a task management request
> related to the command at which it is targeted.  Therefore, my statement is
> that the task manager is required to manage commands and task management
> requests in such a manner as to ensure that a task management request aimed
> at a command received by the task manager prior to the task management
> request not be sent to the device server prior to the command that it is
> aborting.
>
> Thanks,
>
> Kevin D. Butt
> SCSI Architect, Tape Firmware, T10 Standards
> Data Protection & Retention
> MS 6TYA, 9000 S. Rita Rd., Tucson, AZ 85744
> Tel: 520-799-5280
> Fax: 520-799-2723 (T/L:321)
> Email address: *kdbutt at us.ibm.com* <kdbutt at us.ibm.com>
> *http://www-03.ibm.com/servers/storage/*
> <http://www-03.ibm.com/servers/storage/&gt;
>
>
>
> From:        "Knight, Frederick" <*Frederick.Knight at netapp.com*
> <Frederick.Knight at netapp.com>>
> To:	     Kevin D Butt/Tucson/IBM at IBMUS, T10 Reflector <*t10 at t10.org*
> <t10 at t10.org>>, "*george.penokie at lsi.com* <george.penokie at lsi.com>" <
> *george.penokie at lsi.com* <george.penokie at lsi.com>>,
> Date:        06/09/2014 03:15 PM
> Subject:	  RE: SAM UML and question about command priority and Task
> manager/device server
> Sent by:	  <*owner-t10 at t10.org* <owner-t10 at t10.org>>
> ------------------------------
>
>
>
>
> I don’t think your ordering assumptions align with what SAM says (there
> are no ordering requirements for TMFs in SAM).  It is up to the application
> client to do any ordering (if it really wants it).  Here is what SAM
> already says about this:
>
>
>
> *4.4.3 Server request/response ordering*
>
> *  …*
>
> The order in which task management requests are processed is not specified
> by the SCSI architecture model.
>
> The SCSI architecture model does not require in-order delivery of such
> requests or processing by the task
>
> manager in the order received. To guarantee the processing order of task
> management requests referencing a
>
> specific logical unit, an application client should not have more than one
> such task management request pending
>
> to that logical unit.
>
>
>
>		  Fred
>
>
>
> *From:* *owner-t10 at t10.org* <owner-t10 at t10.org> [
> *mailto:owner-t10 at t10.org* <owner-t10 at t10.org>] *On Behalf Of *Kevin D
> Butt
> * Sent:* Monday, June 09, 2014 5:20 PM
> * To:* T10 Reflector; *george.penokie at lsi.com* <george.penokie at lsi.com>
> * Subject:* SAM UML and question about command priority and Task
> manager/device server
>
>
>
> In the last SAM conference call, a couple of us were thinking that the
> command priority should not only be determined by the device server but
> also the task manager.
>
> I was one of those who thought this.	I have since repented of that sin.
> ;-)
>
> I now believe that allowing the task manager to be part of prioritizing
> the commands would actually break the model and it should not be allowed.
> After reviewing this internally, it was pointed out to me that the task
> manager must maintain proper serialization in some aspect if it is going to
> be able to properly manage task management functions.  Those TM functions
> need to be sent after the commands and not before.  This infers that the
> commands and task management functions must be serialized by the task
> manager (in some fashion).  If command priority were to be managed by the
> task manager, this would either break that model or at least make it much
> more complex.
>
> Thanks,
>
> Kevin D. Butt
> SCSI Architect, Tape Firmware, T10 Standards
> Data Protection & Retention
> MS 6TYA, 9000 S. Rita Rd., Tucson, AZ 85744
> Tel: 520-799-5280
> Fax: 520-799-2723 (T/L:321)
> Email address: *kdbutt at us.ibm.com* <kdbutt at us.ibm.com>
> *http://www-03.ibm.com/servers/storage/*
> <http://www-03.ibm.com/servers/storage/&gt;
>
>



More information about the T10 mailing list