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

[Hannes Reinecke]

On 4/7/20 1:54 PM, John Garry wrote:
> On 06/04/2020 10:05, Hannes Reinecke wrote:
> > On 3/11/20 7:22 AM, Christoph Hellwig wrote:
> > > On Tue, Mar 10, 2020 at 09:08:56PM +0000, John Garry wrote:
> > > > On 10/03/2020 18:32, Christoph Hellwig wrote:
> > > > > On Wed, Mar 11, 2020 at 12:25:28AM +0800, John Garry wrote:
> > > > > > From: Hannes Reinecke <hare@xxxxxxxx>
> > > > > >
> > > > > > Allocate a separate 'reserved_cmd_q' for sending reserved commands.
> > > > >
> > > > > Why? Reserved command specifically are not in any way tied to queues.
> > > > > .
> > > >
> > > > So the v1 series used a combination of the sdev queue and the per-host
> > > > reserved_cmd_q. Back then you questioned using the sdev queue for 
> > > > virtio
> > > > scsi, and the unconfirmed conclusion was to use a common per-host q. 
> > > > This is
> > > > the best link I can find now:
> > > >
> > > > https://www.mail-archive.com/linux-scsi@xxxxxxxxxxxxxxx/msg83177.html
> > >
> > > That was just a question on why virtio uses the per-device tags, which
> > > didn't look like it made any sense. What I'm worried about here is
> > > mixing up the concept of reserved tags in the tagset, and queues to use
> > > them. Note that we already have the scsi_get_host_dev to allocate
> > > a scsi_device and thus a request_queue for the host itself. That seems
> > > like the better interface to use a tag for a host wide command vs
> > > introducing a parallel path.
> > Thinking about it some more, I don't think that scsi_get_host_dev() is
> > the best way of handling it.
> > Problem is that it'll create a new scsi_device with <hostno:this_id:0>,
> > which will then show up via eg 'lsscsi'.
>
> are you sure? Doesn't this function just allocate the sdev, but do 
> nothing with it, like probing it?
>
> I bludgeoned it in here for PoC:
>
> https://github.com/hisilicon/kernel-dev/commit/ef0ae8540811e32776f64a5b42bd76cbed17ba47 
>
>
> And then still:
>
> john@ubuntu:~$ lsscsi
> [0:0:0:0] disk SEAGATEÂ ST2000NM0045Â N004Â /dev/sda
> [0:0:1:0] disk SEAGATEÂ ST2000NM0045Â N004Â /dev/sdb
> [0:0:2:0] disk ATASAMSUNG HM320JIÂ 0_01Â /dev/sdc
> [0:0:3:0] disk SEAGATEÂ ST1000NM0023Â 0006Â /dev/sdd
> [0:0:4:0] enclosu HUAWEIExpander 12Gx16Â 128-
> john@ubuntu:~$
>
> Some proper plumbing would be needed, though.
>
> > This would be okay if 'this_id' would have been defined by the driver;
> > sadly, most drivers which are affected here do set 'this_id' to -1.
> > So we wouldn't have a nice target ID to allocate the device from, let
> > alone the problem that we would have to emulate a complete scsi device
> > with all required minimal command support etc.
> > And I'm not quite sure how well that would play with the exising SCSI
> > host template; the device we'll be allocating would have basically
> > nothing in common with the 'normal' SCSI devices.
> >
> > What we could do, though, is to try it the other way round:
> > Lift the request queue from scsi_get_host_dev() into the scsi host
> > itself, so that scsi_get_host_dev() can use that queue, but we also
> > would be able to use it without a SCSI device attached.
>
> wouldn't that limit 1x scsi device per host, not that I know if any more 
> would ever be required? But it does still seem better to use the request 
> queue in the scsi device.
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.

Cheers,

Hannes
--
Dr. Hannes Reinecke            Teamlead Storage & Networking
hare@xxxxxxx                               +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 NÃrnberg
HRB 36809 (AG NÃrnberg), GeschÃftsfÃhrer: Felix ImendÃrffer