Skip to contentSkip to conversation

cpufreq: Auto-register with energy model

PATCH In review Viresh Kumar 3 participants 19 messages Aug 11, 2021 – Aug 12, 2021

Subsystems (6)
lore
Share & tools
lore

Viresh Kumar

author maintainer
#0 lore raw

v2 cover cpufreq: Auto-register with energy model

Many cpufreq drivers register with the energy model for each policy and do exactly the same thing. Follow the footsteps of thermal-cooling, to get it done from the cpufreq core itself.

Provide a new callback, which will be called, if present, by the cpufreq core at the right moment (more on that in the code's comment). Also provide a generic implementation that uses dev_pm_opp_of_register_em().

This also allows us to register with the EM at a later point of time, compared to ->init(), from where the EM core can access cpufreq policy directly using cpufreq_cpu_get() type of helpers and perform other work, like marking few frequencies inefficient, this will be done separately.

This is build/boot tested by the bot for a couple of boards.

https://gitlab.com/vireshk/pmko/-/pipelines/351525873

Note that I haven't picked any of the Reviewed-by tags from the first version since the idea is very much changed here.

V1->V2:
- Add a callback instead of flag.
- Register before governor is initialized.
- Update scmi driver as well.
- Don't unregister from the EM core.

--
Viresh

Viresh Kumar (9):
cpufreq: Auto-register with energy model if asked
cpufreq: dt: Use auto-registration for energy model
cpufreq: imx6q: Use auto-registration for energy model
cpufreq: mediatek: Use auto-registration for energy model
cpufreq: omap: Use auto-registration for energy model
cpufreq: qcom-cpufreq-hw: Use auto-registration for energy model
cpufreq: scpi: Use auto-registration for energy model
cpufreq: vexpress: Use auto-registration for energy model
cpufreq: scmi: Use .register_em() callback

 drivers/cpufreq/cpufreq-dt.c           |  3 +-
 drivers/cpufreq/cpufreq.c              | 12 ++++++
 drivers/cpufreq/imx6q-cpufreq.c        |  2 +-
 drivers/cpufreq/mediatek-cpufreq.c     |  3 +-
 drivers/cpufreq/omap-cpufreq.c         |  2 +-
 drivers/cpufreq/qcom-cpufreq-hw.c      |  3 +-
 drivers/cpufreq/scmi-cpufreq.c         | 55 +++++++++++++++-----------
 drivers/cpufreq/scpi-cpufreq.c         |  3 +-
 drivers/cpufreq/vexpress-spc-cpufreq.c |  3 +-
 include/linux/cpufreq.h                | 14 +++++++
 10 files changed, 65 insertions(+), 35 deletions(-)
Signature · 1 line
2.31.1.272.g89b43f80a514
author maintainer
#1 lore raw

PATCH v2 1/9 cpufreq: Auto-register with energy model if asked

Many cpufreq drivers register with the energy model for each policy and do exactly the same thing. Follow the footsteps of thermal-cooling, to get it done from the cpufreq core itself.

Provide a new callback, which will be called, if present, by the cpufreq core at the right moment (more on that in the code's comment). Also provide a generic implementation that uses dev_pm_opp_of_register_em().

This also allows us to register with the EM at a later point of time, compared to ->init(), from where the EM core can access cpufreq policy directly using cpufreq_cpu_get() type of helpers and perform other work, like marking few frequencies inefficient, this will be done separately.

---
 drivers/cpufreq/cpufreq.c | 12 ++++++++++++
 include/linux/cpufreq.h   | 14 ++++++++++++++
 2 files changed, 26 insertions(+)
drivers/cpufreq/cpufreq.c +12 -0
index 06c526d66dd3..75974e7d6cc5 100644
@@ -1493,6 +1493,18 @@ static int cpufreq_online(unsigned int cpu)
 		write_unlock_irqrestore(&cpufreq_driver_lock, flags);
 	}
 
+	/*
+	 * Register with the energy model before sched_cpufreq_governor_change()
+	 * is called, which will result in rebuilding of the sched domains,
+	 * which should only be done once the energy model is properly
+	 * initialized for the policy first.
+	 *
+	 * Also, this should be called before the policy is registered with
+	 * cooling framework.
+	 */
+	if (cpufreq_driver->register_em)
+		cpufreq_driver->register_em(policy);
+
 	ret = cpufreq_init_policy(policy);
 	if (ret) {
 		pr_err("%s: Failed to initialize policy for cpu: %d (%d)\n",
include/linux/cpufreq.h +14 -0
index 9fd719475fcd..1295621f6c28 100644
@@ -9,10 +9,12 @@
 #define _LINUX_CPUFREQ_H
 
 #include <linux/clk.h>
+#include <linux/cpu.h>
 #include <linux/cpumask.h>
 #include <linux/completion.h>
 #include <linux/kobject.h>
 #include <linux/notifier.h>
+#include <linux/pm_opp.h>
 #include <linux/pm_qos.h>
 #include <linux/spinlock.h>
 #include <linux/sysfs.h>
@@ -373,6 +375,12 @@ struct cpufreq_driver {
 	/* platform specific boost support code */
 	bool		boost_enabled;
 	int		(*set_boost)(struct cpufreq_policy *policy, int state);
+
+	/*
+	 * Set by drivers that want the core to automatically register the
+	 * policy's devices with Energy Model.
+	 */
+	void		(*register_em)(struct cpufreq_policy *policy);
 };
 
