[PATCHv4 3/4] nvme: Introduce frozen controller state

[keith.busch@linux.intel.com (Keith Busch)]

On Mon, Jul 16, 2018@07:09:04PM +0800, Ming Lei wrote:
> On Fri, Jul 13, 2018@02:56:08PM -0600, Keith Busch wrote:
> > +	if (ctrl->state == NVME_CTRL_FROZEN) {
> > +		nvme_wait_freeze(ctrl);
> > +		blk_mq_update_nr_hw_queues(ctrl->tagset, ctrl->queue_count - 1);
> > +		nvme_unfreeze(ctrl);
> > +		if (!nvme_change_ctrl_state(ctrl, NVME_CTRL_LIVE))
> > +			return;
> > +	}
> > +
> 
> Now you move unfreezing into start work function, seems this way can't
> guarantee that nvme_unfreeze() is strictly paired with nvme_start_freeze() done
> in nvme_dev_disable().

The state machine is supposed to guarantee this. We only start freeze
in certain states from nvme_dev_disable, and transition to frozen when
exiting those states. I agree it's not the easiest API to follow, and
I'll double check to see if it's correct. I think I have the
reset/connecting/frozen/live transitions okay for nvme-pci, but I need
to look again at the deleting/dead states.