[Stgt-devel] vtl patch.

[Mark Harvey]

On 4/11/07, FUJITA Tomonori <fujita.tomonori at lab.ntt.co.jp> wrote:
> From: "Mark Harvey" <markh794 at gmail.com>
> Subject: Re: [Stgt-devel] vtl patch.
> Date: Wed, 11 Apr 2007 12:06:18 +1000
>
> > On 4/11/07, FUJITA Tomonori <fujita.tomonori at lab.ntt.co.jp> wrote:
> > > From: "Mark Harvey" <markh794 at gmail.com>
> > > Subject: [Stgt-devel] vtl patch.
> > > Date: Thu, 5 Apr 2007 09:56:10 +1000
> > >
> > > > Apologies for this being an attachment. It is rather large to include
> > > > inline (57k).
> > > >
> > > > Follow up to feedback from earlier postings, I have implemented the following:
> > > >
> > > >
> > > > This patch includes the following changes.
> > > >
> > > > - Patch to tgtadm so multiple '--name, --value' arguments can be passed.
> > >
> > > Why do we need multiple arguments?
> > >
> > > I prefer to keep the current semantics:
> > >
> > > $ tgtadm --lld iscsi --mode logicalunit --op update --tid=1 --lun=0 \
> > >          -n VendorIdent -v QUANTUM
> > > $ tgtadm --lld iscsi --mode logicalunit --op update --tid=1 --lun=0 \
> > >          -n ProductIdent -v SDLT600
> > > $ tgtadm --lld iscsi --mode logicalunit --op update --tid=1 --lun=0 \
> > >          -n ProductRev -v 0001
> > > $ tgtadm --lld iscsi --mode logicalunit --op update --tid=1 --lun=0 \
> > >          -n SerialNumber -v XYZZY10
> > >
> > > rather than:
> > >
> > > tgtadm --lld iscsi --mode logicalunit --op update --tid=1 --lun=0 \
> > >         -n VendorIdent -v QUANTUM       \
> > >         -n ProductIdent -v SDLT600      \
> > >         -n ProductRev -v 0001           \
> > >         -n SerialNumber -v XYZZY10
> > >
> > >
> > > The rest looks ok.
> > >
> >
> > Thanks for the feedback.
> >
> > While setting the Vendor Ident etc can be easly acheived one field at
> > a time, configuring the parameters for the SMC device requires
> > multiple parameters at the same time.
> >
> >
> > e.g.
> > > (Reserve address space for storage slots)
> > > --name=ElementType --value=[1|2|3]
> > > --name=StartAddress --value=number
> > > --name=Quanity --value=number
> > > --name=Sides --value=[1|2]          (Single/double sided)
> >
> > I need to get the element type, starting address and number of slots
> > and if media is single/double sided (or at minimum, the element type
> > and the element address - if configuring a slot per instance of
> > tgtcmd)
>
> I can't find "ElementType" in ssc3_config or smc_config. Will you
> implement it later?

Yes. I am in the process of implementing the this now. Nothing that
actually compiles, let alone run.

I am working on implementing the MODE SENSE/MODE LOG command, with
initial config of page 0x1d (Element Address Assignment).

I hope to post something new tomorrow so what was previously
'theoretical' will have some code example with what I'm trying to
achieve.

> Do you need a configuration like 'only when X is true, Y is
> necessary'?  I think that you can still do that with the current
> semantics, 'one field at a time'. You just need to check whether the
> configuration is valid when a user tries to start a target. Though
> with a 'multiple fields at a time' scheme, the checking would be
> a bit easier.
>
> I don't like '--name a --value b --name b --value c ...' because it
> doesn't look like the standard (GNU) option way.

I can't say I like it either, however it does seem to be the cleanest
way.. I'm open to suggestions.


Only other idea I had was to group up all non-processed arguments from
the getopt_long(), re-format into a argc/argv[] type and pass this
thru to the SMC/SSC module for additional processing. This would allow
the SMC/SSC module to perform their own getopt_long()

Regards
Mark