[Stgt-devel] [PATCH 4/4] OSD command parser
Mon Feb 26 01:27:48 CET 2007
From: Pete Wyckoff <pw at osc.edu>
Subject: Re: [Stgt-devel] [PATCH 4/4] OSD command parser
Date: Sat, 24 Feb 2007 19:20:52 -0500
> fujita.tomonori at lab.ntt.co.jp wrote on Fri, 23 Feb 2007 08:26 +0900:
> > From: Pete Wyckoff <pw at osc.edu>
> > Subject: Re: [Stgt-devel] [PATCH 4/4] OSD command parser
> > Date: Wed, 21 Feb 2007 13:30:58 -0500
> > > We'll still have to do something with bkio_submit_t to support
> > > bidirectional and to support returning both data and sense. I have
> > > such a thing for OSD but have not tried to make all the other
> > > devices fit it pending your perusal.
> > If you send these patches, I'll try to fix and merge them.
> > > The other largish piece I'm carrying around is AHS support for
> > > iscsi.
> > ditto.
> I'm still working on bidirectional support. It's held up as I'm
> having initiator issues, making it somewhat hard to test.
> But the extended CDB and AHS handling should be okay, as well as
> returning data and sense. Let me know if you'd like any changes
> to these next four.
Thanks! I had a quick look over them.
> 01 track-target-type
> Trivial, need this to switch on target type in scsi.c
I'll add SCSI_TARGET_PASSTHROUGH (or whatever we call it).
> 02 pass-scb-length
> Give length of CDB to target.
I'd like to move more members in iscsi_task structure to cmd structure
(cmd structure is renamed to scsi_cmd). I like something like struct
> 03 parse-iscsi-ahs-header-for-ext
> iscsi: parse AHS and handle extended CDB.
> 04 return-sense-data-in-iscsi-res
> Return sense data separate from requsted data. May want
> to fix tgtif to hand it back into kernel too sometime.
I thought about adding sense_buffer to scsi_cmd.
About in-kernel drivers, we need Mike's block layer patch merged into
mainline. That enables tgtd to pass non-aligned buffer to kernel so we
can put sense buffer to scsi_cmd.
I can take on these changes. Give me some time.
More information about the stgt