[PATCH V5 9/9] nvme: pci: support nested EH

[Ming Lei]

On Tue, May 15, 2018 at 06:02:13PM +0800, jianchao.wang wrote:
> Hi ming
> 
> On 05/11/2018 08:29 PM, Ming Lei wrote:
> > +static void nvme_eh_done(struct nvme_eh_work *eh_work, int result)
> > +{
> > +	struct nvme_dev *dev = eh_work->dev;
> > +	bool top_eh;
> > +
> > +	spin_lock(&dev->eh_lock);
> > +	top_eh = list_is_last(&eh_work->list, &dev->eh_head);
> > +	dev->nested_eh--;
> > +
> > +	/* Fail controller if the top EH can't recover it */
> > +	if (!result)
> > +		wake_up_all(&dev->eh_wq);
> > +	else if (top_eh) {
> > +		dev->ctrl_failed = true;
> > +		nvme_eh_sched_fail_ctrl(dev);
> > +		wake_up_all(&dev->eh_wq);
> > +	}
> > +
> > +	list_del(&eh_work->list);
> > +	spin_unlock(&dev->eh_lock);
> > +
> > +	dev_info(dev->ctrl.device, "EH %d: state %d, eh_done %d, top eh %d\n",
> > +			eh_work->seq, dev->ctrl.state, result, top_eh);
> > +	wait_event(dev->eh_wq, nvme_eh_reset_done(dev));
> 
> decrease the nested_eh before it exits, another new EH will have confusing seq number.
> please refer to following log:
> [ 1342.961869] nvme nvme0: Abort status: 0x0
> [ 1342.961878] nvme nvme0: Abort status: 0x0
> [ 1343.148341] nvme nvme0: EH 0: after shutdown, top eh: 1
> [ 1403.828484] nvme nvme0: I/O 21 QID 0 timeout, disable controller
> [ 1403.828603] nvme nvme0: EH 1: before shutdown
> ... waring logs are ignored here 
> [ 1403.984731] nvme nvme0: EH 0: state 4, eh_done -4, top eh 0  // EH0 go to wait
> [ 1403.984786] nvme nvme0: EH 1: after shutdown, top eh: 1
> [ 1464.856290] nvme nvme0: I/O 22 QID 0 timeout, disable controller  // timeout again in EH 1
> [ 1464.856411] nvme nvme0: EH 1: before shutdown // a new EH has a 1 seq number
> 
> Is it expected that the new EH has seq number 1 instead of 2 ?

Right, it has been fixed in my local tree of V5.1:

https://github.com/ming1/linux/commits/v4.17-rc-nvme-timeout.V5.1

And there are also several other fixes in this tree.

All will be merged to V6.

Thanks,
Ming