cpufreq_cooling: Get effective CPU utilization from scheduler

There is nothing schedutil specific in schedutil_cpu_util(), move it to core.c and rename it to sched_cpu_util(), so it can be used from other parts of the kernel as well.

The cpufreq_cooling stuff will make use of this in a later commit.

Signed-off-by Viresh Kumar <[email protected]>

---
 include/linux/sched.h            |  21 ++++++
 kernel/sched/core.c              | 115 ++++++++++++++++++++++++++++++
 kernel/sched/cpufreq_schedutil.c | 116 +------------------------------
 kernel/sched/fair.c              |   6 +-
 kernel/sched/sched.h             |  31 +--------
 5 files changed, 145 insertions(+), 144 deletions(-)

include/linux/sched.h +21 -0

index 063cd120b459..926b944dae5e 100644
@@ -1926,6 +1926,27 @@ extern long sched_getaffinity(pid_t pid, struct cpumask *mask);
 #define TASK_SIZE_OF(tsk)      TASK_SIZE
 #endif
 
+#ifdef CONFIG_SMP
+/**
+ * enum cpu_util_type - CPU utilization type
+ * @FREQUENCY_UTIL:    Utilization used to select frequency
+ * @ENERGY_UTIL:       Utilization used during energy calculation
+ *
+ * The utilization signals of all scheduling classes (CFS/RT/DL) and IRQ time
+ * need to be aggregated differently depending on the usage made of them. This
+ * enum is used within sched_cpu_util() to differentiate the types of
+ * utilization expected by the callers, and adjust the aggregation accordingly.
+ */
+enum cpu_util_type {
+       FREQUENCY_UTIL,
+       ENERGY_UTIL,
+};
+
+/* Returns effective CPU utilization, as seen by the scheduler */
+unsigned long sched_cpu_util(int cpu, enum cpu_util_type type,
+                            unsigned long max);
+#endif /* CONFIG_SMP */
+
 #ifdef CONFIG_RSEQ
 
 /*

kernel/sched/core.c +115 -0

index d2003a7d5ab5..845c976ccd53 100644
@@ -5117,6 +5117,121 @@ struct task_struct *idle_task(int cpu)
        return cpu_rq(cpu)->idle;
 }
 
