Re: [PATCH v10 2/6] PM / Domains: core changes for multiple states

[Axel Haslam <ahaslam@baylibre.com>]

Hi Zhaoyang,

>     @@ -576,6 +584,8 @@ static void pm_genpd_sync_poweroff(struct
>     generic_pm_domain *genpd,
>                  || atomic_read(&genpd->sd_count) > 0)
>                      return;
>
>     +       /* Choose the deepest state when suspending */
>     +       genpd->state_idx = genpd->state_count - 1;
>
> if there is no governor configured, does it mean that we just try the
> deepest one here?
>

If a governor is not configured, there is no way to choose between the 
states, and the state_idx will not change. so it will remain the one set 
at the genpd init function wich is the same as here: the deepest state. 
(state_count -1). In other words, if there is no governor, there
is only one state: the deepest.

About setting the deepest state in this particular place, My 
understanding is that this code will be called in suspend path,
and in this case genpd_poweroff is not called, hence the genpd governor 
is not called either. Because there are many other things happening 
during suspend/resume, i think that suspend/resume is a slow procedure 
that will most likely not honor the device qos constraints, so i set the 
genpd to its deepest state.

Regards
Axel
>              genpd_power_off(genpd, timed);
>
>              genpd->status = GPD_STATE_POWER_OFF;