Re: [PATCH] change scsi_cmd_ioctl to take a gendisk instead of a queue

[Patrick Mansfield <patmans@us.ibm.com>]

On Sun, Feb 01, 2004 at 09:33:34PM +0000, Christoph Hellwig wrote:
> On Sun, Feb 01, 2004 at 12:48:03PM -0800, Patrick Mansfield wrote:
> > Jens or others, any comments?
> > 
> > This patch against a recent bk 2.6 changes scsi_cmd_ioctl to take a
> > gendisk as an argument instead of a request_queue_t. This allows scsi char
> > devices to use the scsi_cmd_ioctl interface.
> 
> Heh, I had pretty much the same patch in 2.5.  I don't rember why
> I dropped it but you're looks at least as nice.

> > If the scsi upper level scsi_cmd_ioctl usage were consolidated in
> > scsi_prep_fn, we could pass a request_queue_t instead of a gendisk to
> > scsi_cmd_ioctl.
> 
> I don't parse that.  You mean taking REQ_PC handling from upper drivers
> to the scsi core?

Yes. That would also clean up the upper level driver code (for example,
sd.c's sd_init_command), but I have not tried this.

Basically for the REQ_BLOCK_PC, setup our own scsi_block_pc_intr in
scsi_lib.c, set up the scsi_cmnd, and set the scsi_cmnd done to
scsi_block_pc_intr. And maybe make still call scsi_init_io.

Then we don't need the:

	drv = *(struct scsi_driver **)req->rq_disk->private_data

And so, in theory ;-), we don't need rq_disk to be set. My reading of
current code shows none of the other scsi_cmd_ioctl users need rq_disk,
so we could then pass a q to scsi_cmd_ioctl. 

Then in the future we could use sg_io instead of the scsi_do_req,
scsi_wait_req, etc., and get rid of REQ_CMD in favor of REQ_BLOCK_PC.

> > +int bio_add_page(struct bio *bio, struct page *page, unsigned int len,
> > +		 unsigned int offset)
> > +{
> > +	return __bio_add_page(bdev_get_queue(bio->bi_bdev), bio, page,
> > +			      len, offset);
> > +}
> 
> Maybe worth inlining in a header? 

Maybe, and then export __bio_add_page.

-- Patrick Mansfield