+#ifdef CONFIG_SMP
+/*
+ * This function computes an effective utilization for the given CPU, to be
+ * used for frequency selection given the linear relation: f = u * f_max.
+ *
+ * The scheduler tracks the following metrics:
+ *
+ *   cpu_util_{cfs,rt,dl,irq}()
+ *   cpu_bw_dl()
+ *
+ * Where the cfs,rt and dl util numbers are tracked with the same metric and
+ * synchronized windows and are thus directly comparable.
+ *
+ * The cfs,rt,dl utilization are the running times measured with rq->clock_task
+ * which excludes things like IRQ and steal-time. These latter are then accrued
+ * in the irq utilization.
+ *
+ * The DL bandwidth number otoh is not a measured metric but a value computed
+ * based on the task model parameters and gives the minimal utilization
+ * required to meet deadlines.
+ */
+unsigned long effective_cpu_util(int cpu, unsigned long util_cfs,
+                                unsigned long max, enum cpu_util_type type,
+                                struct task_struct *p)
+{
+       unsigned long dl_util, util, irq;
+       struct rq *rq = cpu_rq(cpu);
+
+       if (!uclamp_is_used() &&
+           type == FREQUENCY_UTIL && rt_rq_is_runnable(&rq->rt)) {
+               return max;
+       }
+
+       /*
+        * Early check to see if IRQ/steal time saturates the CPU, can be
+        * because of inaccuracies in how we track these -- see
+        * update_irq_load_avg().
+        */
+       irq = cpu_util_irq(rq);
+       if (unlikely(irq >= max))
+               return max;
+
+       /*
+        * Because the time spend on RT/DL tasks is visible as 'lost' time to
+        * CFS tasks and we use the same metric to track the effective
+        * utilization (PELT windows are synchronized) we can directly add them
+        * to obtain the CPU's actual utilization.
+        *
+        * CFS and RT utilization can be boosted or capped, depending on
+        * utilization clamp constraints requested by currently RUNNABLE
+        * tasks.
+        * When there are no CFS RUNNABLE tasks, clamps are released and
+        * frequency will be gracefully reduced with the utilization decay.
+        */
+       util = util_cfs + cpu_util_rt(rq);
+       if (type == FREQUENCY_UTIL)
+               util = uclamp_rq_util_with(rq, util, p);
+
+       dl_util = cpu_util_dl(rq);
+
+       /*
+        * For frequency selection we do not make cpu_util_dl() a permanent part
+        * of this sum because we want to use cpu_bw_dl() later on, but we need
+        * to check if the CFS+RT+DL sum is saturated (ie. no idle time) such
+        * that we select f_max when there is no idle time.
+        *
+        * NOTE: numerical errors or stop class might cause us to not quite hit
+        * saturation when we should -- something for later.
+        */
+       if (util + dl_util >= max)
+               return max;
+
+       /*
+        * OTOH, for energy computation we need the estimated running time, so
+        * include util_dl and ignore dl_bw.
+        */
+       if (type == ENERGY_UTIL)
+               util += dl_util;
+
+       /*
+        * There is still idle time; further improve the number by using the
+        * irq metric. Because IRQ/steal time is hidden from the task clock we
+        * need to scale the task numbers:
+        *
+        *              max - irq
+        *   U' = irq + --------- * U
+        *                 max
+        */
+       util = scale_irq_capacity(util, irq, max);
+       util += irq;
+
+       /*
+        * Bandwidth required by DEADLINE must always be granted while, for
+        * FAIR and RT, we use blocked utilization of IDLE CPUs as a mechanism
+        * to gracefully reduce the frequency when no tasks show up for longer
+        * periods of time.
+        *
+        * Ideally we would like to set bw_dl as min/guaranteed freq and util +
+        * bw_dl as requested freq. However, cpufreq is not yet ready for such
+        * an interface. So, we only do the latter for now.
+        */
+       if (type == FREQUENCY_UTIL)
+               util += cpu_bw_dl(rq);
+
+       return min(max, util);
+}
+
+unsigned long sched_cpu_util(int cpu, enum cpu_util_type type,
+                            unsigned long max)
+{
+       return effective_cpu_util(cpu, cpu_util_cfs(cpu_rq(cpu)), max, type,
+                                 NULL);
+}
+#endif /* CONFIG_SMP */
+
 /**
  * find_process_by_pid - find a process with a matching PID value.
  * @pid: the pid in question.

kernel/sched/cpufreq_schedutil.c +0 -107

index e254745a82cb..a6de75c8b984 100644
@@ -169,122 +169,12 @@ static unsigned int get_next_freq(struct sugov_policy *sg_policy,
        return cpufreq_driver_resolve_freq(policy, freq);
 }
 
