RE: iscsi : default iscsi mode page settings.

["Sanjeev Bhagat (TRIPACE/Zoetermeer)" <sbhagat@tripace.com>]

John, julian, Santosh

Agreed, but it depends where the protocol specific code is implemented.

Let us take example of Windows 2k

the command startes from application and flows down to file system to SCSI
ULP to Fileter drivers, SCSIPORT driver to SCSI Miniport driver and then to
HBA. 

Now in order to implement iSCSI in windows the commands can be diverted into
two paths 

1. after the SCSI ULP (class drivers). A filter driver or a custom written
parallel SCSIPORT driver may divert it to (iscsi protocol specific) TDI
client driver and so to TCP /IP stack 

or 
2. 
  a.)it flows down untill SCSI Miniport driver which can further pass it
(iscsi protocol specific) TDI client driver and then to TCP transport layer
...... etc etc

or 
  b>.)alternatively for step 2 it can go directly to an iSCSI HBA (with
TCP/IP accelerator, Ethernet built in)

In case 1, 2a the FirstburstSize will be set in the iscsi protocol specific
TDI client driver and in case 2b it will be set in SCSI miniport itself.

Clearly it will be set only at the place where the transport protocol
specific layer.


Regards,
Sanjeev



-----Original Message-----
From: Santosh Rao [mailto:santoshr@cup.hp.com]
Sent: Wednesday, September 26, 2001 7:40 PM
To: hufferd@us.ibm.com
Cc: ips@ece.cmu.edu
Subject: Re: iscsi : default iscsi mode page settings.


John,

I suspect that in most cases these mode page settings would not be done, 
since most stacks are still SCSI-2 based.
Hence, any SCSI LLPs that do provide support for these transport specific
mode pages (ex : FCP-2 implementations) would probably be issuing these 
mode sense/select's from within the LLP (miniport driver, HBA driver, etc).

The file system and wedge drivers do not muck with these, as you already
pointed out.

Since these mode pages have transport specific semantics, any SCSI ULP
(ex : class driver) that attempts to issue mode sense/select for transport
specific mode pages would first need to interact with some transport
specific code to determine which transport is being used and what mode
page format is to be used. Thus, any such class driver would need an
upgrade to understand the iscsi transport specific mode pages.

Even if there are some management applications that may be trying to set
these parameters through pass thru mode sense/select commands, they would
need to now be upgraded to understand iscsi specific versions of these
transport mode pages. Further, any such management app. would also need to
develop mechanisms to issue pass-thru text commands or some other
mechanism if it wished to tweak iscsi parameters.

Hence, I agree with you that there should not be any legacy issue here.
Any layer or management app that mucks with iscsi parameters will need an
upgrade anyway and they might as well be upgraded to use a single
mechanism, login keys.

I think we've beaten this issue to death, and suggest that we call for
consensus on this issue and move onto other issues.

Regards,
Santosh

> 
> 
> Does anyone know of where a Legacy function would exist, in the Host which
> would do anything with FirstBurstSize?  I am assuming it would have to be
> in the SCSI Device Driver, and since we have iSCSI instead, that may not
be
> an issue.
> I could be wrong but I do not think that it occures in the SCSI Class
> Drivers.
> And I do not think it ocures in the File Systems, so where does this
Legacy
> function we are worried about occure?
> If it is in the SCSI Device Driver it is NOT an issue.
> 
> .
> .
> .
> John L. Hufferd
> Senior Technical Staff Member (STSM)
> IBM/SSG San Jose Ca
> Main Office (408) 256-0403, Tie: 276-0403,  eFax: (408) 904-4688
> Home Office (408) 997-6136
> Internet address: hufferd@us.ibm.com
> 
> 
> "CONGDON,PAUL (HP-Roseville,ex1)" <paul_congdon@hp.com>@ece.cmu.edu on
> 09/25/2001 10:14:12 PM
> 
> Sent by:  owner-ips@ece.cmu.edu
> 
> 
> To:   "'cbm@rose.hp.com'" <cbm@rose.hp.com>, ips <ips@ece.cmu.edu>
> cc:
> Subject:  RE: iscsi : default iscsi mode page settings.
> 
> 
> 
> 
> >Julian and Santosh,
> >
> >Whether FirstBurstSize should be a text key or not,
> >IMHO, is (mostly) dependent on the current implementations.
> >Santosh rightly pointed out several drawbacks of
> >mandating a new set of mode select/sense operations
> >that are totally iSCSI-specific since the legacy SCSI
> >stack (class drivers, services/midlayer) will have no
> >clue about them.  OTOH, if most of today's SCSI stacks
> >_do_ perform this mode parameter manipulation in disconnect-
> >reconnect mode page, that would be a strong argument to
> >leave FirstBurstSize as a mode page parameter.
> 
> Anything iSCSI specific should attempt to follow other iSCSI
> methods of configuration (e.g. text keys).  We should really
> be driving towards a single method of parameter negotiation
> and configuration - especially when the legacy methods
> have no clue about the new parameters and would require upgrades.
> 
> Paul
> 
> 
> 
> 
> 
> 


-- 
#################################
Santosh Rao
Software Design Engineer,
HP, Cupertino.
email : santoshr@cup.hp.com
Phone : 408-447-3751
#################################