 /* flags */
@@ -1046,4 +1054,10 @@ unsigned int cpufreq_generic_get(unsigned int cpu);
 void cpufreq_generic_init(struct cpufreq_policy *policy,
 		struct cpufreq_frequency_table *table,
 		unsigned int transition_latency);
+
+static inline void cpufreq_register_em_with_opp(struct cpufreq_policy *policy)
+{
+	dev_pm_opp_of_register_em(get_cpu_device(policy->cpu),
+				  policy->related_cpus);
+}
 #endif /* _LINUX_CPUFREQ_H */
Signature · 1 line
2.31.1.272.g89b43f80a514
Reviewed-by on v2 1/9
#2 lore raw

Re: [PATCH V2 1/9] cpufreq: Auto-register with energy model if asked

Quoted 28 lines · Viresh Kumar

Many cpufreq drivers register with the energy model for each policy and do exactly the same thing. Follow the footsteps of thermal-cooling, to get it done from the cpufreq core itself.

Provide a new callback, which will be called, if present, by the cpufreq core at the right moment (more on that in the code's comment). Also provide a generic implementation that uses dev_pm_opp_of_register_em().

This also allows us to register with the EM at a later point of time, compared to ->init(), from where the EM core can access cpufreq policy directly using cpufreq_cpu_get() type of helpers and perform other work, like marking few frequencies inefficient, this will be done separately.

---
 drivers/cpufreq/cpufreq.c | 12 ++++++++++++
 include/linux/cpufreq.h   | 14 ++++++++++++++
 2 files changed, 26 insertions(+)
drivers/cpufreq/cpufreq.c +1 -0
index 06c526d66dd3..75974e7d6cc5 100644
@@ -1493,6 +1493,18 @@ static int cpufreq_online(unsigned int cpu)
 		write_unlock_irqrestore(&cpufreq_driver_lock, flags);
 	}
 
+	/*

drivers/cpufreq/cpufreq.c static int cpufreq_online(unsigned int cpu) · line 1506

Quoted 10 lines
drivers/cpufreq/cpufreq.c +10 -0
+	 * Register with the energy model before sched_cpufreq_governor_change()
+	 * is called, which will result in rebuilding of the sched domains,
+	 * which should only be done once the energy model is properly
+	 * initialized for the policy first.
+	 *
+	 * Also, this should be called before the policy is registered with
+	 * cooling framework.
+	 */
+	if (cpufreq_driver->register_em)
+		cpufreq_driver->register_em(policy);

Maybe move that to the 'if (new_policy)' block above? There is currently no need to re-register the EM on CPU hotplug.

Quoted 34 lines
drivers/cpufreq/cpufreq.c +0 -0
 	ret = cpufreq_init_policy(policy);
 	if (ret) {
 		pr_err("%s: Failed to initialize policy for cpu: %d (%d)\n",
include/linux/cpufreq.h +8 -0
index 9fd719475fcd..1295621f6c28 100644
@@ -9,10 +9,12 @@
 #define _LINUX_CPUFREQ_H
 
 #include <linux/clk.h>
+#include <linux/cpu.h>
 #include <linux/cpumask.h>
 #include <linux/completion.h>
 #include <linux/kobject.h>
 #include <linux/notifier.h>
+#include <linux/pm_opp.h>
 #include <linux/pm_qos.h>
 #include <linux/spinlock.h>
 #include <linux/sysfs.h>
@@ -373,6 +375,12 @@ struct cpufreq_driver {
 	/* platform specific boost support code */
 	bool		boost_enabled;
 	int		(*set_boost)(struct cpufreq_policy *policy, int state);
+
+	/*
+	 * Set by drivers that want the core to automatically register the
+	 * policy's devices with Energy Model.
+	 */
+	void		(*register_em)(struct cpufreq_policy *policy);
 };
 
 /* flags */

include/linux/cpufreq.h unsigned int cpufreq_generic_get(unsigned int cpu); · line 1062

Quoted 9 lines
include/linux/cpufreq.h +6 -0
@@ -1046,4 +1054,10 @@ unsigned int cpufreq_generic_get(unsigned int cpu);
 void cpufreq_generic_init(struct cpufreq_policy *policy,
 		struct cpufreq_frequency_table *table,
 		unsigned int transition_latency);
+
+static inline void cpufreq_register_em_with_opp(struct cpufreq_policy *policy)
+{
+	dev_pm_opp_of_register_em(get_cpu_device(policy->cpu),
+				  policy->related_cpus);
+}

I was thinking this could go in pm_opp.h instead, but it doesn't really matter.