-/*
- * This function computes an effective utilization for the given CPU, to be
- * used for frequency selection given the linear relation: f = u * f_max.
- *
- * The scheduler tracks the following metrics:
- *
- *   cpu_util_{cfs,rt,dl,irq}()
- *   cpu_bw_dl()
- *
- * Where the cfs,rt and dl util numbers are tracked with the same metric and
- * synchronized windows and are thus directly comparable.
- *
- * The cfs,rt,dl utilization are the running times measured with rq->clock_task
- * which excludes things like IRQ and steal-time. These latter are then accrued
- * in the irq utilization.
- *
- * The DL bandwidth number otoh is not a measured metric but a value computed
- * based on the task model parameters and gives the minimal utilization
- * required to meet deadlines.
- */
-unsigned long schedutil_cpu_util(int cpu, unsigned long util_cfs,
-                                unsigned long max, enum schedutil_type type,
-                                struct task_struct *p)
-{
-       unsigned long dl_util, util, irq;
-       struct rq *rq = cpu_rq(cpu);
-
-       if (!uclamp_is_used() &&
-           type == FREQUENCY_UTIL && rt_rq_is_runnable(&rq->rt)) {
-               return max;
-       }
-
-       /*
-        * Early check to see if IRQ/steal time saturates the CPU, can be
-        * because of inaccuracies in how we track these -- see
-        * update_irq_load_avg().
-        */
-       irq = cpu_util_irq(rq);
-       if (unlikely(irq >= max))
-               return max;
-
-       /*
-        * Because the time spend on RT/DL tasks is visible as 'lost' time to
-        * CFS tasks and we use the same metric to track the effective
-        * utilization (PELT windows are synchronized) we can directly add them
-        * to obtain the CPU's actual utilization.
-        *
-        * CFS and RT utilization can be boosted or capped, depending on
-        * utilization clamp constraints requested by currently RUNNABLE
-        * tasks.
-        * When there are no CFS RUNNABLE tasks, clamps are released and
-        * frequency will be gracefully reduced with the utilization decay.
-        */
-       util = util_cfs + cpu_util_rt(rq);
-       if (type == FREQUENCY_UTIL)
-               util = uclamp_rq_util_with(rq, util, p);
-
-       dl_util = cpu_util_dl(rq);
-
-       /*
-        * For frequency selection we do not make cpu_util_dl() a permanent part
-        * of this sum because we want to use cpu_bw_dl() later on, but we need
-        * to check if the CFS+RT+DL sum is saturated (ie. no idle time) such
-        * that we select f_max when there is no idle time.
-        *
-        * NOTE: numerical errors or stop class might cause us to not quite hit
-        * saturation when we should -- something for later.
-        */
-       if (util + dl_util >= max)
-               return max;
-
-       /*
-        * OTOH, for energy computation we need the estimated running time, so
-        * include util_dl and ignore dl_bw.
-        */
-       if (type == ENERGY_UTIL)
-               util += dl_util;
-
-       /*
-        * There is still idle time; further improve the number by using the
-        * irq metric. Because IRQ/steal time is hidden from the task clock we
-        * need to scale the task numbers:
-        *
-        *              max - irq
-        *   U' = irq + --------- * U
-        *                 max
-        */
-       util = scale_irq_capacity(util, irq, max);
-       util += irq;
-
-       /*
-        * Bandwidth required by DEADLINE must always be granted while, for
-        * FAIR and RT, we use blocked utilization of IDLE CPUs as a mechanism
-        * to gracefully reduce the frequency when no tasks show up for longer
-        * periods of time.
-        *
-        * Ideally we would like to set bw_dl as min/guaranteed freq and util +
-        * bw_dl as requested freq. However, cpufreq is not yet ready for such
-        * an interface. So, we only do the latter for now.
-        */
-       if (type == FREQUENCY_UTIL)
-               util += cpu_bw_dl(rq);
-
-       return min(max, util);
-}
-
 static unsigned long sugov_get_util(struct sugov_cpu *sg_cpu)
 {
-       struct rq *rq = cpu_rq(sg_cpu->cpu);

kernel/sched/cpufreq_schedutil.c +0 -0

 }
 
 /**

Quoted 4 lines · Viresh Kumar

There is nothing schedutil specific in schedutil_cpu_util(), move it to core.c and rename it to sched_cpu_util(), so it can be used from other parts of the kernel as well.

The patch does more than this, though.

I would do that in two patches: (1) move the function as is and (2) rename it and rearrange the users.

Quoted 11 lines

static unsigned long sugov_get_util(struct sugov_cpu *sg_cpu) {

kernel/sched/cpufreq_schedutil.c +2 -6

-       struct rq *rq = cpu_rq(sg_cpu->cpu);
-       unsigned long util = cpu_util_cfs(rq);
-       unsigned long max = arch_scale_cpu_capacity(sg_cpu->cpu);
-
-       sg_cpu->max = max;
-       sg_cpu->bw_dl = cpu_bw_dl(rq);
+       sg_cpu->max = arch_scale_cpu_capacity(sg_cpu->cpu);
+       sg_cpu->bw_dl = cpu_bw_dl(cpu_rq(sg_cpu->cpu));

kernel/sched/cpufreq_schedutil.c static unsigned int get_next_freq(struct sugov_policy *sg_policy, · line 177

Quoted 2 lines

kernel/sched/cpufreq_schedutil.c +1 -1

-       return schedutil_cpu_util(sg_cpu->cpu, util, max, FREQUENCY_UTIL, NULL);
+       return sched_cpu_util(sg_cpu->cpu, FREQUENCY_UTIL, sg_cpu->max);

I don't see much value in using this wrapper here TBH and it introduces an otherwise redundant cpu_rq() computation.

