Now, the userspace governor does not support userspace,
if the driver only use the fast-switch and not add target_index(),
it will cause uerspace not work.
So add fast-switch support for userspace governor.

Co-developed-by: Guohua Yan <guohua.yan@unisoc.com>
Signed-off-by: Guohua Yan <guohua.yan@unisoc.com>
Signed-off-by: Xuewen Yan <xuewen.yan@unisoc.com>
---
 drivers/cpufreq/cpufreq_userspace.c | 35 +++++++++++++++++++++++++----
 1 file changed, 31 insertions(+), 4 deletions(-)

diff --git a/drivers/cpufreq/cpufreq_userspace.c b/drivers/cpufreq/cpufreq_userspace.c
index 2c42fee76daa..3a99197246ed 100644
--- a/drivers/cpufreq/cpufreq_userspace.c
+++ b/drivers/cpufreq/cpufreq_userspace.c
@@ -21,6 +21,30 @@ struct userspace_policy {
 	struct mutex mutex;
 };
 
+static int cpufreq_userspace_target_freq(struct cpufreq_policy *policy,
+			unsigned int target_freq, unsigned int relation)
+{
+	int ret;
+
+	if (policy->fast_switch_enabled) {
+		unsigned int idx;
+
+		target_freq = clamp_val(target_freq, policy->min, policy->max);
+
+		if (!policy->freq_table)
+			return target_freq;
+
+		idx = cpufreq_frequency_table_target(policy, target_freq, relation);
+		policy->cached_resolved_idx = idx;
+		policy->cached_target_freq = target_freq;
+		ret = !cpufreq_driver_fast_switch(policy, policy->freq_table[idx].frequency);
+	} else {
+		ret = __cpufreq_driver_target(policy, target_freq, relation);
+	}
+
+	return ret;
+}
+
 /**
  * cpufreq_set - set the CPU frequency
  * @policy: pointer to policy struct where freq is being set
@@ -41,7 +65,7 @@ static int cpufreq_set(struct cpufreq_policy *policy, unsigned int freq)
 
 	userspace->setspeed = freq;
 
-	ret = __cpufreq_driver_target(policy, freq, CPUFREQ_RELATION_L);
+	ret = cpufreq_userspace_target_freq(policy, freq, CPUFREQ_RELATION_L);
  err:
 	mutex_unlock(&userspace->mutex);
 	return ret;
@@ -62,6 +86,8 @@ static int cpufreq_userspace_policy_init(struct cpufreq_policy *policy)
 
 	mutex_init(&userspace->mutex);
 
+	cpufreq_enable_fast_switch(policy);
+
 	policy->governor_data = userspace;
 	return 0;
 }
@@ -72,6 +98,7 @@ static int cpufreq_userspace_policy_init(struct cpufreq_policy *policy)
  */
 static void cpufreq_userspace_policy_exit(struct cpufreq_policy *policy)
 {
+	cpufreq_disable_fast_switch(policy);
 	kfree(policy->governor_data);
 	policy->governor_data = NULL;
 }
@@ -112,13 +139,13 @@ static void cpufreq_userspace_policy_limits(struct cpufreq_policy *policy)
 		 policy->cpu, policy->min, policy->max, policy->cur, userspace->setspeed);
 
 	if (policy->max < userspace->setspeed)
-		__cpufreq_driver_target(policy, policy->max,
+		cpufreq_userspace_target_freq(policy, policy->max,
 					CPUFREQ_RELATION_H);
 	else if (policy->min > userspace->setspeed)
-		__cpufreq_driver_target(policy, policy->min,
+		cpufreq_userspace_target_freq(policy, policy->min,
 					CPUFREQ_RELATION_L);
 	else
-		__cpufreq_driver_target(policy, userspace->setspeed,
+		cpufreq_userspace_target_freq(policy, userspace->setspeed,
 					CPUFREQ_RELATION_L);
 
 	mutex_unlock(&userspace->mutex);
-- 
2.25.1

On 12/9/24 08:14, Xuewen Yan wrote:
> Now, the userspace governor does not support userspace,
> if the driver only use the fast-switch and not add target_index(),

Which driver does that? Is that actually valid?
No mainline driver from what I can see.

