Interpretation of Logical unit Reset behavior

ed nadolski ed.nadolski at sun.com
Tue Oct 22 11:45:25 PDT 2002


* From the T10 Reflector (t10 at t10.org), posted by:
* ed nadolski <ed.nadolski at Sun.COM>
*
If the TARGET RESET does more than a LOGICAL UNIT RESET to each LU, is it a
problem if TARGET RESET is optional?  There might be no way to reset
certain parameters (such as target port negotiated settings) using LOGICAL
UNIT RESET alone.

>From sam3r03 (7.8 TARGET RESET, p. 90):

> NOTE 11 - Previous versions of this standard required TARGET RESET
> support in all SCSI target devices. SCSI transport protocols may
> or may not require that TARGET RESET be supported. SCSI transport
> protocols may require additional actions beyond those specified
> here.

This sounds like it is up to each SCSI transport protocol to mandate TARGET
RESET support where necessary to ensure that all relevant parameters can be
reset. Is this correct?

It would also imply that initiators should not assume that a LOGICAL UNIT
RESET to each LU on a target is equivalent to a TARGET RESET.

Thanks,
Ed

-- 
Edmund Nadolski
Sun Microsystems Inc.
ed.nadolski at sun.com


"Elliott, Robert (Server Storage)" wrote:
> 
> * From the T10 Reflector (t10 at t10.org), posted by:
> * "Elliott, Robert (Server Storage)" <Elliott at hp.com>
> *
> That's what these sentences are trying to allow.
> 
> (spc2r24)
> 5.9.6 Hard reset:
> "Each SCSI transport protocol standard that defines reset events shall
> specify the SCSI target port's action in response to reset events."
> 
> 6.7 TARGET RESET:
> NOTE 12 - ... SCSI transport protocols may require additional actions
> beyond those specified here.
> 
> SPI-5 defines its actions with:
> (spi5r03)
> 4.12.3 When to negotiation
> A port shall set its negotiation required flags to true for all other
> ports after a reset event (see 12.5).
> ...
> After a reset event a port shall set its transfer agreements for all
> other ports to the default transfer agreement (see table 4).
> 
> 12.5.4 Target reset event
> When a SCSI device successfully receives a TARGET RESET message, it
> shall cause a reset event.
> 
> --
> Rob Elliott, elliott at hp.com
> Industry Standard Server Storage Advanced Technology
> Hewlett-Packard
> 
> > -----Original Message-----
> > From: Binford, Charles [mailto:CBinford at pirus.com]
> > Sent: Monday, October 21, 2002 6:11 PM
> > To: t10 at t10.org
> > Subject: RE: Interpretation of Logical unit Reset behavior
> >
> >
> > * From the T10 Reflector (t10 at t10.org), posted by:
> > * "Binford, Charles" <CBinford at Pirus.com>
> > *
> > Robert,  while I agree with your conclusion, I think it is
> > hard to justify your last statement, "TARGET RESET does the
> > target port resetting." from any words I can find in SAM-2.
> > Everywhere I looked Target Reset was merely a LUN reset to
> > each LU.  My reading says the result of a host sending a LUN
> > Reset to each LU, and a host sending a Target Reset would be
> > the same.  I don't see anything to suggest Target Reset
> > should have special handling of the port parameters.  What did I miss?
> >
> > Charles Binford
> > Pirus Networks
> > 316.315.0382 x222
> >
> >
> > -----Original Message-----
> > From: Elliott, Robert (Server Storage) [mailto:Elliott at hp.com]
> > Sent: Friday, October 18, 2002 6:08 PM
> > To: t10 at t10.org
> > Subject: RE: Interpretation of Logical unit Reset behavior
> >
> >
> > * From the T10 Reflector (t10 at t10.org), posted by:
> > * "Elliott, Robert (Server Storage)" <Elliott at hp.com>
> > *
> > This question was raised in March 2001 during the negotiation
> > rewrite, and
> >       01-128r2 (1 May 2001) SPI-4 reset cleanup
> > added the rule that LOGICAL UNIT RESET shall not affect the
> > negotiation.
> >
> >
> > Negotiated settings are a property of the target port shared
> > by all the logical units. In a target with multiple logical
> > units, reset of one of the logical units shouldn't affect the
> > others' ability to communicate.
> >
> > SAM-2 just says the logical unit should be returned to a
> > power-on-like state, not the target port.  TARGET RESET does
> > the target port resetting.
> >
> > --
> > Rob Elliott, elliott at hp.com
> > Industry Standard Server Storage Advanced Technology Hewlett-Packard