Quoted 10 lines · Rafael J. Wysocki

Quoted 4 lines · Viresh Kumar

There is nothing schedutil specific in schedutil_cpu_util(), move it to core.c and rename it to sched_cpu_util(), so it can be used from other parts of the kernel as well.

The patch does more than this, though.

I would do that in two patches: (1) move the function as is and (2) rename it and rearrange the users.

Sure.

Quoted 11 lines

Quoted 11 lines

static unsigned long sugov_get_util(struct sugov_cpu *sg_cpu) {

kernel/sched/cpufreq_schedutil.c +2 -6

-       struct rq *rq = cpu_rq(sg_cpu->cpu);
-       unsigned long util = cpu_util_cfs(rq);
-       unsigned long max = arch_scale_cpu_capacity(sg_cpu->cpu);
-
-       sg_cpu->max = max;
-       sg_cpu->bw_dl = cpu_bw_dl(rq);
+       sg_cpu->max = arch_scale_cpu_capacity(sg_cpu->cpu);
+       sg_cpu->bw_dl = cpu_bw_dl(cpu_rq(sg_cpu->cpu));

Quoted 5 lines

kernel/sched/cpufreq_schedutil.c static unsigned int get_next_freq(struct sugov_policy *sg_policy, · line 177

Quoted 2 lines

kernel/sched/cpufreq_schedutil.c +1 -1

-       return schedutil_cpu_util(sg_cpu->cpu, util, max, FREQUENCY_UTIL, NULL);
+       return sched_cpu_util(sg_cpu->cpu, FREQUENCY_UTIL, sg_cpu->max);

I don't see much value in using this wrapper here TBH and it introduces an otherwise redundant cpu_rq() computation.

You want to call effective_cpu_util() here instead, right ?

Several parts of the kernel are already using the effective CPU utilization (as seen by the scheduler) to get the current load on the CPU, do the same here instead of depending on the idle time of the CPU, which isn't that accurate comparatively.

This is also the right thing to do as it makes the cpufreq governor (schedutil) align better with the cpufreq_cooling driver, as the power requested by cpufreq_cooling governor will exactly match the next frequency requested by the schedutil governor since they are both using the same metric to calculate load.

Note that, this (and CPU frequency scaling in general) doesn't work that well with idle injection as that is done from rt threads and is counted as load while it tries to do quite the opposite. That should be solved separately though.

This was tested on ARM Hikey6220 platform with hackbench, sysbench and schbench. None of them showed any regression or significant improvements. Schbench is the most important ones out of these as it creates the scenario where the utilization numbers provide a better estimate of the future.

Scenario 1: The CPUs were mostly idle in the previous polling window of the IPA governor as the tasks were sleeping and here are the details from traces (load is in %):

Old: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=203 load={{0x35,0x1,0x0,0x31,0x0,0x0,0x64,0x0}} dynamic_power=1339
New: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=600 load={{0x60,0x46,0x45,0x45,0x48,0x3b,0x61,0x44}} dynamic_power=3960

Here, the "Old" line gives the load and requested_power (dynamic_power here) numbers calculated using the idle time based implementation, while "New" is based on the CPU utilization from scheduler.

As can be clearly seen, the load and requested_power numbers are simply incorrect in the idle time based approach and the numbers collected from CPU's utilization are much closer to the reality.

Scenario 2: The CPUs were busy in the previous polling window of the IPA governor:

Old: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=800 load={{0x64,0x64,0x64,0x64,0x64,0x64,0x64,0x64}} dynamic_power=5280
New: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=708 load={{0x4d,0x5c,0x5c,0x5b,0x5c,0x5c,0x51,0x5b}} dynamic_power=4672

As can be seen, the idle time based load is 100% for all the CPUs as it took only the last window into account, but in reality the CPUs aren't that loaded as shown by the utilization numbers.

