Re: [PATCH] clk: walk orphan list on clock provider registration

[Jerome Brunet <jbrunet@baylibre.com>]

On Mon 02 Dec 2019 at 19:19, Stephen Boyd <sboyd@kernel.org> wrote:

> Quoting Jerome Brunet (2019-11-29 08:16:58)
>> So far, we walked the orphan list every time a new clock was registered
>> in CCF. This was fine since the clocks were only referenced by name.
>> 
>> Now that the clock can be referenced through DT, it is not enough:
>> * Controller A register first a reference clocks from controller B
>>   through DT.
>> * Controller B register all its clocks then register the provider.
>> 
>> Each time controller B registers a new clock, the orphan list is walked
>> but it can't match since the provider is registered yet. When the
>> provider is finally registered, the orphan list is not walked unless
>> another clock is registered afterward.
>> 
>> This can lead to situation where some clocks remain orphaned even if
>> the parent is available.
>> 
>> Walking the orphan list on provider registration solves the problem.
>> 
>> Fixes: fc0c209c147f ("clk: Allow parents to be specified without string names")
>> Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
>> ---
>
> Sounds right. Thanks for making the fix!
>
> I suspect there should be a reported-by tag though?

laboriously, yes

>
>>  drivers/clk/clk.c | 59 +++++++++++++++++++++++++++++------------------
>>  1 file changed, 37 insertions(+), 22 deletions(-)
>> 
>> diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c
>> index ef4416721777..917ba37c3b9d 100644
>> --- a/drivers/clk/clk.c
>> +++ b/drivers/clk/clk.c
>> @@ -3249,6 +3249,34 @@ static inline void clk_debug_unregister(struct clk_core *core)
>>  }
>>  #endif
>>  
>> +static void __clk_core_reparent_orphan(void)
>
> Maybe drop the double underscore. clk_core prefix already means "private
> to this file".

I've done it this way just to keep coherent with __clk_core_init(),
which the code was extracted from

In the end, I don't mind so I'll drop it in the v2

>
>> +{
>> +       struct clk_core *orphan;
>> +       struct hlist_node *tmp2;
>> +
>> +       /*
>> +        * walk the list of orphan clocks and reparent any that newly finds a
>> +        * parent.
>> +        */
>> +       hlist_for_each_entry_safe(orphan, tmp2, &clk_orphan_list, child_node) {
>> +               struct clk_core *parent = __clk_init_parent(orphan);
>> +
>> +               /*
>> +                * We need to use __clk_set_parent_before() and _after() to
>> +                * to properly migrate any prepare/enable count of the orphan
>> +                * clock. This is important for CLK_IS_CRITICAL clocks, which
>> +                * are enabled during init but might not have a parent yet.
>> +                */
>> +               if (parent) {
>> +                       /* update the clk tree topology */
>> +                       __clk_set_parent_before(orphan, parent);
>> +                       __clk_set_parent_after(orphan, parent, NULL);
>> +                       __clk_recalc_accuracies(orphan);
>> +                       __clk_recalc_rates(orphan, 0);
>> +               }
>> +       }
>> +}
>> +
>>  /**
>>   * __clk_core_init - initialize the data structures in a struct clk_core
>>   * @core:      clk_core being initialized
>> @@ -3259,8 +3287,6 @@ static inline void clk_debug_unregister(struct clk_core *core)
>>  static int __clk_core_init(struct clk_core *core)
>>  {
>>         int ret;
>> -       struct clk_core *orphan;
>> -       struct hlist_node *tmp2;
>>         unsigned long rate;
>>  
>>         if (!core)
>> @@ -3416,27 +3442,8 @@ static int __clk_core_init(struct clk_core *core)
>>                 clk_enable_unlock(flags);
>>         }
>>  
>> -       /*
>> -        * walk the list of orphan clocks and reparent any that newly finds a
>> -        * parent.
>> -        */
>> -       hlist_for_each_entry_safe(orphan, tmp2, &clk_orphan_list, child_node) {
>> -               struct clk_core *parent = __clk_init_parent(orphan);
>> +       __clk_core_reparent_orphan();
>>  
>> -               /*
>> -                * We need to use __clk_set_parent_before() and _after() to
>> -                * to properly migrate any prepare/enable count of the orphan
>> -                * clock. This is important for CLK_IS_CRITICAL clocks, which
>> -                * are enabled during init but might not have a parent yet.
>> -                */
>> -               if (parent) {
>> -                       /* update the clk tree topology */
>> -                       __clk_set_parent_before(orphan, parent);
>> -                       __clk_set_parent_after(orphan, parent, NULL);
>> -                       __clk_recalc_accuracies(orphan);
>> -                       __clk_recalc_rates(orphan, 0);
>> -               }
>> -       }
>>  
>>         kref_init(&core->ref);
>>  out:
>> @@ -4288,6 +4295,10 @@ int of_clk_add_provider(struct device_node *np,
>>         mutex_unlock(&of_clk_mutex);
>>         pr_debug("Added clock from %pOF\n", np);
>>  
>> +       clk_prepare_lock();
>> +       __clk_core_reparent_orphan();
>> +       clk_prepare_unlock();
>> +
>
> Maybe make a locked version of this function and an unlocked version?
>
>>         ret = of_clk_set_defaults(np, true);
>>         if (ret < 0)
>>                 of_clk_del_provider(np);
>> @@ -4323,6 +4334,10 @@ int of_clk_add_hw_provider(struct device_node *np,
>>         mutex_unlock(&of_clk_mutex);
>>         pr_debug("Added clk_hw provider from %pOF\n", np);
>>  
>> +       clk_prepare_lock();
>> +       __clk_core_reparent_orphan();
>> +       clk_prepare_unlock();
>> +
>
> So we don't duplicate this twice.
>

Sure.

>>         ret = of_clk_set_defaults(np, true);
>>         if (ret < 0)