> it will cause uerspace not work.

s/uerspace/userspace
to not work?

> So add fast-switch support for userspace governor.
> 
> Co-developed-by: Guohua Yan <guohua.yan@unisoc.com>
> Signed-off-by: Guohua Yan <guohua.yan@unisoc.com>
> Signed-off-by: Xuewen Yan <xuewen.yan@unisoc.com>
> ---
>  drivers/cpufreq/cpufreq_userspace.c | 35 +++++++++++++++++++++++++----
>  1 file changed, 31 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/cpufreq/cpufreq_userspace.c b/drivers/cpufreq/cpufreq_userspace.c
> index 2c42fee76daa..3a99197246ed 100644
> --- a/drivers/cpufreq/cpufreq_userspace.c
> +++ b/drivers/cpufreq/cpufreq_userspace.c
> @@ -21,6 +21,30 @@ struct userspace_policy {
>  	struct mutex mutex;
>  };
>  
> +static int cpufreq_userspace_target_freq(struct cpufreq_policy *policy,
> +			unsigned int target_freq, unsigned int relation)
> +{
> +	int ret;

not really necessary

> +
> +	if (policy->fast_switch_enabled) {
> +		unsigned int idx;
> +
> +		target_freq = clamp_val(target_freq, policy->min, policy->max);
> +
> +		if (!policy->freq_table)
> +			return target_freq;
> +
> +		idx = cpufreq_frequency_table_target(policy, target_freq, relation);
> +		policy->cached_resolved_idx = idx;
> +		policy->cached_target_freq = target_freq;
> +		ret = !cpufreq_driver_fast_switch(policy, policy->freq_table[idx].frequency);
> +	} else {
> +		ret = __cpufreq_driver_target(policy, target_freq, relation);

NIT: could save the indent if you reverse conditions and ret early on !fast_switch

> +	}
> +
> +	return ret;
> +}
> +
>  /**
>   * cpufreq_set - set the CPU frequency
>   * @policy: pointer to policy struct where freq is being set
> @@ -41,7 +65,7 @@ static int cpufreq_set(struct cpufreq_policy *policy, unsigned int freq)
>  
>  	userspace->setspeed = freq;
>  
> -	ret = __cpufreq_driver_target(policy, freq, CPUFREQ_RELATION_L);
> +	ret = cpufreq_userspace_target_freq(policy, freq, CPUFREQ_RELATION_L);
>   err:
>  	mutex_unlock(&userspace->mutex);
>  	return ret;
> @@ -62,6 +86,8 @@ static int cpufreq_userspace_policy_init(struct cpufreq_policy *policy)
>  
>  	mutex_init(&userspace->mutex);
>  
> +	cpufreq_enable_fast_switch(policy);
> +
>  	policy->governor_data = userspace;
>  	return 0;
>  }
> @@ -72,6 +98,7 @@ static int cpufreq_userspace_policy_init(struct cpufreq_policy *policy)
>   */
>  static void cpufreq_userspace_policy_exit(struct cpufreq_policy *policy)
>  {
> +	cpufreq_disable_fast_switch(policy);
>  	kfree(policy->governor_data);
>  	policy->governor_data = NULL;
>  }
> @@ -112,13 +139,13 @@ static void cpufreq_userspace_policy_limits(struct cpufreq_policy *policy)
>  		 policy->cpu, policy->min, policy->max, policy->cur, userspace->setspeed);
>  
>  	if (policy->max < userspace->setspeed)
> -		__cpufreq_driver_target(policy, policy->max,
> +		cpufreq_userspace_target_freq(policy, policy->max,
>  					CPUFREQ_RELATION_H);
>  	else if (policy->min > userspace->setspeed)
> -		__cpufreq_driver_target(policy, policy->min,
> +		cpufreq_userspace_target_freq(policy, policy->min,
>  					CPUFREQ_RELATION_L);
>  	else
> -		__cpufreq_driver_target(policy, userspace->setspeed,
> +		cpufreq_userspace_target_freq(policy, userspace->setspeed,
>  					CPUFREQ_RELATION_L);
>  
>  	mutex_unlock(&userspace->mutex);

