[PATCH] ARM: ux500: send cpufreq notification for all cpus

[PATCH] ARM: ux500: send cpufreq notification for all cpus

[Vincent Guittot]

The same clock is used for all cpus so we must notify the frequency change
for each one in order to update the configuration of all twd clockevents.

Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
---
 drivers/cpufreq/db8500-cpufreq.c |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/drivers/cpufreq/db8500-cpufreq.c b/drivers/cpufreq/db8500-cpufreq.c
index 0d8dd1c..4cb5785 100644
--- a/drivers/cpufreq/db8500-cpufreq.c
+++ b/drivers/cpufreq/db8500-cpufreq.c
@@ -76,13 +76,13 @@ static int db8500_cpufreq_target(struct cpufreq_policy *policy,
 
 	freqs.old = policy->cur;
 	freqs.new = freq_table[idx].frequency;
-	freqs.cpu = policy->cpu;
 
 	if (freqs.old == freqs.new)
 		return 0;
 
 	/* pre-change notification */
-	cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
+	for_each_online_cpu(freqs.cpu)
+		cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
 
 	/* request the PRCM unit for opp change */
 	if (prcmu_set_arm_opp(idx2opp[idx])) {
@@ -91,7 +91,8 @@ static int db8500_cpufreq_target(struct cpufreq_policy *policy,
 	}
 
 	/* post change notification */
-	cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
+	for_each_online_cpu(freqs.cpu)
+		cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
 
 	return 0;
 }
-- 
1.7.4.1

[PATCH] ARM: ux500: send cpufreq notification for all cpus

[Vincent Guittot]

The same clock is used for all cpus so we must notify the frequency change
for each one in order to update the configuration of all twd clockevents.

Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
---
 drivers/cpufreq/db8500-cpufreq.c |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/drivers/cpufreq/db8500-cpufreq.c b/drivers/cpufreq/db8500-cpufreq.c
index 0d8dd1c..4cb5785 100644
--- a/drivers/cpufreq/db8500-cpufreq.c
+++ b/drivers/cpufreq/db8500-cpufreq.c
@@ -76,13 +76,13 @@ static int db8500_cpufreq_target(struct cpufreq_policy *policy,
 
 	freqs.old = policy->cur;
 	freqs.new = freq_table[idx].frequency;
-	freqs.cpu = policy->cpu;
 
 	if (freqs.old == freqs.new)
 		return 0;
 
 	/* pre-change notification */
-	cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
+	for_each_online_cpu(freqs.cpu)
+		cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
 
 	/* request the PRCM unit for opp change */
 	if (prcmu_set_arm_opp(idx2opp[idx])) {
@@ -91,7 +91,8 @@ static int db8500_cpufreq_target(struct cpufreq_policy *policy,
 	}
 
 	/* post change notification */
-	cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
+	for_each_online_cpu(freqs.cpu)
+		cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
 
 	return 0;
 }
-- 
1.7.4.1

Re: [PATCH] ARM: ux500: send cpufreq notification for all cpus

[Linus Walleij]

On Wed, Aug 24, 2011 at 8:44 AM, Vincent Guittot
<vincent.guittot@linaro.org> wrote:

> The same clock is used for all cpus so we must notify the frequency change
> for each one in order to update the configuration of all twd clockevents.
>
> Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>

Acked-by: Linus Walleij <linus.walleij@linaro.org>

Yours,
Linus Walleij

[PATCH] ARM: ux500: send cpufreq notification for all cpus

[Linus Walleij]

On Wed, Aug 24, 2011 at 8:44 AM, Vincent Guittot
<vincent.guittot@linaro.org> wrote:

> The same clock is used for all cpus so we must notify the frequency change
> for each one in order to update the configuration of all twd clockevents.
>
> Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>

Acked-by: Linus Walleij <linus.walleij@linaro.org>

Yours,
Linus Walleij

Re: [PATCH] ARM: ux500: send cpufreq notification for all cpus

[Kyungmin Park]

On Wed, Aug 24, 2011 at 4:42 PM, Linus Walleij <linus.walleij@linaro.org> wrote:
> On Wed, Aug 24, 2011 at 8:44 AM, Vincent Guittot
> <vincent.guittot@linaro.org> wrote:
>
>> The same clock is used for all cpus so we must notify the frequency change
>> for each one in order to update the configuration of all twd clockevents.

BTW, does it same as other ARM cpus?. e.g., exynos4 has same codes as this.

How to you think it? Need to modify the same way?

