Menu
▾
▴
Re: [Iscsitarget-devel] Fwd: Newbie to this group
|
From: Ming Z. <bla...@gm...> - 2008-07-11 01:47:54
|
On Thu, 2008-07-10 at 18:42 -0700, john M wrote:
> Hi all,
>
> I intergrated the iet target with our Raid core. As of now, whatever
> data i need to send to Raid Core or receive from Raid core, i am doing
> byte copy from "tio" or to "tio" structure. My RaidCore return
> SGList for Read commands and would give some buffer pointer to copy
> the DMA data into for the write commands.
>
> To completely utilize the throughput, i need a interface which has
> Scatter Gatter List instead of "tio". Can anyone put some light on
> how would i go to do this. Or point me at the place which has this
> interface already implemented.
did not see your code, but you can always allocate a sg list that point
to the buffer you get from tio, and pass that to raid core. so there is
not copy any more.
>
>
> -JM
>
>
> On Mon, Jun 30, 2008 at 10:23 AM, john M <joh...@gm...> wrote:
> Ming/Arne,
> Thanks for your help, i figured out what was going wrong.
> -JM
>
>
>
> On Mon, Jun 30, 2008 at 7:01 AM, Ming Zhang
> <bla...@gm...> wrote:
> i could not spot anything wrong easily base on this
> partial view. since
> u get the BUG(), then u can print out that command
> content and see which
> command it is. how it was executed....
>
>
>
> On Sun, 2008-06-29 at 11:23 -0700, john M wrote:
> > Arne,
> > Here is my code snippet:
> >
> > static int cmnd_execute(struct iscsi_cmnd *cmnd)
> > {
> > int type =
> cmnd->conn->session->target->trgt_param.target_type;
> > assert(target_type_array[type]->execute_cmnd);
> > return
> > target_type_array[type]->execute_cmnd(cmnd);
> /* This
> > would call ProcessCommand which is a filter driver
> to raid core*/
> > }
> >
> > int ProcessCmd(cmnd)
> > {
> > int err ;
> > err = SendCmndToRaidCore(cmnd) ;
> > return err;
> > }
> >
> > SentRspToInitator(cmnd, SGL *RspSGL)
> > {
> > int tmplen = 0, rsplen = 0 ;
> > /* Caluclate the SCSI Respone length */
> > for(i = 0; i < num_sge; i++){
> > tmplen += RspSGL->sge[i].length ;
> /* Itertate thr
> > all the SGE and add the lengths
> >
> >
> > rsplen = tmplen ;
> >
> > tio = cmnd->tio = alloc(rsplen) ;
> /*
> > Allocate the memory for response length */
> > data =
> > page_address(tio->pvec[0]);
> /*This is the original
> code */
> >
> > for(i = 0; i < num_sge; i++)
> > {
> /*
> > Copy the response into the tio buffer */
> > memcpy(data,
> bus_to_virt(sgl->sge[i].address),
> > sgl->sge[i].length) ;
> > data += (sgl->sge[i].length-1) ;
> > }
> >
> > tio_set(tio, tmplen, 0) ;
> >
> > do_send_data_rsp(cmnd);
> >
> > return 0;
> > }
> >
> >
> > -JM
> >
> >
> > On Sun, Jun 29, 2008 at 3:48 AM, Arne Redlich
> <ag...@po...>
> > wrote:
> > Am Samstag, den 28.06.2008, 18:40 -0700
> schrieb john M:
> > >
> > > Hi Team,
> > >
> > > I am trying to interface the IET iscsi
> target with our Raid
> > core.
> > >
> > > I hacked the disk_execute_cmnd to send the
> SCSI CDB's to
> > RaidCore
> > > insted of File-IO/Block-IO. When my
> RaidCore comes with the
> > > Data/Response for the particular SCSI
> command, i fill in
> > the
> > > Data/Response into the tio->pvec[idx] part
> of the struct
> > iscsi_cmnd
> > > and call the "do_send_data_rsp(cmnd)"
> function to send the
> > data back
> > > to Initiator.
> > >
> > > The Report LUNS command goes well, but the
> next command
> > INQUIRY
> > > Carshes at the following place as shown
> below. i wrote the
> > code
> > > similar to the existing interface. Looks
> like i need to
> > delete the
> > > "struct iscsi_cmnd" from the linked
> list. But the existing
> > code
> > > doesn't call list_del(cmnd->list) or
> > list_del_init(&cmnd->list); in
> > > the disk_execute_cmnd function. Can
> anyone point out where
> > am i going
> > > wrong?
> >
> >
> > You'll have to show us (your modifications
> to) the code.
> >
> > Cheers,
> > Arne
> >
> >
>
> --
>
> Ming Zhang
>
>
> @#$%^ purging memory... (*!%
> http://blackmagic02881.wordpress.com/
> http://www.linkedin.com/in/blackmagic02881
> --------------------------------------------
>
>
>
>
>
--
Ming Zhang
@#$%^ purging memory... (*!%
http://blackmagic02881.wordpress.com/
http://www.linkedin.com/in/blackmagic02881
--------------------------------------------
|