On Mon, Dec 9, 2024 at 6:36 PM Christian Loehle
<christian.loehle@arm.com> wrote:
>
> On 12/9/24 08:14, Xuewen Yan wrote:
> > Now, the userspace governor does not support userspace,
> > if the driver only use the fast-switch and not add target_index(),
>
> Which driver does that? Is that actually valid?
> No mainline driver from what I can see.
>

Yes, indeed no mainline driver, It's on our own driver.

> > it will cause uerspace not work.
>
> s/uerspace/userspace
> to not work?
>
> > So add fast-switch support for userspace governor.
> >
> > Co-developed-by: Guohua Yan <guohua.yan@unisoc.com>
> > Signed-off-by: Guohua Yan <guohua.yan@unisoc.com>
> > Signed-off-by: Xuewen Yan <xuewen.yan@unisoc.com>
> > ---
> >  drivers/cpufreq/cpufreq_userspace.c | 35 +++++++++++++++++++++++++----
> >  1 file changed, 31 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/cpufreq/cpufreq_userspace.c b/drivers/cpufreq/cpufreq_userspace.c
> > index 2c42fee76daa..3a99197246ed 100644
> > --- a/drivers/cpufreq/cpufreq_userspace.c
> > +++ b/drivers/cpufreq/cpufreq_userspace.c
> > @@ -21,6 +21,30 @@ struct userspace_policy {
> >       struct mutex mutex;
> >  };
> >
> > +static int cpufreq_userspace_target_freq(struct cpufreq_policy *policy,
> > +                     unsigned int target_freq, unsigned int relation)
> > +{
> > +     int ret;
>
> not really necessary

In cpufreq_set(), we need the return value.

>
> > +
> > +     if (policy->fast_switch_enabled) {
> > +             unsigned int idx;
> > +
> > +             target_freq = clamp_val(target_freq, policy->min, policy->max);
> > +
> > +             if (!policy->freq_table)
> > +                     return target_freq;
> > +
> > +             idx = cpufreq_frequency_table_target(policy, target_freq, relation);
> > +             policy->cached_resolved_idx = idx;
> > +             policy->cached_target_freq = target_freq;
> > +             ret = !cpufreq_driver_fast_switch(policy, policy->freq_table[idx].frequency);
> > +     } else {
> > +             ret = __cpufreq_driver_target(policy, target_freq, relation);
>
> NIT: could save the indent if you reverse conditions and ret early on !fast_switch

Okay, thanks!