So, with the first comment above fixed:

Thanks,
Quentin

Reviewed-by on v2 1/9
#3 lore raw

Re: [PATCH V2 1/9] cpufreq: Auto-register with energy model if asked

In reply to Viresh Kumar
Quoted 60 lines · Viresh Kumar

Many cpufreq drivers register with the energy model for each policy and do exactly the same thing. Follow the footsteps of thermal-cooling, to get it done from the cpufreq core itself.

Provide a new callback, which will be called, if present, by the cpufreq core at the right moment (more on that in the code's comment). Also provide a generic implementation that uses dev_pm_opp_of_register_em().

This also allows us to register with the EM at a later point of time, compared to ->init(), from where the EM core can access cpufreq policy directly using cpufreq_cpu_get() type of helpers and perform other work, like marking few frequencies inefficient, this will be done separately.

---
  drivers/cpufreq/cpufreq.c | 12 ++++++++++++
  include/linux/cpufreq.h   | 14 ++++++++++++++
  2 files changed, 26 insertions(+)
drivers/cpufreq/cpufreq.c +12 -0
index 06c526d66dd3..75974e7d6cc5 100644
@@ -1493,6 +1493,18 @@ static int cpufreq_online(unsigned int cpu)
  		write_unlock_irqrestore(&cpufreq_driver_lock, flags);
  	}
 
+	/*
+	 * Register with the energy model before sched_cpufreq_governor_change()
+	 * is called, which will result in rebuilding of the sched domains,
+	 * which should only be done once the energy model is properly
+	 * initialized for the policy first.
+	 *
+	 * Also, this should be called before the policy is registered with
+	 * cooling framework.
+	 */
+	if (cpufreq_driver->register_em)
+		cpufreq_driver->register_em(policy);
+
  	ret = cpufreq_init_policy(policy);
  	if (ret) {
  		pr_err("%s: Failed to initialize policy for cpu: %d (%d)\n",
include/linux/cpufreq.h +2 -0
index 9fd719475fcd..1295621f6c28 100644
@@ -9,10 +9,12 @@
  #define _LINUX_CPUFREQ_H
 
  #include <linux/clk.h>
+#include <linux/cpu.h>
  #include <linux/cpumask.h>
  #include <linux/completion.h>
  #include <linux/kobject.h>
  #include <linux/notifier.h>
+#include <linux/pm_opp.h>
  #include <linux/pm_qos.h>
  #include <linux/spinlock.h>
  #include <linux/sysfs.h>

include/linux/cpufreq.h struct cpufreq_driver { · line 381

Quoted 7 lines
include/linux/cpufreq.h +4 -0
@@ -373,6 +375,12 @@ struct cpufreq_driver {
  	/* platform specific boost support code */
  	bool		boost_enabled;
  	int		(*set_boost)(struct cpufreq_policy *policy, int state);
+
+	/*
+	 * Set by drivers that want the core to automatically register the
+	 * policy's devices with Energy Model.

It covers one use case. I would add also something about customized EM setup, which might be provided by drivers and implemented then in this callback. It won't be only for automatic registration.

Quoted 7 lines
include/linux/cpufreq.h +2 -0
+	 */
+	void		(*register_em)(struct cpufreq_policy *policy);
  };
 
  /* flags */

include/linux/cpufreq.h unsigned int cpufreq_generic_get(unsigned int cpu); · line 1063

Quoted 10 lines
include/linux/cpufreq.h +6 -0
@@ -1046,4 +1054,10 @@ unsigned int cpufreq_generic_get(unsigned int cpu);
  void cpufreq_generic_init(struct cpufreq_policy *policy,
  		struct cpufreq_frequency_table *table,
  		unsigned int transition_latency);
+
+static inline void cpufreq_register_em_with_opp(struct cpufreq_policy *policy)
+{
+	dev_pm_opp_of_register_em(get_cpu_device(policy->cpu),
+				  policy->related_cpus);
+}
  #endif /* _LINUX_CPUFREQ_H */

There rest is OK,

author maintainer
#4 lore raw

PATCH v2 2/9 cpufreq: dt: Use auto-registration for energy model

In reply to Viresh Kumar

Set the newly added .register_em() callback with cpufreq_register_em_with_opp() to automatically register with the EM core.

---
 drivers/cpufreq/cpufreq-dt.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/cpufreq/cpufreq-dt.c +1 -2
index ece52863ba62..8fcaba541539 100644
@@ -143,8 +143,6 @@ static int cpufreq_init(struct cpufreq_policy *policy)
 		cpufreq_dt_attr[1] = &cpufreq_freq_attr_scaling_boost_freqs;
 	}
 
-	dev_pm_opp_of_register_em(cpu_dev, policy->cpus);
-
 	return 0;
 
 out_clk_put:
@@ -184,6 +182,7 @@ static struct cpufreq_driver dt_cpufreq_driver = {
 	.exit = cpufreq_exit,
 	.online = cpufreq_online,
 	.offline = cpufreq_offline,
+	.register_em = cpufreq_register_em_with_opp,
 	.name = "cpufreq-dt",
 	.attr = cpufreq_dt_attr,
 	.suspend = cpufreq_generic_suspend,
Signature · 1 line
2.31.1.272.g89b43f80a514
author maintainer
#5 lore raw

PATCH v2 3/9 cpufreq: imx6q: Use auto-registration for energy model

In reply to Viresh Kumar

Set the newly added .register_em() callback with cpufreq_register_em_with_opp() to automatically register with the EM core.

