Re: [Qemu-devel] Re: [PATCH] scsi-generic: bugfixes for 'SCSIRequest' conversion

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

On Thu, 2010-11-25 at 09:50 +0100, Gerd Hoffmann wrote:
> On 11/25/10 09:46, Benjamin Herrenschmidt wrote:
> 
> > So far tho, it appears that I can (at least with scsi-disk) rely on
> > always been eventually called with SCSI_REASON_DONE so my code (and
> > maybe the usb-msd code too, I haven't verified) relies on that to
> > complete requests... Is that incorrect ?
> 
> Yes.

Well, so far it works :-) But I suppose I must be lucky.. I must admit
that it's very unclear how that SCSI "stack" is meant to be used from
the HBA standpoint.

Right now, I've somewhat come up with:

  - client request occurs
  - call device send_command()
  - if result is 0, assume my complete() was called with 
    SCSI_REASON_DONE
  - else, use sign of result for transfer direction, store the
    absolute value as the total expected transfer len and call
    the device scsi_data_read()/write() and wait for complete()
  - when complete() is called:
    - if SCSI_REASON_DONE, complete client request
    - else perform the client "DMA" for "arg" bytes
    - call the device scsi_data_read()/write() again

So far it seems to work with scsi-disk but maybe I miss something in
which case I would very much enjoy being corrected :-)

Cheers,
Ben.