Re: [PATCH] ibmvscsis: Move to struct ibmvscsis_vdev usage in fabric configfs handlers

[FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>]

On Mon, 01 Nov 2010 15:25:46 -0700
"Nicholas A. Bellinger" <nab@linux-iscsi.org> wrote:

> > I think that vio_dev->unit_address can be used to identify a nexus
> > uniquely. It's not WWPN though. It's a connection between an initiator
> > lpar and a target lpar.
> 
> Ok, then I think we will want to generate an emulated SRP WWN in
> userspace based on partially based on system_id (via existing sysfs code
> I assume..?) and drop the ->unit_address check in ibmvscsis_make_tpg()
> and continue use the dev_name(vio_dev->dev) for the TPGT part of the WWN
> +TPGT. 

Why we want to generate fake WWNs?


>  This also means we need a TCM/ibmvscsis internal method to
> ensure that each ibmvscsis_vdev is only mapped 1:1 for ibmvscsis_tport
> +ibmvscsi_tpg target_core_fabric_configfs.c callers..

That's why my initiatal driver uses tpg for vdev. They are always
1:1. Having ibmvscsis_vdev (a separate data structure for vdev) makes
no point.


> >From there VIO SRP WWPNs present in /sys/kernel/config/target/ibmvscsis/
> explictly configured by enduser can be saved into 
> /etc/target/ibmvscsis_start.sh and made persisent across reboots.
> 
> Btw, I still need to add the SRP transportID encoding/decoding for
> target_core_fabric_lib.c and TCM_Loop code, so while I am doing this I
> will have a look at writing up some python code to generate these for
> our VIO SRP WWPNs.