Re: [PATCH RFC v2 02/24] scsi: allocate separate queue for reserved commands

[John Garry]

On 07/04/2020 17:30, Christoph Hellwig wrote:
> On Tue, Apr 07, 2020 at 04:00:10PM +0200, Hannes Reinecke wrote:
> > My concern is this:
> >
> > struct scsi_device *scsi_get_host_dev(struct Scsi_Host *shost)
> > {
> > 	[ .. ]
> > 	starget = scsi_alloc_target(&shost->shost_gendev, 0, shost->this_id);
> > 	[ .. ]
> >
> > and we have typically:
> >
> > drivers/scsi/hisi_sas/hisi_sas_v3_hw.c: .this_id                = -1,
> >
> > It's _very_ uncommon to have a negative number as the SCSI target device; in
> > fact, it _is_ an unsigned int already.
> >
> > But alright, I'll give it a go; let's see what I'll end up with.
>
> But this shouldn't be exposed anywhere.  And I prefer that over having
> magic requests/scsi_cmnd that do not have a valid ->device pointer.
> .

(just looking at this again)

Hi Christoph,

So how would this look added in scsi_lib.c:

struct scsi_cmnd *scsi_get_reserved_cmd(struct Scsi_Host *shost)
{
	struct scsi_cmnd *scmd;
	struct request *rq;
	struct scsi_device *sdev = scsi_get_host_dev(shost);

	if (!sdev)
		return NULL;

	rq = blk_mq_alloc_request(sdev->request_queue,
				  REQ_OP_DRV_OUT | REQ_NOWAIT,
				  BLK_MQ_REQ_RESERVED);
	if (IS_ERR(rq)) // fix tidy-up
		return NULL;
	WARN_ON(rq->tag == -1);
	scmd = blk_mq_rq_to_pdu(rq);
	scmd->request = rq;
	scmd->device = sdev;

	return scmd;
}
EXPORT_SYMBOL_GPL(scsi_get_reserved_cmd);

void scsi_put_reserved_cmd(struct scsi_cmnd *scmd)
{
	struct request *rq = blk_mq_rq_from_pdu(scmd);

	if (blk_mq_rq_is_reserved(rq)) {
		struct scsi_device *sdev = scmd->device;
		blk_mq_free_request(rq);
		scsi_free_host_dev(sdev);
	}
}
EXPORT_SYMBOL_GPL(scsi_put_reserved_cmd);

Not sure if we want a static scsi_device per host, or alloc and free 
dynamically.

(@Hannes, I also have some proper patches for libsas if you want to add it)

Cheers,
John