Thank you,
Kyungmin Park
>>
>> Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
>
> Acked-by: Linus Walleij <linus.walleij@linaro.org>
>
> Yours,
> Linus Walleij
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>

[PATCH] ARM: ux500: send cpufreq notification for all cpus

[Kyungmin Park]

On Wed, Aug 24, 2011 at 4:42 PM, Linus Walleij <linus.walleij@linaro.org> wrote:
> On Wed, Aug 24, 2011 at 8:44 AM, Vincent Guittot
> <vincent.guittot@linaro.org> wrote:
>
>> The same clock is used for all cpus so we must notify the frequency change
>> for each one in order to update the configuration of all twd clockevents.

BTW, does it same as other ARM cpus?. e.g., exynos4 has same codes as this.

How to you think it? Need to modify the same way?

Thank you,
Kyungmin Park
>>
>> Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
>
> Acked-by: Linus Walleij <linus.walleij@linaro.org>
>
> Yours,
> Linus Walleij
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>

Re: [PATCH] ARM: ux500: send cpufreq notification for all cpus

[Vincent Guittot]

On 24 August 2011 09:56, Kyungmin Park <kmpark@infradead.org> wrote:
> On Wed, Aug 24, 2011 at 4:42 PM, Linus Walleij <linus.walleij@linaro.org> wrote:
>> On Wed, Aug 24, 2011 at 8:44 AM, Vincent Guittot
>> <vincent.guittot@linaro.org> wrote:
>>
>>> The same clock is used for all cpus so we must notify the frequency change
>>> for each one in order to update the configuration of all twd clockevents.
>
> BTW, does it same as other ARM cpus?. e.g., exynos4 has same codes as this.
>
> How to you think it? Need to modify the same way?
>

yes, each ARM platform which uses twd timer should check that the
notification is sent for each cpu.

> Thank you,
> Kyungmin Park
>>>
>>> Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
>>
>> Acked-by: Linus Walleij <linus.walleij@linaro.org>
>>
>> Yours,
>> Linus Walleij
>>
>> _______________________________________________
>> linux-arm-kernel mailing list
>> linux-arm-kernel@lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>>
>

[PATCH] ARM: ux500: send cpufreq notification for all cpus

[Vincent Guittot]

On 24 August 2011 09:56, Kyungmin Park <kmpark@infradead.org> wrote:
> On Wed, Aug 24, 2011 at 4:42 PM, Linus Walleij <linus.walleij@linaro.org> wrote:
>> On Wed, Aug 24, 2011 at 8:44 AM, Vincent Guittot
>> <vincent.guittot@linaro.org> wrote:
>>
>>> The same clock is used for all cpus so we must notify the frequency change
>>> for each one in order to update the configuration of all twd clockevents.
>
> BTW, does it same as other ARM cpus?. e.g., exynos4 has same codes as this.
>
> How to you think it? Need to modify the same way?
>

yes, each ARM platform which uses twd timer should check that the
notification is sent for each cpu.

> Thank you,
> Kyungmin Park
>>>
>>> Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
>>
>> Acked-by: Linus Walleij <linus.walleij@linaro.org>
>>
>> Yours,
>> Linus Walleij
>>
>> _______________________________________________
>> linux-arm-kernel mailing list
>> linux-arm-kernel at lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>>
>

Re: [PATCH] ARM: ux500: send cpufreq notification for all cpus

[Santosh]

On Wednesday 24 August 2011 12:14 PM, Vincent Guittot wrote:
> The same clock is used for all cpus so we must notify the frequency change
> for each one in order to update the configuration of all twd clockevents.
>
> Signed-off-by: Vincent Guittot<vincent.guittot@linaro.org>
> ---
>   drivers/cpufreq/db8500-cpufreq.c |    7 ++++---
>   1 files changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/cpufreq/db8500-cpufreq.c b/drivers/cpufreq/db8500-cpufreq.c
> index 0d8dd1c..4cb5785 100644
> --- a/drivers/cpufreq/db8500-cpufreq.c
> +++ b/drivers/cpufreq/db8500-cpufreq.c
> @@ -76,13 +76,13 @@ static int db8500_cpufreq_target(struct cpufreq_policy *policy,
>
>   	freqs.old = policy->cur;
>   	freqs.new = freq_table[idx].frequency;
> -	freqs.cpu = policy->cpu;
>
>   	if (freqs.old == freqs.new)
>   		return 0;
>
>   	/* pre-change notification */
> -	cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
> +	for_each_online_cpu(freqs.cpu)
> +		cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);

Blind notfier on all online CPU's will race. Take a example whete CPU
is marked online but not yet registered to CPUFREQ. I have seen such a
race in OMAP and fixed that using the policy->cpus field.

