SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


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

    RE: iSCSI: login issue - partial consensus call


    • To: <ips@ece.cmu.edu>
    • Subject: RE: iSCSI: login issue - partial consensus call
    • From: "Martin, Nick" <Nick.Martin@compaq.com>
    • Date: Thu, 6 Sep 2001 15:33:14 -0500
    • content-class: urn:content-classes:message
    • Content-Transfer-Encoding: 8bit
    • Content-Type: text/plain;charset="us-ascii"
    • Sender: owner-ips@ece.cmu.edu
    • Thread-Index: AcE3BdDjwRzS7qL3EdWqmgBQiwLxogAAeApQ
    • Thread-Topic: iSCSI: login issue - partial consensus call

    Santosh,
    
    If an iSCSI HBA driver is going to live within the SCSI subsystems in
    today's operating systems, it will be told almost nothing about the host
    through those interfaces.  When the driver initializes, it is likely to
    be told only the PCI addresses of its adapter(s).  It is provided
    interfaces only to expose its SCSI request, ioctl request, and hardware
    interrupt service routines.  It will not get the host name, or an IP
    address via interfaces in the SCSI subsystem.  Interfaces to collect
    information such as IP address, and host name which are available to
    network drivers, are not available to many storage drivers.  These and
    anything else needed will have to be collected from other interfaces.
    Some OS environments will provide a way for the SCSI driver to initiate
    queries for this information, some will allow static pre-configuration,
    others will need to wait for something to poke the information in
    through custom ioctl interfaces, or retrieve or construct what is needed
    from non-volatile storage on the HBA card.  
    
    This is partly a chicken and egg problem.  The host name for example is
    normally retrieved from the system disk which can not be read until the
    storage driver has completed its initialization.  If the boot storage
    controller could ask the OS for the host name, it would not get a good
    answer.
    
    Operating systems will eventually comprehend the need for driver APIs
    for iSCSI which blend network and storage requirements, but initial
    implementations will be entirely supplied by the HBA or driver vendor.
    
    IMHO, if a driver can initialize (bootstrap) without information from
    other sources, you make life much easier for the driver writers,
    testers, and users.  If all iSCSI drivers operating within the same host
    need to agree on something which is not available through a pre-existing
    interface, there will be problems.  If the administrator has to manually
    set the same InitiatorName (and/or AccessID) for each iSCSI driver, that
    is one thing.  If he has to manually set non-overlapping ranges of ISIDs
    for each driver instance that will be more error prone.
    
    The suggestion by Michael Schoberg to eliminate initiator assigned ISID,
    and only use TSID which is already guaranteed to be unique makes sense
    to me.  
    
    As Michael noted, we are a bit off of the consensus topic.
    
    Thanks,
    Nick
    -----Original Message-----
    From: Santosh Rao [mailto:santoshr@cup.hp.com]
    Sent: Thursday, September 06, 2001 1:57 PM
    To: ips@ece.cmu.edu
    Subject: Re: iSCSI: login issue - partial consensus call
    
    
    "Martin, Nick" wrote:
    
    > The problem is that iSCSI does not and will not specify how two HBAs
    > from different vendors installed in the same Initiator should or could
    > get a range of ISIDs for their exclusive use.  This will be operating
    > system specific and vendor defined.  It is uncertain that the same
    tool
    > or repository would be used by all HBA vendors in any environment.
    > Given this, accidental overlap in ISID space is not unlikely.
    
    In the above scenario, if the HBAs do not have an interface to receive a
    range of ISIDs from the OS iscsi driver, then, they must also lack an
    interface to be handed down an iscsi initiator node name from the
    OS driver. Without using a common iscsi initiator node name, the ISID
    issue
    does not arise for multiple HBAs. [since each would be using its own
    iscsi
    name].
    
    OTOH, if the HBAs have an interface to be handed down an iscsi initiator
    node name from the OS driver, there's no reason for them to not provide
    an
    interface to be handed down an ISID or a range of ISIDs for their use.
    
    I am in favor of option (A) since it is the simplest and a well designed
    HBA should be providing interfaces for the OS driver to assign the HBA
    an
    iscsi initiator node name as well as a range of ISIDs, thus, rendering
    Nick's scenario unlikely.
    
    >
    > Given that there is no one way to play right, we must make sure that
    > everyone can at least play nice.
    >
    > My expectation is that sessions are infrequently established and long
    > lived.  ISIDs may be re-used at will by their current owners.  When no
    > "already owned" ISIDs are available, or an attempt to re-use an
    "already
    > owned" ISID failed, and HBA would need to a) "probe" for a new
    available
    > ISID or b) fail the request to establish the session.  Session
    recovery
    > should not be attempted unless a session is known to have failed.
    >
    > If tools are available, and the administrator has used them correctly,
    > then HBAs will not collide in ISID space.  If the tools are not
    > available or were not used correctly, I would hope the second HBA can
    > still attempt to come up without impacting the sessions established by
    > the first.
    
    If the tools were not available, how does the 2nd HBA get assigned the
    same
    iscsi initiator node name as the first ? The ISID issue is only
    applicable
    if both HBAs are sharing the same iscsi initiator node name.
    
    I would like to understand the rationale behind an HBA interface
    allowing
    an initiator node name to be assigned, but not an ISID or a range of
    ISIDs
    (?).
    
    
    Regards,
    Santosh
    
    
    


Home

Last updated: Fri Sep 07 12:17:11 2001
6423 messages in chronological order