SRP and login with initiator port identifier already in use

Edward A. Gardner eag at ophidian.com
Wed May 23 19:57:31 PDT 2001


* From the T10 Reflector (t10 at t10.org), posted by:
* "Edward A. Gardner" <eag at ophidian.com>
*
I think I agree with Bob Nixon's response.  The following is somewhat more
detailed.

Multiple independent paths between the same initiator and target ports (the
same I_T nexus) mean just that.  They are between the *same* initiator and
target ports -- that is, there is only one initiator port and only one
target port.  Done confuse these with ports on IB adapters.

It's analogous to FCP with a fabric that uses multiple paths and therefore
provides out of order delivery.  Except that the multiple paths / channels
are visible to the initiator, so it can choose on a request by request basis
whether to enforce ordering (use the same path/channel) or allow out of
order behavior for performance (use any path/channel).

>The logical units behind the target treat the login as if a separate
>initiator port identifier were being used.  In most regards, everything
that
>comes in would run independently.

False.  All the paths have the same initiator port identifier.  All are part
of the same I_T nexus.  All share the same reservation, persistent
reservation and UA context.  One could also send a command down one path,
then abort it with an ABORT TASK down the other.  There's a race condition
with this, I don't recommend it, but one could do this -- might make a good
compliance test :-).

>If a reservation has been established, however, the permission would be
>checked only on the initiator port identifier sent during login.  Thus one
>reservation can apply to multiple "initiator ports".

It is true that "permission would be checked only on the initiator port
identifier sent during login".  That is the *only* initiator port identifier
that exists.  True, the reservation may apply to multiple paths/channels,
which may involve multiple IB adapters and/or ports.  But those are *not*
SCSI initiator ports.

>UAs and other things for "other initiators" would treat this new channel
>nexus as having a different initiator.

False.  It's the same initiator port.

Edward A. Gardner               eag at ophidian.com
Ophidian Designs                719 593-8866 voice
1262 Hofstead Terrace           719 593-8989 fax
Colorado Springs, CO  80907     719 210-7200 cell


*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org




More information about the T10 mailing list