Quoted 11 lines · Viresh Kumar

Scenario 1: The CPUs were mostly idle in the previous polling window of the IPA governor as the tasks were sleeping and here are the details from traces (load is in %):

Old: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=203 load={{0x35,0x1,0x0,0x31,0x0,0x0,0x64,0x0}} dynamic_power=1339
New: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=600 load={{0x60,0x46,0x45,0x45,0x48,0x3b,0x61,0x44}} dynamic_power=3960

Here, the "Old" line gives the load and requested_power (dynamic_power here) numbers calculated using the idle time based implementation, while "New" is based on the CPU utilization from scheduler.

Quoted 3 lines

As can be clearly seen, the load and requested_power numbers are simply incorrect in the idle time based approach and the numbers collected from CPU's utilization are much closer to the reality.

It is contradicting to what you have put in 'Scenario 1' description, isn't it?

Frequency at 1.2GHz, 75% total_load, power 4W... I'd say if CPUs were mostly idle than 1.3W would better reflect that state.

What was the IPA period in your setup?

It depends on your platform IPA period (e.g. 100ms) and your current runqueues state (at that sampling point in time). The PELT decay/rise period is different. I am not sure if you observe the system avg load for last e.g. 100ms looking at these signals. Maybe IPA period is too short/long and couldn't catch up with PELT signals?
But we won't too short averaging, since 16ms is a display tick.

IMHO based on this result it looks like the util could lost older information from the past or didn't converge yet to this low load yet.

Quoted 7 lines

Scenario 2: The CPUs were busy in the previous polling window of the IPA governor:

Old: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=800 load={{0x64,0x64,0x64,0x64,0x64,0x64,0x64,0x64}} dynamic_power=5280
New: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=708 load={{0x4d,0x5c,0x5c,0x5b,0x5c,0x5c,0x51,0x5b}} dynamic_power=4672

Quoted 3 lines

As can be seen, the idle time based load is 100% for all the CPUs as it took only the last window into account, but in reality the CPUs aren't that loaded as shown by the utilization numbers.

This is also odd. The ~88% of total_load, looks like started decaying or didn't converge yet to 100% or some task vanished?

Quoted 12 lines · Lukasz Luba

Quoted 11 lines · Viresh Kumar

Scenario 1: The CPUs were mostly idle in the previous polling window of the IPA governor as the tasks were sleeping and here are the details from traces (load is in %):

   Old: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=203 load={{0x35,0x1,0x0,0x31,0x0,0x0,0x64,0x0}} dynamic_power=1339
   New: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=600 load={{0x60,0x46,0x45,0x45,0x48,0x3b,0x61,0x44}} dynamic_power=3960

Here, the "Old" line gives the load and requested_power (dynamic_power here) numbers calculated using the idle time based implementation, while "New" is based on the CPU utilization from scheduler.

Quoted 6 lines

Quoted 3 lines

As can be clearly seen, the load and requested_power numbers are simply incorrect in the idle time based approach and the numbers collected from CPU's utilization are much closer to the reality.

It is contradicting to what you have put in 'Scenario 1' description, isn't it?

At least I didn't think so when I wrote this and am still not sure :)

Quoted 2 lines

Frequency at 1.2GHz, 75% total_load, power 4W... I'd say if CPUs were mostly idle than 1.3W would better reflect that state.

The CPUs were idle because the tasks were sleeping, but once the tasks resume to work, we need a frequency that matches the real load of the tasks. This is exactly what schedutil would ask for as well as it uses the same metric and so we should be looking to ask for the same power budget..

Quoted 2 lines

What was the IPA period in your setup?

It is 100 ms by default, though I remember that I tried with 10 ms as well.

Quoted 21 lines

It depends on your platform IPA period (e.g. 100ms) and your current runqueues state (at that sampling point in time). The PELT decay/rise period is different. I am not sure if you observe the system avg load for last e.g. 100ms looking at these signals. Maybe IPA period is too short/long and couldn't catch up with PELT signals?
But we won't too short averaging, since 16ms is a display tick.

IMHO based on this result it looks like the util could lost older information from the past or didn't converge yet to this low load yet.

Quoted 10 lines

Scenario 2: The CPUs were busy in the previous polling window of the IPA governor:

   Old: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=800 load={{0x64,0x64,0x64,0x64,0x64,0x64,0x64,0x64}} dynamic_power=5280
   New: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=708 load={{0x4d,0x5c,0x5c,0x5b,0x5c,0x5c,0x51,0x5b}} dynamic_power=4672

As can be seen, the idle time based load is 100% for all the CPUs as it took only the last window into account, but in reality the CPUs aren't that loaded as shown by the utilization numbers.

