Re: [PATCH 1/2] [tgt]: Add proper CDB passthrough for SG_IO backstores

["Nicholas A. Bellinger" <nab@linux-iscsi.org>]

On Sun, 2010-05-30 at 12:36 +0300, Boaz Harrosh wrote:
> On 05/18/2010 08:13 AM, Nicholas A. Bellinger wrote:
> > From: Nicholas Bellinger <nab@linux-iscsi.org>
> > 
> > This patch updates usr/bs_sg.c:bs_sg_cmd_submit() to process incoming scsi-generic IO
> > of type SG_DXFER_TO_DEV and SG_DXFER_FROM_DEV using a single new struct scsi_cmnd->sg_iovec
> > set using existing scsi_cmd.h scsi_get_[out,in]_[buffer,length]() macros.
> > 
> > It also makes set_cmd_failed() properly return SAM_STAT_CHECK_CONDITION, which is expected
> > by the bs_sg_cmd_submit() caller in usr/scsi.c:scsi_cmd_perform()
> > 
> > Finally, this patch adds struct backingstore_template->bs_passthrough=1 that is used by
> > bs_sg.c here and the next patch to hand off struct scsi_cmd for CDB passthrough.
> > 
> > Signed-off-by: Nicholas A. Bellinger <nab@linux-iscsi.org>
> > ---
> >  usr/bs_sg.c     |   31 ++++++++++++++++++++++++-------
> 
> Why not add a new bs_bsg.c backing store that goes through BSG and not SG
> and do it properly for all type of targets and commands?
> 
> Just a thought.

Greetings Boaz,

I completely agree with you here and think that BSG is the next logical
direction for expanding our target capabilities and for adding some
manner of I_T nexus auto-configuration between STGT userspace fabric
modules <-> TCM_Loop multi-fabric target port LUNs.  That said, I will
start groking BSG and look at adding a new bs_bsg backstore for STGT
soon for extended user -> kernel passthrough ops soon.

Also, the BSG direction is going to make alot of sense with the QEMU-KVM
HBA Emulation <-> TCM_Loop efforts too.   However I believe at least for
megraid_sas that TMRs are handled in the driver for LDs and not
processed directly by the 8708EM2 PowerPC Core firmware. ;)

Best,

--nab