---
 drivers/cpufreq/imx6q-cpufreq.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/cpufreq/imx6q-cpufreq.c +1 -1
index 5bf5fc759881..90beb26ed34e 100644
@@ -192,7 +192,6 @@ static int imx6q_cpufreq_init(struct cpufreq_policy *policy)
 	policy->clk = clks[ARM].clk;
 	cpufreq_generic_init(policy, freq_table, transition_latency);
 	policy->suspend_freq = max_freq;
-	dev_pm_opp_of_register_em(cpu_dev, policy->cpus);
 
 	return 0;
 }
@@ -204,6 +203,7 @@ static struct cpufreq_driver imx6q_cpufreq_driver = {
 	.target_index = imx6q_set_target,
 	.get = cpufreq_generic_get,
 	.init = imx6q_cpufreq_init,
+	.register_em = cpufreq_register_em_with_opp,
 	.name = "imx6q-cpufreq",
 	.attr = cpufreq_generic_attr,
 	.suspend = cpufreq_generic_suspend,
Signature · 1 line
2.31.1.272.g89b43f80a514
author maintainer
#6 lore raw

PATCH v2 4/9 cpufreq: mediatek: Use auto-registration for energy model

In reply to Viresh Kumar

Set the newly added .register_em() callback with cpufreq_register_em_with_opp() to automatically register with the EM core.

---
 drivers/cpufreq/mediatek-cpufreq.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/cpufreq/mediatek-cpufreq.c +1 -2
index 87019d5a9547..866163883b48 100644
@@ -448,8 +448,6 @@ static int mtk_cpufreq_init(struct cpufreq_policy *policy)
 	policy->driver_data = info;
 	policy->clk = info->cpu_clk;
 
-	dev_pm_opp_of_register_em(info->cpu_dev, policy->cpus);
-
 	return 0;
 }
 
@@ -471,6 +469,7 @@ static struct cpufreq_driver mtk_cpufreq_driver = {
 	.get = cpufreq_generic_get,
 	.init = mtk_cpufreq_init,
 	.exit = mtk_cpufreq_exit,
+	.register_em = cpufreq_register_em_with_opp,
 	.name = "mtk-cpufreq",
 	.attr = cpufreq_generic_attr,
 };
Signature · 1 line
2.31.1.272.g89b43f80a514
author maintainer
#7 lore raw

PATCH v2 5/9 cpufreq: omap: Use auto-registration for energy model

In reply to Viresh Kumar

Set the newly added .register_em() callback with cpufreq_register_em_with_opp() to automatically register with the EM core.

---
 drivers/cpufreq/omap-cpufreq.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/cpufreq/omap-cpufreq.c +1 -1
index e035ee216b0f..1b50df06c6bc 100644
@@ -131,7 +131,6 @@ static int omap_cpu_init(struct cpufreq_policy *policy)
 
 	/* FIXME: what's the actual transition time? */
 	cpufreq_generic_init(policy, freq_table, 300 * 1000);
-	dev_pm_opp_of_register_em(mpu_dev, policy->cpus);
 
 	return 0;
 }
@@ -150,6 +149,7 @@ static struct cpufreq_driver omap_driver = {
 	.get		= cpufreq_generic_get,
 	.init		= omap_cpu_init,
 	.exit		= omap_cpu_exit,
+	.register_em	= cpufreq_register_em_with_opp,
 	.name		= "omap",
 	.attr		= cpufreq_generic_attr,
 };
Signature · 1 line
2.31.1.272.g89b43f80a514
author maintainer
#8 lore raw

PATCH v2 6/9 cpufreq: qcom-cpufreq-hw: Use auto-registration for energy model

In reply to Viresh Kumar

Set the newly added .register_em() callback with cpufreq_register_em_with_opp() to automatically register with the EM core.

---
 drivers/cpufreq/qcom-cpufreq-hw.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/cpufreq/qcom-cpufreq-hw.c +1 -2
index f86859bf76f1..c2e71c430fbf 100644
@@ -362,8 +362,6 @@ static int qcom_cpufreq_hw_cpu_init(struct cpufreq_policy *policy)
 		goto error;
 	}
 
-	dev_pm_opp_of_register_em(cpu_dev, policy->cpus);
-
 	if (policy_has_boost_freq(policy)) {
 		ret = cpufreq_enable_boost_support();
 		if (ret)
@@ -412,6 +410,7 @@ static struct cpufreq_driver cpufreq_qcom_hw_driver = {
 	.get		= qcom_cpufreq_hw_get,
 	.init		= qcom_cpufreq_hw_cpu_init,
 	.exit		= qcom_cpufreq_hw_cpu_exit,
+	.register_em	= cpufreq_register_em_with_opp,
 	.fast_switch    = qcom_cpufreq_hw_fast_switch,
 	.name		= "qcom-cpufreq-hw",
 	.attr		= qcom_cpufreq_hw_attr,
Signature · 1 line
2.31.1.272.g89b43f80a514
author maintainer
#9 lore raw

PATCH v2 7/9 cpufreq: scpi: Use auto-registration for energy model

In reply to Viresh Kumar

Set the newly added .register_em() callback with cpufreq_register_em_with_opp() to automatically register with the EM core.