Quoted 2 lines

This is also odd. The ~88% of total_load, looks like started decaying or didn't converge yet to 100% or some task vanished?

They must have decayed a bit because of the idle period, so looks okay that way.

Several parts of the kernel are already using the effective CPU utilization (as seen by the scheduler) to get the current load on the CPU, do the same here instead of depending on the idle time of the CPU, which isn't that accurate comparatively.

This is also the right thing to do as it makes the cpufreq governor (schedutil) align better with the cpufreq_cooling driver, as the power requested by cpufreq_cooling governor will exactly match the next frequency requested by the schedutil governor since they are both using the same metric to calculate load.

Note that, this (and CPU frequency scaling in general) doesn't work that well with idle injection as that is done from rt threads and is counted as load while it tries to do quite the opposite. That should be solved separately though.

This was tested on ARM Hikey6220 platform with hackbench, sysbench and schbench. None of them showed any regression or significant improvements. Schbench is the most important ones out of these as it creates the scenario where the utilization numbers provide a better estimate of the future.

Scenario 1: The CPUs were mostly idle in the previous polling window of the IPA governor as the tasks were sleeping and here are the details from traces (load is in %):

Old: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=203 load={{0x35,0x1,0x0,0x31,0x0,0x0,0x64,0x0}} dynamic_power=1339
New: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=600 load={{0x60,0x46,0x45,0x45,0x48,0x3b,0x61,0x44}} dynamic_power=3960

Here, the "Old" line gives the load and requested_power (dynamic_power here) numbers calculated using the idle time based implementation, while "New" is based on the CPU utilization from scheduler.

As can be clearly seen, the load and requested_power numbers are simply incorrect in the idle time based approach and the numbers collected from CPU's utilization are much closer to the reality.

Scenario 2: The CPUs were busy in the previous polling window of the IPA governor:

Old: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=800 load={{0x64,0x64,0x64,0x64,0x64,0x64,0x64,0x64}} dynamic_power=5280
New: thermal_power_cpu_get_power: cpus=00000000,000000ff freq=1200000 total_load=708 load={{0x4d,0x5c,0x5c,0x5b,0x5c,0x5c,0x51,0x5b}} dynamic_power=4672

As can be seen, the idle time based load is 100% for all the CPUs as it took only the last window into account, but in reality the CPUs aren't that loaded as shown by the utilization numbers.

Signed-off-by Viresh Kumar <[email protected]>

---
  drivers/thermal/cpufreq_cooling.c | 68 ++++++++++++++++++++++++-------
  1 file changed, 54 insertions(+), 14 deletions(-)

drivers/thermal/cpufreq_cooling.c +53 -13

index cc2959f22f01..a364a2fd84b1 100644
@@ -76,7 +76,9 @@ struct cpufreq_cooling_device {
  	struct em_perf_domain *em;
  	struct cpufreq_policy *policy;
  	struct list_head node;
+#ifndef CONFIG_SMP
  	struct time_in_idle *idle_time;
+#endif
  	struct freq_qos_request qos_req;
  };
 

@@ -132,14 +134,35 @@ static u32 cpu_power_to_freq(struct cpufreq_cooling_device *cpufreq_cdev,
  }
 
  /**
- * get_load() - get load for a cpu since last updated
- * @cpufreq_cdev:	&struct cpufreq_cooling_device for this cpu
- * @cpu:	cpu number
- * @cpu_idx:	index of the cpu in time_in_idle*
+ * get_load() - get load for a cpu
+ * @cpufreq_cdev: struct cpufreq_cooling_device for the cpu
+ * @cpu: cpu number
+ * @cpu_idx: index of the cpu in time_in_idle array
   *
   * Return: The average load of cpu @cpu in percentage since this
   * function was last called.
   */
+#ifdef CONFIG_SMP
+static u32 get_load(struct cpufreq_cooling_device *cpufreq_cdev, int cpu,
+		    int cpu_idx)
+{
+	unsigned long max = arch_scale_cpu_capacity(cpu);
+	unsigned long util;
+
+	util = sched_cpu_util(cpu, ENERGY_UTIL, max);
+	return (util * 100) / max;
+}
+
+static inline int allocate_idle_time(struct cpufreq_cooling_device *cpufreq_cdev)
+{
+	return 0;
+}
+
+static inline void free_idle_time(struct cpufreq_cooling_device *cpufreq_cdev)
+{
+}
+
+#else /* !CONFIG_SMP */
  static u32 get_load(struct cpufreq_cooling_device *cpufreq_cdev, int cpu,
  		    int cpu_idx)
  {

