Re: [PATCH 2/2] power: bq24190_charger: Use PM runtime autosuspend

[Liam Breck <liam@networkimprov.net>]

On Mon, Jan 30, 2017 at 3:55 PM, Tony Lindgren <tony@atomide.com> wrote:
> * Liam Breck <liam@networkimprov.net> [170126 15:57]:
>> On Tue, Jan 24, 2017 at 10:29 AM, Tony Lindgren <tony@atomide.com> wrote:
>> > @@ -1430,10 +1473,20 @@ static int bq24190_probe(struct i2c_client *client,
>> >  static int bq24190_remove(struct i2c_client *client)
>> >  {
>> >         struct bq24190_dev_info *bdi = i2c_get_clientdata(client);
>> > +       int error;
>> > +
>> > +       error = pm_runtime_get_sync(bdi->dev);
>> > +       if (error < 0) {
>> > +               dev_warn(bdi->dev, "pm_runtime_get failed: %i\n", error);
>> > +               pm_runtime_put_noidle(bdi->dev);
>> > +       }
>> >
>> > -       pm_runtime_get_sync(bdi->dev);
>> >         bq24190_register_reset(bdi);
>> > -       pm_runtime_put_sync(bdi->dev);
>> > +
>> > +       if (!error) {
>> > +               pm_runtime_dont_use_autosuspend(bdi->dev);
>> > +               pm_runtime_put_sync(bdi->dev);
>> > +       }
>>
>> Should put_sync() be called before dont_use_autosuspend(), as in probe()?
>
> Yeah good to have them paired. With pm_runtime_put_noidle() we're
> active on error, so we can just call pm_runtime_dont_use_autosuspend()
> and pm_runtime_put_sync() unconditionally.

I meant should the two pm_runtime_* calls above be flipped, so
put_sync() is first. That's the order in probe()