SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


    [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

    Re: iscsi : default iscsi mode page settings - Consensus call



    Ralph Weber wrote:
    
    > I read SPC-2 to promise applications that such
    > a value will appear in the proper field in the
    > Disconnect-Reconnect mode page.
    
    Ralph,
    
    I don't see where in SPC-2 is it stated that "such a value
    (MaximumBurstSize) will appear in the disconnect-reconnect mode page of
    each protocol." 
    
    I read the following in SPC-2 definition of disconnect-reconnect mode
    page :
    
    "The parameters appropriate to each protocol and their interpretation
    for that protocol may be specified in the individual protocol
    documents."
    
    Further below in that same section :
    
    "If a parameter that is not appropriate for the specific protocol
    implemented by the SCSI device is non-zero, the device server shall
    return CHECK CONDITION status. The sense key shall be set to ILLEGAL
    REQUEST and the asc set to ILLEGAL FIELD IN PARAMETER LIST."
    
    Since there are other fields in this mode page that differ from one SCSI
    transport to another, I don't see how an application that is performing
    a mode sense/select operation can be guaranteed reliable semantics when
    it uses the SPC-2 definition of this page.
    
    Any such application will need to be aware of all the SCSI transport
    specific versions of this page and use the appropriate version based on
    the transport in use.
    
    Also, the mode sense(6) & mode sense(10) sections state that :
    "If an application issues a mode sense command with a page code that is
    not supported by the target, the device server shall return CHECK
    CONDITION status and the sense key shall be set to ILLEGAL REQUEST and
    the asc to INVALID FIELD IN CDB."
    
    The above implies that specific implementations are allowed to not
    support certain page codes [like the disconnect-reconnect mode page]. In
    the case of iscsi, a class of targets (iscsi targets) will not be
    supporting this page.
    
    Again, from the above, I don't see how an application would be
    guaranteed that every SCSI target would support the disconnect-reconnect
    mode page. 
    
    Please help me understand where in the SPC-2 definition is the
    application assured that :
    
    a) the disconnect-reconnect page code is required to be supported on all
    targets.
    
    b) all or any specific fields of the disconnect-reconnect field
    definitions will be available to the application in each of the SCSI
    transports.
    
    Thanks,
    Santosh
    
    
    
    > Recourse to the SPI-4 definition of First Burst Size
    > is a lame defense since SPI-4 simply has no First
    > Burst Size concept to present in the Disconnect-
    > Reconnect mode page. First Burst Size was invented
    > for FCP-2 and never had any meaning for the
    > parallel SCSI bus.
    > 
    > Since SCSI-2 there has been a promise that the
    > Maximum Burst Size field in the Disconnect-Reconnect
    > mode page will have meaning. Any application that
    > sees zero (the preferred value for "reserved") in
    > that field has every right to assume it means what
    > SPC-2 says zero means, to whit: "A value of zero
    > indicates there is no limit on the amount of data
    > transferred per data transfer operation."
    > 
    > I disapprove of the proposal that First Burst Size
    > be made "reserved" in the Disconnect-Reconnect
    > mode page, since iSCSI has a First Burst Size
    > value that could appear in the mode page. But
    > owning to the parallel SCSI usage, the most that
    > can be said about iSCSI making the field reserved
    > is that doing so would be very bad form.
    > 
    > There is no basis for insisting that either of
    > these parameters be "changeable" via a MODE SELECT
    > command, and I leave that choice to iSCSI.
    > 
    > However, you all might consider remaining silent on
    > the subject and thus allowing vendors whose customers
    > have real needs the option of making the parameters
    > changeable without failing to meet iSCSI compliance
    > tests.
    > 
    > Knowing when to say nothing also is part of the
    > art of standards writing.
    > 
    > Thanks for your consideration.
    > 
    > Ralph Weber
    > SPC-2 technical editor
    
    -- 
    ##################################
    Santosh Rao
    Software Design Engineer,
    HP-UX iSCSI Driver Team,
    Hewlett Packard, Cupertino.
    email : santoshr@cup.hp.com
    Phone : 408-447-3751
    ##################################
    


Home

Last updated: Mon Oct 01 13:17:20 2001
6920 messages in chronological order