[Stgt-devel] Initial patch to add Tape and Library support

[Mark Harvey]

On 3/23/07, Douglas Gilbert <dougg at torque.net> wrote:
> FUJITA Tomonori wrote:
> > From: "Mark Harvey" <markh794 at gmail.com>
> > Subject: [Stgt-devel] Initial patch to add Tape and Library support
> > Date: Thu, 22 Mar 2007 13:37:57 +1100
> >
> >> Sorry for the 'attachment' but it seems a little too big to include in-line.
> >>
> >> I can chop it up if requested.
> >>
> >> Note: Most (read all except Inquiry) commands are not supported yet.
> >
> > diff -urN -x TAGS -x .svn reference/doc/README.vtl vtl/doc/README.vtl
> > --- reference/doc/README.vtl  1970-01-01 10:00:00.000000000 +1000
> > +++ vtl/doc/README.vtl        2007-03-22 10:57:24.000000000 +1100
> > @@ -0,0 +1,35 @@
> > +Initial support for SCSI Streaming and SCSI Medium Changer devices.
> > +markh794 at gmail dot com
> > +mark_harvey at symantec dot com
> > +
> > +
> > +To set up a changer and tape device..
> > +./usr/tgtadm --lld iscsi --op new --mode target --tid 1 -T iqn.2007-03:marks-vtl_smc:`hostname` --target-type changer
> > +./usr/tgtadm --lld iscsi --op new --mode target --tid 2 -T iqn.2007-03:marks-vtl_ssc:`hostname` --target-type tape
> > +
> > +I'm hijacking the 'backing device' for use as the configuration file.
> >
> > This isn't a good idea. tgt doesn't use any configuration files (or
> > database). Everything should be configured via tgtadm. We leave
> > storing configurations to other projects like Openfiler.
>
> Do the current stgt kernel based transport drivers come
> up in an "off the bus state"? If they are visible to
> the transport before tgtd and tgtadm have got their acts
> together then you have a problem.
>
> I noticed a problem with a MPT SAS HBA. Its firmware settings
> make it a SAS target and (perhaps prior to any mpt* drivers
> being loaded) SAS SMP inititiators in that SAS domain will
> be prompted to re-run discovery during which the firmware
> identifiers itself as a SSP target. Then an INQUIRY gets sent
> as _all_ SCSI targets should respond to that, even if there
> are no logical units.
>
> Doug Gilbert

I noticed a sense key of "Inquiry Data has changed" ASC/ASCQ of 3f/03
which I have always assumed would be used for this type of scenario.

It's always been on my 'list of things to test'. Just to see if the
Linux SCSI upper layers would actually re-issue an Inquiry and
re-configure to the updated settings.