On 4/13/07, Douglas Gilbert <dougg at torque.net> wrote: > Mark Harvey wrote: > > SSC / SMC patches to SCSI Tgt code. > > > > It is based from git commit : > > ================================ > > commit 47e15c234ca298d31b1709db592eef9db4bce0ca > > Author: Albert Pauw <albert.pauw at gmail.com> > > Date: Sat Apr 7 17:31:54 2007 +0900 > > > > some shortcut options missing from tgtadm > > > > Signed-off-by: Albert Pauw <albert.pauw at gmail.com> > > ================================ > > > > > > Additions from last patch include : > > - Updated syntax to add/configure limited set of SMC / SSC parameters > > - Start implementing structures for MODE SELECT / MODE SENSE op codes > > - Start implementing structures for LOG SELECT / LOG SENSE op codes > > It will be interesting to see how you go with MODE SELECT > and LOG SELECT. As far as I can see it requires the SCSI > parser to: > a) decode the "parameter list length" field in the > cdb > b) go away and fetch that much data from the data_out > c) return to parser and process the parameter list > together with other fields in the cdb > > In the case of the SAS and FCP transports, step b) > probably means going all the way back to the > initiator (low level details: the target sends a XFER_RDY > frame to entice the initiator to send the data_out buffer). > There may be less work involved with iSCSI and SRP. > > Folks who want persistent reservations will also want > this problem solved as PERSISTENT RESERVE OUT has the > same structure as MODE SELECT. > > Doug Gilbert > I'll let you know how I go.. It will either be tomorrow, or a little over a week. (Taking a weeks break from computers, and taking the family on holidays :) I managed to get this working without too much problems with the Intel iSCSI reference code.. So the iSCSI initiator is doing the 'right thing' Cheers Mark |