>
> > +     }
> > +
> > +     return ret;
> > +}
> > +
> >  /**
> >   * cpufreq_set - set the CPU frequency
> >   * @policy: pointer to policy struct where freq is being set
> > @@ -41,7 +65,7 @@ static int cpufreq_set(struct cpufreq_policy *policy, unsigned int freq)
> >
> >       userspace->setspeed = freq;
> >
> > -     ret = __cpufreq_driver_target(policy, freq, CPUFREQ_RELATION_L);
> > +     ret = cpufreq_userspace_target_freq(policy, freq, CPUFREQ_RELATION_L);
> >   err:
> >       mutex_unlock(&userspace->mutex);
> >       return ret;
> > @@ -62,6 +86,8 @@ static int cpufreq_userspace_policy_init(struct cpufreq_policy *policy)
> >
> >       mutex_init(&userspace->mutex);
> >
> > +     cpufreq_enable_fast_switch(policy);
> > +
> >       policy->governor_data = userspace;
> >       return 0;
> >  }
> > @@ -72,6 +98,7 @@ static int cpufreq_userspace_policy_init(struct cpufreq_policy *policy)
> >   */
> >  static void cpufreq_userspace_policy_exit(struct cpufreq_policy *policy)
> >  {
> > +     cpufreq_disable_fast_switch(policy);
> >       kfree(policy->governor_data);
> >       policy->governor_data = NULL;
> >  }
> > @@ -112,13 +139,13 @@ static void cpufreq_userspace_policy_limits(struct cpufreq_policy *policy)
> >                policy->cpu, policy->min, policy->max, policy->cur, userspace->setspeed);
> >
> >       if (policy->max < userspace->setspeed)
> > -             __cpufreq_driver_target(policy, policy->max,
> > +             cpufreq_userspace_target_freq(policy, policy->max,
> >                                       CPUFREQ_RELATION_H);
> >       else if (policy->min > userspace->setspeed)
> > -             __cpufreq_driver_target(policy, policy->min,
> > +             cpufreq_userspace_target_freq(policy, policy->min,
> >                                       CPUFREQ_RELATION_L);
> >       else
> > -             __cpufreq_driver_target(policy, userspace->setspeed,
> > +             cpufreq_userspace_target_freq(policy, userspace->setspeed,
> >                                       CPUFREQ_RELATION_L);
> >
> >       mutex_unlock(&userspace->mutex);
>

BR

On 12/10/24 03:27, Xuewen Yan wrote:
> On Mon, Dec 9, 2024 at 6:36 PM Christian Loehle
> <christian.loehle@arm.com> wrote:
>>
>> On 12/9/24 08:14, Xuewen Yan wrote:
>>> Now, the userspace governor does not support userspace,
>>> if the driver only use the fast-switch and not add target_index(),
>>
>> Which driver does that? Is that actually valid?
>> No mainline driver from what I can see.
>>
> 
> Yes, indeed no mainline driver, It's on our own driver.

Fair enough. 
There seems to be handling for that case in cpufreq anyway.

> 
>>> it will cause uerspace not work.
>>
>> s/uerspace/userspace
>> to not work?
>>
>>> So add fast-switch support for userspace governor.
>>>
>>> Co-developed-by: Guohua Yan <guohua.yan@unisoc.com>
>>> Signed-off-by: Guohua Yan <guohua.yan@unisoc.com>
>>> Signed-off-by: Xuewen Yan <xuewen.yan@unisoc.com>
>>> ---
>>>  drivers/cpufreq/cpufreq_userspace.c | 35 +++++++++++++++++++++++++----
>>>  1 file changed, 31 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/drivers/cpufreq/cpufreq_userspace.c b/drivers/cpufreq/cpufreq_userspace.c
>>> index 2c42fee76daa..3a99197246ed 100644
>>> --- a/drivers/cpufreq/cpufreq_userspace.c
>>> +++ b/drivers/cpufreq/cpufreq_userspace.c
>>> @@ -21,6 +21,30 @@ struct userspace_policy {
>>>       struct mutex mutex;
>>>  };
>>>
>>> +static int cpufreq_userspace_target_freq(struct cpufreq_policy *policy,
>>> +                     unsigned int target_freq, unsigned int relation)
>>> +{
>>> +     int ret;
>>
>> not really necessary
> 
> In cpufreq_set(), we need the return value.

Sorry for not being clear enough, I suggested rewriting it
like this, although personal preference.

---
static int cpufreq_userspace_target_freq(struct cpufreq_policy *policy,
			unsigned int target_freq, unsigned int relation)
{
	unsigned int idx;

	if (!policy->fast_switch_enabled)
		return __cpufreq_driver_target(policy, target_freq, relation);

	target_freq = clamp_val(target_freq, policy->min, policy->max);
	if (!policy->freq_table)
		return target_freq;

	idx = cpufreq_frequency_table_target(policy, target_freq, relation);
	policy->cached_resolved_idx = idx;
	policy->cached_target_freq = target_freq;
	return !cpufreq_driver_fast_switch(policy, policy->freq_table[idx].frequency);
}