Re: [PATCH v1 2/2] PM: runtime: Defer suspending suppliers

[Rafael J. Wysocki]

On Fri, Mar 19, 2021 at 2:30 PM Ulf Hansson <ulf.hansson@xxxxxxxxxx> wrote:
>
> On Thu, 18 Mar 2021 at 19:15, Rafael J. Wysocki <rjw@xxxxxxxxxxxxx> wrote:
> >
> > From: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
> >
> > Because the PM-runtime status of the device is not updated in
> > __rpm_callback(), attempts to suspend the suppliers of the given
> > device triggered by the rpm_put_suppliers() call in there may fail.
> >
> > To fix this (1) modify __rpm_callback() to avoid attempting to
> > actually suspend the suppliers, but only decrease their PM-runtime
> > usage counters and (2) make rpm_suspend() try to suspend the suppliers
> > after changing the device's PM-runtime status, in analogy with the
> > handling of the device's parent.
> >
> > Link: https://lore.kernel.org/linux-pm/CAPDyKFqm06KDw_p8WXsM4dijDbho4bb6T4k50UqqvR1_COsp8g@xxxxxxxxxxxxxx/
> > Fixes: 21d5c57b3726 ("PM / runtime: Use device links")
> > Reported-by: elaine.zhang <zhangqing@xxxxxxxxxxxxxx>
> > Diagnosed-by: Ulf Hansson <ulf.hansson@xxxxxxxxxx>
> > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
>
> Just a minor nitpick, see below. In any case:
>
> Reviewed-by: Ulf Hansson <ulf.hansson@xxxxxxxxxx>

Thanks!

>
> > ---
> >  drivers/base/power/runtime.c |   45 +++++++++++++++++++++++++++++++++++++------
> >  1 file changed, 39 insertions(+), 6 deletions(-)
> >
> > Index: linux-pm/drivers/base/power/runtime.c
> > ===================================================================
> > --- linux-pm.orig/drivers/base/power/runtime.c
> > +++ linux-pm/drivers/base/power/runtime.c
> > @@ -305,7 +305,7 @@ static int rpm_get_suppliers(struct devi
> >         return 0;
> >  }
> >
> > -static void rpm_put_suppliers(struct device *dev)
> > +static void __rpm_put_suppliers(struct device *dev, bool try_to_suspend)
> >  {
> >         struct device_link *link;
> >
> > @@ -313,10 +313,30 @@ static void rpm_put_suppliers(struct dev
> >                                 device_links_read_lock_held()) {
> >
> >                 while (refcount_dec_not_one(&link->rpm_active))
> > -                       pm_runtime_put(link->supplier);
> > +                       pm_runtime_put_noidle(link->supplier);
> > +
> > +               if (try_to_suspend)
> > +                       pm_request_idle(link->supplier);
> >         }
> >  }
> >
> > +static void rpm_put_suppliers(struct device *dev)
> > +{
> > +       __rpm_put_suppliers(dev, true);
> > +}
> > +
> > +static void rpm_try_to_suspend_suppliers(struct device *dev)
>
> Maybe "rpm_suspend_suppliers" is sufficient for the name of the
> function, but I have no strong opinion.

OK

In addition to this, spin_unlock_irq()/spin_lock_irq() need to be used
around the call to it in rpm_suspend(), so I'll send a v2.  I guess
that the R-by still applies, though. :-)

> > +{
> > +       struct device_link *link;
> > +       int idx = device_links_read_lock();
> > +
> > +       list_for_each_entry_rcu(link, &dev->links.suppliers, c_node,
> > +                               device_links_read_lock_held())
> > +               pm_request_idle(link->supplier);
> > +
> > +       device_links_read_unlock(idx);
> > +}
> > +