Re: IG clarifications: Login Response & Reject reason codes
"Mallikarjun C." <[email protected]>
| Newsgroups | gmane.ietf.ips |
|---|---|
| Message-ID | <[email protected]> |
Julian, As I read through relevant sections of 3720, I see the following in section 5.3: The Login Phase is only implemented via Login Request and Responses. The whole Login Phase is considered as a single task and has a single Initiator Task Tag (similar to the linked SCSI commands). Section 10.10.3 makes a similar point about text requests/responses. I do not thus believe that the current text allows the flexibility to pipeline multiple outstanding Login/Text Requests/Responses. That of course means that I guess I am reaching the opposite conclusion than the one I started with - the clarification the anonymous reviewer had wanted me to put in wrt Login Response is not necessary. If OTOH you believe we need to explicitly allow pipelining overriding this 3720 text, please let me know. Thanks. Mallikarjun ----- Original Message ---- From: Julian Satran <[email protected]> To: Mallikarjun C. <[email protected]> Cc: IPS <[email protected]> Sent: Friday, December 29, 2006 3:15:41 AM Subject: Re: [Ips] IG clarifications: Login Response & Reject reason codes Mallikarjun, If we allow for multiple outstanding Login Requests we will have to allow for multiple outstanding Login Responses. Regards, Julo "Mallikarjun C." <[email protected]> 28/12/06 21:45 To Julian Satran/Haifa/IBM@IBMIL cc IPS <[email protected]> Subject Re: [Ips] IG clarifications: Login Response & Reject reason codes Julian, Thanks for the inputs, I somehow overlooked to respond to this email earlier. I agree with you on the "Task in progress" Reject reason code. There is no Unless I hear new objections, I would like to note "Negotiation Reset" Reject reason code as obsolete. Based on the silence so far, like you, I do not believe it is in use. On the "no more than one outstanding Login Response" semantics, your comments seem to be around Login Request PDU (and I believe we have such text for Login Request PDU) - do you see any cases wherein there can be multiple outstanding Login Responses on the wire? The request I got was to clarify if there is a plausible use case. Thanks! Mallikarjun ----- Original Message ---- From: Julian Satran <[email protected]> To: Mallikarjun C. <[email protected]> Cc: IPS <[email protected]> Sent: Monday, December 11, 2006 1:34:41 PM Subject: Re: [Ips] IG clarifications: Login Response & Reject reason codes "Mallikarjun C." <[email protected]> wrote on 11/12/2006 20:15:23: > All: > > I received some offline feedback on the implementers' guide draft > from a few reviewers who preferred to be anonymous. Please review & comment. > > 1) RFC 3720 does not explicitly call out that there cannot be more > than one outstanding Login-Response PDU on one iSCSI connection at > any given time (although the C-bit text indirectly implies it). > > This is intentional. At the time we where playing with the idea of pipelining the login. However it is common practice to have a single outstanding Login Request. I think that the only thing that becomes problematic is the phase change request (there you can have only one outstanding). There is already text that says that all changes to key values become final only at the end (when consistency can be reasonably checked). > Section 10.10 on Text Request PDU (which should cover Login Request > PDU semantics as well) says: "An initiator MUST have at most one > outstanding Text Request on a connection at any given time." > Essentially, an analog for Login/Text Response is missing (or so it seems). > > 2) RFC 3720 does not specify the use case for Reject reason code > "Task in progress" (0x07). > > I vaguely recall we put in this reason code for task reassignment > attempts while a task is in progress, but then we subsequently added > a TMF response reason code for that case (Julian?). So I'm not sure > if reason code 0x07 is used by implementations any longer. > The reject can used when the initiator attempts to start a new task but a task with the same ITT is still active for those cases when the target can't be sure this is a protocol error (e.g., a race between a logout and a reissued SCSI command). > The other non-obvious case is that of a "negotiation reset" Reject > reason code. What is this used for by implementations, if at all? > If I don't hear any objections, I will deprecate these two reason codes. > The negotiating can't be continued by one of the parties but the partial results (e.g., previous stage) are OK and no renegotiation is deemed necessary. I have no clue if somebody uses it but I felt at the time that the purists will object if I'd suggest restarting the login :-) > Mallikarjun > > > > ____________________________________________________________________________________ > Do you Yahoo!? > Everyone is raving about the all-new Yahoo! Mail beta. > http://new.mail.yahoo.com > > _______________________________________________ > Ips mailing list > [email protected] > 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 __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com _______________________________________________ Ips mailing list [email protected] https://www1.ietf.org/mailman/listinfo/ips