[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Google
  Web www.spinics.net

Re: Response Fence Flag



Not really.  Current draft text is intentionally written to not have any dependencies on T10 dynamics.  The point is that iSCSI needs such a notion for succinctly describing the proper iSCSI protocol actions in a few places - ACA, TMF, Persistent reserve/Abort to name a few.  We certainly hope it will be approved by T10 and be a part of SAM-4 soon, but that isn't required per se for describing what iSCSI needs for its correct behavior.
 
IPS WG has adopted what it needs in the past - staying ahead of T10 review/approval cycle if necessary.  I_T nexus loss notification, iSCSI target/port naming, clearing effects are a few I recall.
 
Mallikarjun


 
----- Original Message ----
From: Eddy Quicksall <Quicksall_iSCSI@xxxxxxxxxxxxx>
To: Robert Snively <rsnively@xxxxxxxxxxx>; "Elliott, Robert (Server Storage)" <Elliott@xxxxxx>; ips@xxxxxxxx
Sent: Friday, January 5, 2007 8:58:47 AM
Subject: Re: Response Fence Flag

From an earlier email I think that Response Fence is only a proposal in T10 (http://www.t10.org:80/doc06.htm). If so shouldn't iSCSI wait a bit until this has been ratified?
 
Eddy
----- Original Message -----
From: Robert Snively
To: Elliott, Robert (Server Storage) ; ips@xxxxxxxx
Sent: Thursday, January 04, 2007 1:01 PM
Subject: RE: Response Fence Flag

The problem is that the definition of "previous" in this context is most likely flawed.
 
The "Response Fence" is merely an internal and untestable state machine input that
attempts to order the transmission of responses in the target port's queue.
It  does nothing to improve the overall ordering of the responses at the application
client.
 
Drivers for generic SCSI transports must always be aware that requested operations
report completion with a more or less arbitrary time relationship to each other and must be
designed to tolerate that. 
 
Using LOGICAL UNIT RESET as an example, the operation does not act only on
the device server.  As it passes down through the initiator stack, it may clean out
previously offered commands that have not yet been transported.  Then as it
passes up through the target stack, it may clean out previously offered commands
that have not yet been operated on.  As it passes through the device server
execution path, it may clean out previously offered commands that have begun
execution and some that have completed execution without yet responding.  As its
response passes back down the target stack, it may clean up commands
that have completed for which the response has not yet been transmitted, and as
it passes back up the initiator stack, it may clean up commands for which the
response has been transmitted, but not yet been made available to the driver or
application.  SAM-4 attempts to define the passage through the target port,
through the device server, and back to the target port.  However, beyond the
initiator port is outside the SAM-4 scope, and even where it is defined,
the "shall ensure"s are unlikely to be testable and in some cases not implementable. 
 
Will I sometimes get command complete indications for commands after
a LOGICAL UNIT RESET response is received?  You bet.  Response Fence doesn't
change that.
 
Will I sometimes get command aborted indications for commands after a
LOGICAL UNIT RESET is response is received?  You bet.  Response Fence
doesn't change that.
 
Is this a problem?  It never has been in the past, why is it now?
 
Can you test the corner conditions for a fence operation?  I doubt it.  The
existence of the "Response Fence" argument is internal to the target
implementation and invisible to testing.
 
Bob
 


From: Elliott, Robert (Server Storage) [mailto:Elliott@xxxxxx]
Sent: Wednesday, January 03, 2007 3:35 PM
To: ips@xxxxxxxx
Subject: RE: Response Fence Flag

SAM-3 and SAM-4 require all transport protocols to tag the TMF responses with the requests:

"Each SCSI transport protocol shall allow a Received Task Management Function Executed confirming completion of the requested task to be associated with the corresponding Send Task Management Request."

Although iSCSI was based on SAM-2, it complies with that rule. The Fence concept is asking for more than that - it wants the target to be able to ensure that all previous commands/TMFs are complete before delivering a particular TMF response (e.g., for a LOGICAL UNIT RESET).  Since iSCSI doesn't have ACKs, the target must wait for the next PDU from the initiator with an updated ExpStatSN.

06-341r0 discussed an alternative - add a Delivery Result output to the Send Command Complete and Task Management Function Executed confirmations (as previously proposed in 04-072r0 for a different purpose). This would let the device server/task manager wait for all the previous commands/TMF responses to complete (be ACKed) before proceeding to make additional calls (e.g., Task Management Function Executed for a LOGICAL UNIT RESET).

However, iSCSI allows the target port to send a NOP-IN to force delivery of a NOP-OUT with ExpStatSN, rather than passively wait for a PDU to show up.  The device server/task manager must instruct the target port when to do this, and the Request Fence argument serves that purpose.  Target ports using protocols without such a force mechanism are still OK - they will just wait for protocol-specific delivery confirmations (e.g. ACKs).

--
Rob Elliott, elliott@xxxxxx
Hewlett-Packard Industry Standard Server Storage Advanced Technology
https://ecardfile.com/id/RobElliott


 


From: Robert Snively [mailto:rsnively@xxxxxxxxxxx]
Sent: Wednesday, January 03, 2007 4:46 PM
To: Elliott, Robert (Server Storage); ips@xxxxxxxx
Subject: RE: Response Fence Flag

I need a little tutorial here.  Note that this is an architectural hack that
is very "un-SCSI" and is not required if the response is qualified with identification
information associated with the original request.  I know that such qualification
exists for commands, but does it exist for task management requests in
iSCSI? 
 
I know that it does exist for parallel SCSI (which is strictly interlocked)
and for SAS and Fibre Channel SCSI, making such "fences" unrequired by
those technologies.
 
Note that Rob's previous revision of 06-341 is available on www.t10.org.
 
I would hate to see such a hack creep into the SCSI architecture.
 
Bob
 


From: Elliott, Robert (Server Storage) [mailto:Elliott@xxxxxx]
Sent: Wednesday, January 03, 2007 1:56 PM
To: ips@xxxxxxxx
Subject: RE: [Ips] Response Fence Flag

If T10 agrees, T10 proposal 06-341 will add it to SAM-4.
 

--
Rob Elliott, elliott@xxxxxx
Hewlett-Packard Industry Standard Server Storage Advanced Technology
https://ecardfile.com/id/RobElliott




From: Eddy Quicksall [mailto:Quicksall_iSCSI@xxxxxxxxxxxxx]
Sent: Wednesday, January 03, 2007 2:00 PM
To: ips@xxxxxxxx
Subject: Response Fence Flag

Section 3.3.1 talks about a Response Fence Flag:
 
SCSI protocol layer instructs the SCSI transport layer of a
"Response Fence" associated with the response in question when
the "Send Command Complete" protocol data service (SAM-2, clause
5.4.2) ...
 
But I don't see any reference to that in SAM-2.
 
Is this strictly an iSCSI flag? Where is the flag specified?
 
Eddy

_______________________________________________
Ips mailing list
Ips@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/ips
_______________________________________________
Ips mailing list
Ips@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/ips


__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
_______________________________________________
Ips mailing list
Ips@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/ips

[IETF]     [Linux iSCSI]     [Linux SCSI]     [Linux Resources]     [Yosemite News]     [IETF Announcements]     [IETF Discussion]     [SCSI]

Add to Google Powered by Linux