Re: [PATCH 4/5] memstick: rtsx_usb_ms: Support runtime power management

[Kai-Heng Feng]

at 21:29, Ulf Hansson <ulf.hansson@xxxxxxxxxx> wrote:

> [...]
>
> > -#ifdef CONFIG_PM_SLEEP
> > -static int rtsx_usb_ms_suspend(struct device *dev)
> > +#ifdef CONFIG_PM
> > +static int rtsx_usb_ms_runtime_suspend(struct device *dev)
> >  {
> >         struct rtsx_usb_ms *host = dev_get_drvdata(dev);
> >         struct memstick_host *msh = host->msh;
> >
> > -       dev_dbg(ms_dev(host), "--> %s\n", __func__);
> > -
> > +       host->suspend = true;
> >         memstick_suspend_host(msh);
>
> I missed this one. Does this really work? To me, this looks like doing
> things upside-down.
>
> To suspend the host, you first need to runtime resume it, because
> mmc_suspend_host() calls into one of the host ops and my touch the
> device, right?
>
> If you want to suspend the host (actually the naming is wrong, as it's
> about suspending/power-iff the memstick card), that should be done via
> when the memstick core finds that the card is removed or during system
> wide suspend.

Do you mean the logic was wrong even before my modification?

Kai-Heng

> > +
> >         return 0;
> >  }
> >
> > -static int rtsx_usb_ms_resume(struct device *dev)
> > +static int rtsx_usb_ms_runtime_resume(struct device *dev)
> >  {
> >         struct rtsx_usb_ms *host = dev_get_drvdata(dev);
> >         struct memstick_host *msh = host->msh;
> >
> > -       dev_dbg(ms_dev(host), "--> %s\n", __func__);
> > -
> >         memstick_resume_host(msh);
>
> According to the above, this seems not correct to me.
>
> > +       host->suspend = false;
> > +       schedule_delayed_work(&host->poll_card, 100);
> > +
> >         return 0;
> >  }
> > -#endif /* CONFIG_PM_SLEEP */
>
> [...]
>
> Kind regards
> Uffe