---
 drivers/cpufreq/scpi-cpufreq.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/cpufreq/scpi-cpufreq.c +1 -2
index d6a698a1b5d1..bda3e7d42964 100644
@@ -163,8 +163,6 @@ static int scpi_cpufreq_init(struct cpufreq_policy *policy)
 
 	policy->fast_switch_possible = false;
 
-	dev_pm_opp_of_register_em(cpu_dev, policy->cpus);
-
 	return 0;
 
 out_free_cpufreq_table:
@@ -200,6 +198,7 @@ static struct cpufreq_driver scpi_cpufreq_driver = {
 	.init	= scpi_cpufreq_init,
 	.exit	= scpi_cpufreq_exit,
 	.target_index	= scpi_cpufreq_set_target,
+	.register_em	= cpufreq_register_em_with_opp,
 };
 
 static int scpi_cpufreq_probe(struct platform_device *pdev)
Signature · 1 line
2.31.1.272.g89b43f80a514
author maintainer
#10 lore raw

PATCH v2 8/9 cpufreq: vexpress: Use auto-registration for energy model

In reply to Viresh Kumar

Set the newly added .register_em() callback with cpufreq_register_em_with_opp() to automatically register with the EM core.

---
 drivers/cpufreq/vexpress-spc-cpufreq.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/cpufreq/vexpress-spc-cpufreq.c +1 -2
index ab56813b7256..284b6bd040b1 100644
@@ -440,8 +440,6 @@ static int ve_spc_cpufreq_init(struct cpufreq_policy *policy)
 	policy->freq_table = freq_table[cur_cluster];
 	policy->cpuinfo.transition_latency = 1000000; /* 1 ms */
 
-	dev_pm_opp_of_register_em(cpu_dev, policy->cpus);
-
 	if (is_bL_switching_enabled())
 		per_cpu(cpu_last_req_freq, policy->cpu) =
 						clk_get_cpu_rate(policy->cpu);
@@ -475,6 +473,7 @@ static struct cpufreq_driver ve_spc_cpufreq_driver = {
 	.get			= ve_spc_cpufreq_get_rate,
 	.init			= ve_spc_cpufreq_init,
 	.exit			= ve_spc_cpufreq_exit,
+	.register_em		= cpufreq_register_em_with_opp,
 	.attr			= cpufreq_generic_attr,
 };
 
Signature · 1 line
2.31.1.272.g89b43f80a514
author maintainer
#11 lore raw

PATCH v2 9/9 cpufreq: scmi: Use .register_em() callback

In reply to Viresh Kumar

Set the newly added .register_em() callback to register with the EM after the cpufreq policy is properly initialized.

---
 drivers/cpufreq/scmi-cpufreq.c | 55 ++++++++++++++++++++--------------
 1 file changed, 32 insertions(+), 23 deletions(-)
drivers/cpufreq/scmi-cpufreq.c +32 -23
index 75f818d04b48..b916c9e22921 100644
@@ -22,7 +22,9 @@
 
 struct scmi_data {
 	int domain_id;
+	int nr_opp;
 	struct device *cpu_dev;
+	cpumask_var_t opp_shared_cpus;
 };
 
 static struct scmi_protocol_handle *ph;
@@ -123,9 +125,6 @@ static int scmi_cpufreq_init(struct cpufreq_policy *policy)
 	struct device *cpu_dev;
 	struct scmi_data *priv;
 	struct cpufreq_frequency_table *freq_table;
-	struct em_data_callback em_cb = EM_DATA_CB(scmi_get_cpu_power);
-	cpumask_var_t opp_shared_cpus;
-	bool power_scale_mw;
 
 	cpu_dev = get_cpu_device(policy->cpu);
 	if (!cpu_dev) {
@@ -133,9 +132,15 @@ static int scmi_cpufreq_init(struct cpufreq_policy *policy)
 		return -ENODEV;
 	}
 
-	if (!zalloc_cpumask_var(&opp_shared_cpus, GFP_KERNEL))
+	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
+	if (!priv)
 		return -ENOMEM;
 
+	if (!zalloc_cpumask_var(&priv->opp_shared_cpus, GFP_KERNEL)) {
+		ret = -ENOMEM;
+		goto out_free_priv;
+	}
+
 	/* Obtain CPUs that share SCMI performance controls */
 	ret = scmi_get_sharing_cpus(cpu_dev, policy->cpus);
 	if (ret) {
@@ -148,14 +153,14 @@ static int scmi_cpufreq_init(struct cpufreq_policy *policy)
 	 * The OPP 'sharing cpus' info may come from DT through an empty opp
 	 * table and opp-shared.
 	 */
-	ret = dev_pm_opp_of_get_sharing_cpus(cpu_dev, opp_shared_cpus);
-	if (ret || !cpumask_weight(opp_shared_cpus)) {
+	ret = dev_pm_opp_of_get_sharing_cpus(cpu_dev, priv->opp_shared_cpus);
+	if (ret || !cpumask_weight(priv->opp_shared_cpus)) {
 		/*
 		 * Either opp-table is not set or no opp-shared was found.
 		 * Use the CPU mask from SCMI to designate CPUs sharing an OPP
 		 * table.
 		 */
-		cpumask_copy(opp_shared_cpus, policy->cpus);
+		cpumask_copy(priv->opp_shared_cpus, policy->cpus);
 	}
 
