attached integrate-iscsi-netlink.patch - incorporate the 
open-iscsi/linux-iscsi netlink interface into the iscsi transport class.

This patch also removes the iscsi_host class and breaks the 
iscsi_transport class into a iscsi_connection and iscsi_session class. 
The iscsi_host class is removed becuase the open-iscsi/linux-iscsi 
netlink framework pushes login to usersapce so there is no need to store 
long strings like the initiatorname and alias in the kernel anymore. And 
breaking out the session and connection from the old 
/sys/class/iscsi_transport was necessary becuase the session and 
connection creation and destruction is driven from userspace and it may 
end up where you have a session but no scsi target (in which case 
/sys/class/iscsi_transport would not show up previously). Having the 
session/connection represented in this case is useful for iscsi async 
events and their userpace handlers.

We followed the FC transport class strategy for the sysfs layout which 
uses struct devices and driver model transport_classes so the format 
looks like the following:

[root@mina class]# cd iscsi_session/
[root@mina iscsi_session]# ls
session2
[root@mina iscsi_session]# cd session2/
[root@mina session2]# ls
data_pdu_in_order  device  first_burst_len  initial_r2t 
max_outstanding_r2t
data_seq_in_order  erl     immediate_data   max_burst_len
[root@mina class]# cd iscsi_connection/
[root@mina iscsi_connection]# ls
connection2:0
[root@mina iscsi_connection]# cd connection2\:0/
[root@mina connection2:0]# ls
device  header_digest  ifmarker  max_recv_dlength  max_xmit_dlength 
ofmarker

A session's connection is linked by the device links:

[root@mina host2]# tree
.
|-- detach_state
|-- power
|   `-- state
`-- session2
     |-- connection2:0
     |   |-- detach_state
     |   `-- power
     |       `-- state
     |-- detach_state
     `-- power
         `-- state

We are perfectly fine with adding a different parent of the session or 
making the class_devices linked rather than the devices or ditching the 
device usage for kobjects or something else. We were not sure if the FC 
transport class model was the preferred model for us to follow though.


Thanks,

Linux-iscsi Team

Signed-off-by: Alex Aizman <itn780@yahoo.com>
Signed-off-by: Dmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>