Regards
Santosh

[PATCH] ARM: ux500: send cpufreq notification for all cpus

[Santosh]

On Wednesday 24 August 2011 12:14 PM, Vincent Guittot wrote:
> The same clock is used for all cpus so we must notify the frequency change
> for each one in order to update the configuration of all twd clockevents.
>
> Signed-off-by: Vincent Guittot<vincent.guittot@linaro.org>
> ---
>   drivers/cpufreq/db8500-cpufreq.c |    7 ++++---
>   1 files changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/cpufreq/db8500-cpufreq.c b/drivers/cpufreq/db8500-cpufreq.c
> index 0d8dd1c..4cb5785 100644
> --- a/drivers/cpufreq/db8500-cpufreq.c
> +++ b/drivers/cpufreq/db8500-cpufreq.c
> @@ -76,13 +76,13 @@ static int db8500_cpufreq_target(struct cpufreq_policy *policy,
>
>   	freqs.old = policy->cur;
>   	freqs.new = freq_table[idx].frequency;
> -	freqs.cpu = policy->cpu;
>
>   	if (freqs.old == freqs.new)
>   		return 0;
>
>   	/* pre-change notification */
> -	cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
> +	for_each_online_cpu(freqs.cpu)
> +		cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);

Blind notfier on all online CPU's will race. Take a example whete CPU
is marked online but not yet registered to CPUFREQ. I have seen such a
race in OMAP and fixed that using the policy->cpus field.

Regards
Santosh

Re: [PATCH] ARM: ux500: send cpufreq notification for all cpus

[Vincent Guittot]

On 24 August 2011 10:26, Santosh <santosh.shilimkar@ti.com> wrote:
> On Wednesday 24 August 2011 12:14 PM, Vincent Guittot wrote:
>>
>> The same clock is used for all cpus so we must notify the frequency change
>> for each one in order to update the configuration of all twd clockevents.
>>
>> Signed-off-by: Vincent Guittot<vincent.guittot@linaro.org>
>> ---
>>  drivers/cpufreq/db8500-cpufreq.c |    7 ++++---
>>  1 files changed, 4 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/cpufreq/db8500-cpufreq.c
>> b/drivers/cpufreq/db8500-cpufreq.c
>> index 0d8dd1c..4cb5785 100644
>> --- a/drivers/cpufreq/db8500-cpufreq.c
>> +++ b/drivers/cpufreq/db8500-cpufreq.c
>> @@ -76,13 +76,13 @@ static int db8500_cpufreq_target(struct cpufreq_policy
>> *policy,
>>
>>        freqs.old = policy->cur;
>>        freqs.new = freq_table[idx].frequency;
>> -       freqs.cpu = policy->cpu;
>>
>>        if (freqs.old == freqs.new)
>>                return 0;
>>
>>        /* pre-change notification */
>> -       cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
>> +       for_each_online_cpu(freqs.cpu)
>> +               cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
>
> Blind notfier on all online CPU's will race. Take a example whete CPU
> is marked online but not yet registered to CPUFREQ. I have seen such a
> race in OMAP and fixed that using the policy->cpus field.
>

You're right, I'm going to use policy->cpus instead

> Regards
> Santosh
>
>

[PATCH] ARM: ux500: send cpufreq notification for all cpus

[Vincent Guittot]

On 24 August 2011 10:26, Santosh <santosh.shilimkar@ti.com> wrote:
> On Wednesday 24 August 2011 12:14 PM, Vincent Guittot wrote:
>>
>> The same clock is used for all cpus so we must notify the frequency change
>> for each one in order to update the configuration of all twd clockevents.
>>
>> Signed-off-by: Vincent Guittot<vincent.guittot@linaro.org>
>> ---
>> ?drivers/cpufreq/db8500-cpufreq.c | ? ?7 ++++---
>> ?1 files changed, 4 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/cpufreq/db8500-cpufreq.c
>> b/drivers/cpufreq/db8500-cpufreq.c
>> index 0d8dd1c..4cb5785 100644
>> --- a/drivers/cpufreq/db8500-cpufreq.c
>> +++ b/drivers/cpufreq/db8500-cpufreq.c
>> @@ -76,13 +76,13 @@ static int db8500_cpufreq_target(struct cpufreq_policy
>> *policy,
>>
>> ? ? ? ?freqs.old = policy->cur;
>> ? ? ? ?freqs.new = freq_table[idx].frequency;
>> - ? ? ? freqs.cpu = policy->cpu;
>>
>> ? ? ? ?if (freqs.old == freqs.new)
>> ? ? ? ? ? ? ? ?return 0;
>>
>> ? ? ? ?/* pre-change notification */
>> - ? ? ? cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
>> + ? ? ? for_each_online_cpu(freqs.cpu)
>> + ? ? ? ? ? ? ? cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
>
> Blind notfier on all online CPU's will race. Take a example whete CPU
> is marked online but not yet registered to CPUFREQ. I have seen such a
> race in OMAP and fixed that using the policy->cpus field.
>

You're right, I'm going to use policy->cpus instead

> Regards
> Santosh
>
>