 	 /*
@@ -180,7 +185,7 @@ static int scmi_cpufreq_init(struct cpufreq_policy *policy)
 			goto out_free_opp;
 		}
 
-		ret = dev_pm_opp_set_sharing_cpus(cpu_dev, opp_shared_cpus);
+		ret = dev_pm_opp_set_sharing_cpus(cpu_dev, priv->opp_shared_cpus);
 		if (ret) {
 			dev_err(cpu_dev, "%s: failed to mark OPPs as shared: %d\n",
 				__func__, ret);
@@ -188,21 +193,13 @@ static int scmi_cpufreq_init(struct cpufreq_policy *policy)
 			goto out_free_opp;
 		}
 
-		power_scale_mw = perf_ops->power_scale_mw_get(ph);
-		em_dev_register_perf_domain(cpu_dev, nr_opp, &em_cb,
-					    opp_shared_cpus, power_scale_mw);
-	}
-
-	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
-	if (!priv) {
-		ret = -ENOMEM;
-		goto out_free_opp;
+		priv->nr_opp = nr_opp;
 	}
 
 	ret = dev_pm_opp_init_cpufreq_table(cpu_dev, &freq_table);
 	if (ret) {
 		dev_err(cpu_dev, "failed to init cpufreq table: %d\n", ret);
-		goto out_free_priv;
+		goto out_free_opp;
 	}
 
 	priv->cpu_dev = cpu_dev;
@@ -223,17 +220,16 @@ static int scmi_cpufreq_init(struct cpufreq_policy *policy)
 	policy->fast_switch_possible =
 		perf_ops->fast_switch_possible(ph, cpu_dev);
 
-	free_cpumask_var(opp_shared_cpus);
 	return 0;
 
-out_free_priv:
-	kfree(priv);
-
 out_free_opp:
 	dev_pm_opp_remove_all_dynamic(cpu_dev);
 
 out_free_cpumask:
-	free_cpumask_var(opp_shared_cpus);
+	free_cpumask_var(priv->opp_shared_cpus);
+
+out_free_priv:
+	kfree(priv);
 
 	return ret;
 }
@@ -244,11 +240,23 @@ static int scmi_cpufreq_exit(struct cpufreq_policy *policy)
 
 	dev_pm_opp_free_cpufreq_table(priv->cpu_dev, &policy->freq_table);
 	dev_pm_opp_remove_all_dynamic(priv->cpu_dev);
+	free_cpumask_var(priv->opp_shared_cpus);
 	kfree(priv);
 
 	return 0;
 }
 
+static void scmi_cpufreq_register_em(struct cpufreq_policy *policy)
+{
+	struct em_data_callback em_cb = EM_DATA_CB(scmi_get_cpu_power);
+	bool power_scale_mw = perf_ops->power_scale_mw_get(ph);
+	struct scmi_data *priv = policy->driver_data;
+
+	em_dev_register_perf_domain(get_cpu_device(policy->cpu), priv->nr_opp,
+				    &em_cb, priv->opp_shared_cpus,
+				    power_scale_mw);
+}
+
 static struct cpufreq_driver scmi_cpufreq_driver = {
 	.name	= "scmi",
 	.flags	= CPUFREQ_HAVE_GOVERNOR_PER_POLICY |
@@ -261,6 +269,7 @@ static struct cpufreq_driver scmi_cpufreq_driver = {
 	.get	= scmi_cpufreq_get_rate,
 	.init	= scmi_cpufreq_init,
 	.exit	= scmi_cpufreq_exit,
+	.register_em	= scmi_cpufreq_register_em,
 };
 
 static int scmi_cpufreq_probe(struct scmi_device *sdev)
Signature · 1 line
2.31.1.272.g89b43f80a514
#12 lore raw

Re: [PATCH V2 9/9] cpufreq: scmi: Use .register_em() callback

Quoted 13 lines · Viresh Kumar

Set the newly added .register_em() callback to register with the EM after the cpufreq policy is properly initialized.

---
 drivers/cpufreq/scmi-cpufreq.c | 55 ++++++++++++++++++++--------------
 1 file changed, 32 insertions(+), 23 deletions(-)
drivers/cpufreq/scmi-cpufreq.c +0 -0
index 75f818d04b48..b916c9e22921 100644

drivers/cpufreq/scmi-cpufreq.c line 27

Quoted 6 lines
drivers/cpufreq/scmi-cpufreq.c +2 -0
@@ -22,7 +22,9 @@
 
 struct scmi_data {
 	int domain_id;
+	int nr_opp;
 	struct device *cpu_dev;
+	cpumask_var_t opp_shared_cpus;

Can we use policy->related_cpus and friends directly in the callback instead? That should simplify the patch a bit.