@@ -162,6 +185,26 @@ static u32 get_load(struct cpufreq_cooling_device *cpufreq_cdev, int cpu,
  	return load;
  }
 
+static int allocate_idle_time(struct cpufreq_cooling_device *cpufreq_cdev)
+{
+	unsigned int num_cpus = cpumask_weight(cpufreq_cdev->policy->related_cpus);
+
+	cpufreq_cdev->idle_time = kcalloc(num_cpus,
+					 sizeof(*cpufreq_cdev->idle_time),
+					 GFP_KERNEL);
+	if (!cpufreq_cdev->idle_time)
+		return -ENOMEM;
+
+	return 0;
+}
+
+static void free_idle_time(struct cpufreq_cooling_device *cpufreq_cdev)
+{
+	kfree(cpufreq_cdev->idle_time);
+	cpufreq_cdev->idle_time = NULL;
+}
+#endif /* CONFIG_SMP */
+
  /**
   * get_dynamic_power() - calculate the dynamic power
   * @cpufreq_cdev:	&cpufreq_cooling_device for this cdev

@@ -487,7 +530,7 @@ __cpufreq_cooling_register(struct device_node *np,
  	struct thermal_cooling_device *cdev;
  	struct cpufreq_cooling_device *cpufreq_cdev;
  	char dev_name[THERMAL_NAME_LENGTH];
-	unsigned int i, num_cpus;
+	unsigned int i;
  	struct device *dev;
  	int ret;
  	struct thermal_cooling_device_ops *cooling_ops;

@@ -498,7 +541,6 @@ __cpufreq_cooling_register(struct device_node *np,
  		return ERR_PTR(-ENODEV);
  	}
 
-
  	if (IS_ERR_OR_NULL(policy)) {
  		pr_err("%s: cpufreq policy isn't valid: %p\n", __func__, policy);
  		return ERR_PTR(-EINVAL);

@@ -516,12 +558,10 @@ __cpufreq_cooling_register(struct device_node *np,
  		return ERR_PTR(-ENOMEM);
 
  	cpufreq_cdev->policy = policy;
-	num_cpus = cpumask_weight(policy->related_cpus);
-	cpufreq_cdev->idle_time = kcalloc(num_cpus,
-					 sizeof(*cpufreq_cdev->idle_time),
-					 GFP_KERNEL);
-	if (!cpufreq_cdev->idle_time) {
-		cdev = ERR_PTR(-ENOMEM);
+
+	ret = allocate_idle_time(cpufreq_cdev);
+	if (ret) {
+		cdev = ERR_PTR(ret);
  		goto free_cdev;
  	}
 

@@ -581,7 +621,7 @@ __cpufreq_cooling_register(struct device_node *np,
  remove_ida:
  	ida_simple_remove(&cpufreq_ida, cpufreq_cdev->id);
  free_idle_time:
-	kfree(cpufreq_cdev->idle_time);
+	free_idle_time(cpufreq_cdev);
  free_cdev:
  	kfree(cpufreq_cdev);
  	return cdev;

LGTM. It has potential. We will see how far we can improve IPA with this model.

Reviewed-by Lukasz Luba <[email protected]>

As can be clearly seen, the load and requested_power numbers are simply incorrect in the idle time based approach and the numbers collected from CPU's utilization are much closer to the reality.

PELT time-scaling can make the util signals behave strangely from an external PoV. For instance, on a big.LITTLE system, the rq util of a LITTLE CPU may suddenly drop if it was stuck on a too-low OPP for some time and eventually reached the "right" OPP (i.e. got idle time).

Also, as Peter pointed out in [1], task migrations can easily confuse an external observer that considers util to be "an image of the recent past".