> >
> >
> >
> > > -----Original Message-----
> > > From: gerry.houlder at seagate.com [mailto:gerry.houlder at seagate.com]
> > > Sent: Friday, October 18, 2002 4:26 PM
> > > To: t10 at t10.org
> > > Subject: Interpretation of Logical unit Reset behavior
> > >
> > >
> > > * From the T10 Reflector (t10 at t10.org), posted by:
> > > * gerry.houlder at seagate.com
> > > *
> > > A contradiction has been pointed out for the description of
> > > the Logical Unit Reset behavior. In particular, note the
> > > requirements I have marked with double brackets << >>.
> > >
> > > From SPI - 4 or 5:
> > >
> > >    16.5.6 LOGICAL UNIT RESET
> > >    The LOGICAL UNIT RESET message is defined in the SCSI
> > Architecture
> > >    Model-2 standard.
> > >    Receipt of the LOGICAL UNIT RESET message after an I_T_L
> > > nexus has been
> > >    established is a logical
> > >    unit reset event as defined in the SCSI Architecture
> > > Model-2 standard.
> > >    In addition to the requirements in the SCSI Architecture Model-2
> > >    standard the SCSI target port shall go to
> > >    the BUS FREE phase following the successful receipt of the
> > > LOGICAL UNIT
> > >    RESET message.
> > >
> > >    <<A logical unit reset has no effect on the transfer agreement.>>
> > >
> > > From SCSI Architecture Model-2 standard
> > >
> > >    5.9.7 Logical unit reset
> > >    A logical unit reset is:
> > >    a) The action in response to a LOGICAL UNIT RESET task management
> > >    request (see 6.6) or some other
> > >    logical unit reset event; or
> > >    b) One of the actions in response to a TARGET RESET task
> > management
> > >    function (see 6.7) or a hard reset
> > >    (see 5.9.6).
> > >    The definition of logical unit reset events is dependent
> > > on the SCSI
> > >    transport protocol.
> > >    To process a logical unit reset the logical unit shall:
> > >    a) Abort all tasks as described in 5.7;
> > >    b) Clear a CA (see 5.9.1.6) or ACA (see 5.9.1.7)
> > > condition, if one is
> > >    present;
> > >    c) Release all reservations established using the reserve/release
> > >    management method (persistent reserva-tions
> > >    shall not be affected);
> > >    <<d) Return the logical unit's operating mode to the
> > > appropriate initial
> > >    conditions, similar to those conditions that
> > >    would be found following device power-on.>> The MODE
> > > SELECT parameters
> > >    (see SPC-2) shall be restored
> > >    to their last saved values if saved values have been
> > > established. MODE
> > >    SELECT parameters for which no
> > >    saved values have been established shall be returned to
> > > their default
> > >    values;
> > >    e) Set an unit attention condition (see 5.9.5); and
> > >    f) Initiate a logical unit reset for all dependent logical
> > > units (see
> > >    4.13).
> > >    In addition to the above, the logical unit shall perform
> > > any additional
> > >    functions required by the applicable standards.
> > >
> > > The wording from SPI says "no effect on transfer agreement",
> > > meaning the target has to remember the transfer agreement
> > > even though everything else is being reset. The SAM wording
> > > says "similar to power on", which certainly includes
> > > resetting the transfer agreement. These conditions certainly
> > > conflict with each other. How do we resolve the conflicting wording?
> > >
> > > The single LUN targets I am familiar with reset all the same
> > > things (including the transfer agreement) for both of these
> > > functions. Are these devices in violation of the standard?
> > >
> > > *
> > > * For T10 Reflector information, send a message with
> > > * 'info t10' (no quotes) in the message body to majordomo at t10.org
> > >
> > *
> > * For T10 Reflector information, send a message with
> > * 'info t10' (no quotes) in the message body to majordomo at t10.org
> > *
> > * For T10 Reflector information, send a message with
> > * 'info t10' (no quotes) in the message body to majordomo at t10.org
> >
> *
> * For T10 Reflector information, send a message with
> * 'info t10' (no quotes) in the message body to majordomo at t10.org

-- 
Edmund Nadolski
Sun Microsystems Inc.
ed.nadolski at sun.com
*
* 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