Also, we can probably afford calling dev_pm_opp_get_opp_count() from the em_register callback as it is not a hot path, which would avoid wasting some 'resident' memory here that is only used during init.

Thanks,
Quentin

#13 lore raw

Re: [PATCH V2 9/9] cpufreq: scmi: Use .register_em() callback

Quoted 15 lines · Quentin Perret
Quoted 14 lines · Viresh Kumar

Set the newly added .register_em() callback to register with the EM after the cpufreq policy is properly initialized.

---
  drivers/cpufreq/scmi-cpufreq.c | 55 ++++++++++++++++++++--------------
  1 file changed, 32 insertions(+), 23 deletions(-)
drivers/cpufreq/scmi-cpufreq.c +0 -0
index 75f818d04b48..b916c9e22921 100644
@@ -22,7 +22,9 @@
 
Quoted 7 lines

drivers/cpufreq/scmi-cpufreq.c line 27

Quoted 5 lines
drivers/cpufreq/scmi-cpufreq.c +2 -0
  struct scmi_data {
  	int domain_id;
+	int nr_opp;
  	struct device *cpu_dev;
+	cpumask_var_t opp_shared_cpus;

Can we use policy->related_cpus and friends directly in the callback

Unfortunately not. This tricky setup code was introduced because we may have a platform with per-CPU policy, so single bit set in policy->related_cpus, but we want EAS to be still working on set of CPUs. That's why we construct temporary cpumask and pass it to EM.

Quoted 9 lines

instead? That should simplify the patch a bit.

Also, we can probably afford calling dev_pm_opp_get_opp_count() from the em_register callback as it is not a hot path, which would avoid wasting some 'resident' memory here that is only used during init.

Thanks,
Quentin

#14 lore raw

Re: [PATCH V2 9/9] cpufreq: scmi: Use .register_em() callback

Quoted 25 lines · Lukasz Luba
Quoted 9 lines · Quentin Perret
Quoted 8 lines · Viresh Kumar

Set the newly added .register_em() callback to register with the EM after the cpufreq policy is properly initialized.

---
  drivers/cpufreq/scmi-cpufreq.c | 55 ++++++++++++++++++++--------------
  1 file changed, 32 insertions(+), 23 deletions(-)
Quoted 12 lines

drivers/cpufreq/scmi-cpufreq.c line 26

Quoted 10 lines
drivers/cpufreq/scmi-cpufreq.c +2 -0
index 75f818d04b48..b916c9e22921 100644
@@ -22,7 +22,9 @@
  struct scmi_data {
  	int domain_id;
+	int nr_opp;
  	struct device *cpu_dev;
+	cpumask_var_t opp_shared_cpus;

Can we use policy->related_cpus and friends directly in the callback

Quoted 4 lines

Unfortunately not. This tricky setup code was introduced because we may have a platform with per-CPU policy, so single bit set in policy->related_cpus, but we want EAS to be still working on set of CPUs. That's why we construct temporary cpumask and pass it to EM.

Aha, I see this now. Hmm, those platforms better have AMUs then, otherwise PELT signals will be wonky ...

I was going to suggest using dev_pm_opp_get_sharing_cpus() from the callback instead, but maybe that's overkill as we'd need to allocate a temporary cpumask and all. So n/m this patch should be fine as is.

#15 lore raw

Re: [PATCH V2 9/9] cpufreq: scmi: Use .register_em() callback

Quoted 26 lines · Quentin Perret
Quoted 25 lines · Lukasz Luba
Quoted 21 lines · Quentin Perret
Quoted 13 lines · Viresh Kumar

Set the newly added .register_em() callback to register with the EM after the cpufreq policy is properly initialized.

---
   drivers/cpufreq/scmi-cpufreq.c | 55 ++++++++++++++++++++--------------
   1 file changed, 32 insertions(+), 23 deletions(-)
drivers/cpufreq/scmi-cpufreq.c +0 -0
index 75f818d04b48..b916c9e22921 100644

drivers/cpufreq/scmi-cpufreq.c line 26

Quoted 5 lines
drivers/cpufreq/scmi-cpufreq.c +2 -0
@@ -22,7 +22,9 @@
   struct scmi_data {
   	int domain_id;
+	int nr_opp;
   	struct device *cpu_dev;
+	cpumask_var_t opp_shared_cpus;

Can we use policy->related_cpus and friends directly in the callback

Quoted 7 lines
Quoted 4 lines

Unfortunately not. This tricky setup code was introduced because we may have a platform with per-CPU policy, so single bit set in policy->related_cpus, but we want EAS to be still working on set of CPUs. That's why we construct temporary cpumask and pass it to EM.

Aha, I see this now. Hmm, those platforms better have AMUs then, otherwise PELT signals will be wonky ...

