Re: [PATCH] nvme-pci: Use non-operational power state instead of D3 on Suspend-to-Idle

[Rafael J. Wysocki]

On Thu, May 9, 2019 at 11:21 PM Keith Busch <kbusch@xxxxxxxxxx> wrote:
>
> On Thu, May 09, 2019 at 10:54:04PM +0200, Rafael J. Wysocki wrote:
> > On Thu, May 9, 2019 at 9:33 PM Keith Busch <kbusch@xxxxxxxxxx> wrote:
> > >  #include <linux/io-64-nonatomic-lo-hi.h>
> > > @@ -2851,6 +2852,8 @@ static int nvme_suspend(struct device *dev)
> > >         struct pci_dev *pdev = to_pci_dev(dev);
> > >         struct nvme_dev *ndev = pci_get_drvdata(pdev);
> > >
> > > +       if (!pm_suspend_via_firmware())
> > > +               return nvme_set_power(&ndev->ctrl, ndev->ctrl.npss);
> >
> > You probably want to call pci_save_state(pdev) in the branch above to
> > prevent pci_pm_suspend_noirq() from calling pci_prepare_to_sleep()
> > going forward, so I would write this routine as
> >
> > if (pm_suspend_via_firmware()) {
> >         nvme_dev_disable(ndev, true);
> >         return 0;
> > }
> >
> > pci_save_state(pdev)
> > return nvme_set_power(&ndev->ctrl, ndev->ctrl.npss);
>
> Ah, good point. I'll make sure that's added and will wait to see hear if
> there's any other feedback.
>
> I am trying to test the paths by faking out PS capabilities, and have
> a question on how to force each:
>
> Running "rtcwake -m freeze ...", that takes the !pm_suspend_via_firmware()
> path as I expected.
>
> But trying to test the original path, I thought using "-m mem" would
> have been a suspend via firmware, but that is still returning false.
>
> Is that expected?

Yes, if s2idle is the default on that platform.  You should be able to
switch over to S3 by writing "deep" into /sys/power/mem_sleep as long
as it is supported on that platform at all.

> I've only tried this on one platform so far, so might
> just be this particular one is missing a firmware capability.

You can check that by looking into /sys/power/mem_sleep (if there is
only "[s2idle]" in there, S3 is not supported).