[dm-devel] Re: [PATCH RFC 0/4] use scatter lists for all block pc requests and simplify hw handlers

[James Bottomley]

On Tue, 2005-06-07 at 13:10 +0100, Christoph Hellwig wrote:
> shouldn't blk_rq_map_kern handle a 0 buffer and do nothing more than
> blk_get_request?  It's not exactly a criticial fastpath and that would make life
> easier for the callers.

Yes ... and it should probably do bio bouncing as well, since there's
nothing special we have to do on completion to clean it up.

I also think we might need a blk_rq_kern_iovec call that would take a
vector of user I/O's and map it to a multiple bio request.  This would
allow me to strip all of the user mapping out of sg, and also allow the
block layer SG_IO to pick up this API (currently it returns -ENOTSUPP if
you try it).

> > +		if (req->rq_disk) {
> > +			drv = *(struct scsi_driver **)req->rq_disk->private_data;
> > +			if (unlikely(!drv->init_command(cmd))) {
> > +				scsi_release_buffers(cmd);
> > +				scsi_put_command(cmd);
> > +				return BLKPREP_KILL;
> > +			}
> > +		} else {
> > +			memcpy(cmd->cmnd, req->cmd, sizeof(cmd->cmnd));
> > +			if (rq_data_dir(req) == WRITE)
> > +				cmd->sc_data_direction = DMA_TO_DEVICE;
> > +			else if (req->data_len)
> > +				cmd->sc_data_direction = DMA_FROM_DEVICE;
> > +			else
> > +				cmd->sc_data_direction = DMA_NONE;
> > +			
> > +			cmd->transfersize = req->data_len;
> > +			cmd->allowed = 3;
> > +			cmd->timeout_per_command = req->timeout;
> 
> most of this could probably be done in the midlayer always instead of the
> upper drivers.

I don't think so ... this is all struct scsi_cmnd handling.  SCSI needs
to allocate and process its internal command structures.

However, I think we have a case for the number of retries being carried
by the generic request structure.

James