That's the plan, to have the AMUs. We suggest that, but reality would tell... ;)

Quoted 5 lines

I was going to suggest using dev_pm_opp_get_sharing_cpus() from the callback instead, but maybe that's overkill as we'd need to allocate a temporary cpumask and all. So n/m this patch should be fine as is.

author maintainer
#16 lore raw

Re: [PATCH V2 9/9] cpufreq: scmi: Use .register_em() callback

In reply to Quentin Perret
Quoted 3 lines · Quentin Perret

Also, we can probably afford calling dev_pm_opp_get_opp_count() from the em_register callback as it is not a hot path, which would avoid wasting some 'resident' memory here that is only used during init.

We also need to make sure that OPPs are available in init(), else we fail. So, we can't really move that out.

Signature · 1 line
viresh
#17 lore raw

Re: [PATCH V2 9/9] cpufreq: scmi: Use .register_em() callback

In reply to Viresh Kumar
Quoted 9 lines · Viresh Kumar

Set the newly added .register_em() callback to register with the EM after the cpufreq policy is properly initialized.

---
  drivers/cpufreq/scmi-cpufreq.c | 55 ++++++++++++++++++++--------------
  1 file changed, 32 insertions(+), 23 deletions(-)

drivers/cpufreq/scmi-cpufreq.c static int scmi_cpufreq_exit(struct cpufreq_policy *policy) · line 257

Quoted 9 lines
drivers/cpufreq/scmi-cpufreq.c +9 -0
+static void scmi_cpufreq_register_em(struct cpufreq_policy *policy)
+{
+	struct em_data_callback em_cb = EM_DATA_CB(scmi_get_cpu_power);
+	bool power_scale_mw = perf_ops->power_scale_mw_get(ph);
+	struct scmi_data *priv = policy->driver_data;
+
+	em_dev_register_perf_domain(get_cpu_device(policy->cpu), priv->nr_opp,
+				    &em_cb, priv->opp_shared_cpus,
+				    power_scale_mw);

I would free the priv->opp_shared_cpus mask here, since we don't need it anymore and memory can be reclaimed. Don't forget this setup would be called N CPUs times, on this per-CPU policy platform.

If freed here, then also there wouldn't be a need to free it in scmi_cpufreq_exit() so you can remove it from there.

Quoted 7 lines
drivers/cpufreq/scmi-cpufreq.c +2 -0
+}
+
  static struct cpufreq_driver scmi_cpufreq_driver = {
  	.name	= "scmi",
  	.flags	= CPUFREQ_HAVE_GOVERNOR_PER_POLICY |

drivers/cpufreq/scmi-cpufreq.c static struct cpufreq_driver scmi_cpufreq_driver = { · line 275

Quoted 7 lines
drivers/cpufreq/scmi-cpufreq.c +1 -0
@@ -261,6 +269,7 @@ static struct cpufreq_driver scmi_cpufreq_driver = {
  	.get	= scmi_cpufreq_get_rate,
  	.init	= scmi_cpufreq_init,
  	.exit	= scmi_cpufreq_exit,
+	.register_em	= scmi_cpufreq_register_em,
  };
 
  static int scmi_cpufreq_probe(struct scmi_device *sdev)

I will test&review this patch on Monday when I re-flash custom FW to my Juno (just to be sure that this per-CPU cpufreq + shared EM/EAS works).

author maintainer
#18 lore raw

Re: [PATCH V2 9/9] cpufreq: scmi: Use .register_em() callback

Quoted 11 lines · Lukasz Luba
Quoted 7 lines · Viresh Kumar

Set the newly added .register_em() callback to register with the EM after the cpufreq policy is properly initialized.

---
  drivers/cpufreq/scmi-cpufreq.c | 55 ++++++++++++++++++++--------------
  1 file changed, 32 insertions(+), 23 deletions(-)
Quoted 12 lines

drivers/cpufreq/scmi-cpufreq.c static int scmi_cpufreq_exit(struct cpufreq_policy *policy) · line 257

Quoted 9 lines
drivers/cpufreq/scmi-cpufreq.c +9 -0
+static void scmi_cpufreq_register_em(struct cpufreq_policy *policy)
+{
+	struct em_data_callback em_cb = EM_DATA_CB(scmi_get_cpu_power);
+	bool power_scale_mw = perf_ops->power_scale_mw_get(ph);
+	struct scmi_data *priv = policy->driver_data;
+
+	em_dev_register_perf_domain(get_cpu_device(policy->cpu), priv->nr_opp,
+				    &em_cb, priv->opp_shared_cpus,
+				    power_scale_mw);

I would free the priv->opp_shared_cpus mask here, since we don't need it anymore and memory can be reclaimed.

Yes, we don't need it anymore, but this isn't a good place to undo what init() has done. Moreover, it is possible that register_em() may not get called at all, if some error has occurred after init() has successfully returned. It is always better to use exit() for such things. It won't hurt a lot to keep this around anyway.

Quoted 2 lines

Don't forget this
setup would be called N CPUs times, on this per-CPU policy platform.

Yes, but EM will just ignore this call. Though I have made a change here now to check for non-zero nr_opp to avoid the unnecessary call.

Quoted 2 lines

If freed here, then also there wouldn't be a need to free it in scmi_cpufreq_exit() so you can remove it from there.

Signature · 1 line
viresh