[RFC PATCH v1 0/2] Add platform specific callbacks for cpu cooling

[RFC PATCH v1 0/2] Add platform specific callbacks for cpu cooling

[zhang.lyra]

From: Jeson Gao <jeson.gao@unisoc.com>

Reducing frequency to too low will affect the system performance, and
sometimes the temperature could not cool down to an ideal value as well
even if the frequency was reduced to the lowest.

This patch provides users a callback to set its own frequency limit to
prevent cpu frequence from being reduced too low. And also give an interface
to allow users to register specific cooling operations, such as unplug cores
or cpu isolation.

Any comments are very appreciated.

jeson.gao (2):
  thermal/drivers/cpufreq_cooling: Add platform callback functions
  thermal/drivers/sprd_cpu_cooling: Add platform mitigation thermal
    driver

 drivers/thermal/Kconfig            |   8 +
 drivers/thermal/Makefile           |   1 +
 drivers/thermal/cpufreq_cooling.c  |  53 +++++
 drivers/thermal/sprd_cpu_cooling.c | 340 +++++++++++++++++++++++++++++
 include/linux/cpu_cooling.h        |  30 +++
 5 files changed, 432 insertions(+)
 create mode 100644 drivers/thermal/sprd_cpu_cooling.c

-- 
2.20.1

[RFC PATCH v1 1/2] thermal/drivers/cpufreq_cooling: Add platform callback functions

[zhang.lyra]

From: Jeson Gao <jeson.gao@unisoc.com>

Reducing frequency to too low will affect the system performance, and
sometimes the temperature could not cool down to an ideal value as well
even if the frequency was reduced to the lowest.

This patch provides users a callback to set its own frequency limit to
prevent cpu frequence from being reduced too low. And also give an interface
to allow users to register specific cooling operations.

Signed-off-by: Jeson Gao <jeson.gao@unisoc.com>
Signed-off-by: Chunyan Zhang <zhang.lyra@gmail.com>
---
 drivers/thermal/cpufreq_cooling.c | 53 +++++++++++++++++++++++++++++++
 include/linux/cpu_cooling.h       | 30 +++++++++++++++++
 2 files changed, 83 insertions(+)

diff --git a/drivers/thermal/cpufreq_cooling.c b/drivers/thermal/cpufreq_cooling.c
index e297e135c031..da1cbb22c975 100644
--- a/drivers/thermal/cpufreq_cooling.c
+++ b/drivers/thermal/cpufreq_cooling.c
@@ -64,6 +64,7 @@ struct time_in_idle {
  * @node: list_head to link all cpufreq_cooling_device together.
  * @idle_time: idle time stats
  * @qos_req: PM QoS contraint to apply
+ * @plat_ops: point to platform callback function.
  *
  * This structure is required for keeping information of each registered
  * cpufreq_cooling_device.
@@ -78,6 +79,7 @@ struct cpufreq_cooling_device {
 	struct list_head node;
 	struct time_in_idle *idle_time;
 	struct freq_qos_request qos_req;
+	struct cpufreq_cooling_plat_ops *plat_ops;
 };
 
 static DEFINE_IDA(cpufreq_ida);
@@ -313,12 +315,24 @@ static int cpufreq_power2state(struct thermal_cooling_device *cdev,
 	u32 last_load, normalised_power;
 	struct cpufreq_cooling_device *cpufreq_cdev = cdev->devdata;
 	struct cpufreq_policy *policy = cpufreq_cdev->policy;
+	struct cpufreq_cooling_plat_ops *plat_ops = cpufreq_cdev->plat_ops;
 
 	last_load = cpufreq_cdev->last_load ?: 1;
 	normalised_power = (power * 100) / last_load;
 	target_freq = cpu_power_to_freq(cpufreq_cdev, normalised_power);
 
 	*state = get_level(cpufreq_cdev, target_freq);
+	if (*state == cpufreq_cdev->max_level &&
+			plat_ops && plat_ops->cpufreq_plat_min_freq_limit) {
+		plat_ops->cpufreq_plat_min_freq_limit(policy, &target_freq);
+		*state = get_level(cpufreq_cdev, target_freq);
+	}
+
+	if (plat_ops && plat_ops->cpufreq_plat_cpu_ctrl)
+		plat_ops->cpufreq_plat_cpu_ctrl(policy,
+				last_load, normalised_power,
+				cpu_freq_to_power(cpufreq_cdev, target_freq));
+
 	trace_thermal_power_cpu_limit(policy->related_cpus, target_freq, *state,
 				      power);
 	return 0;
@@ -684,3 +698,42 @@ void cpufreq_cooling_unregister(struct thermal_cooling_device *cdev)
 	kfree(cpufreq_cdev);
 }
 EXPORT_SYMBOL_GPL(cpufreq_cooling_unregister);
+
+/**
+ * cpufreq_cooling_plat_ops_register - register platform callback function.
+ * @cdev: thermal cooling device pointer.
+ * @plat_ops: platform callback function pointer.
+ */
+int  cpufreq_cooling_plat_ops_register(struct thermal_cooling_device *cdev,
+			struct cpufreq_cooling_plat_ops *plat_ops)
+{
+	struct cpufreq_cooling_device *cpufreq_cdev;
+
+	if (!cdev && !cdev->devdata && !plat_ops)
+		return -EINVAL;
+
+	cpufreq_cdev = cdev->devdata;
+	cpufreq_cdev->plat_ops = plat_ops;
+
+	return 0;
+
+}
+EXPORT_SYMBOL_GPL(cpufreq_cooling_plat_ops_register);
+
+/**
+ * cpufreq_cooling_plat_ops_unregister - unregister platform callback function.
+ * @cdev: thermal cooling device pointer.
+ */
+int  cpufreq_cooling_plat_ops_unregister(struct thermal_cooling_device *cdev)
+{
+	struct cpufreq_cooling_device *cpufreq_cdev;
+
+	if (!cdev && !cdev->devdata)
+		return -EINVAL;
+
+	cpufreq_cdev = cdev->devdata;
+	cpufreq_cdev->plat_ops = NULL;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(cpufreq_cooling_plat_ops_unregister);
diff --git a/include/linux/cpu_cooling.h b/include/linux/cpu_cooling.h
index 65501d8f9778..3934918e4ab7 100644
--- a/include/linux/cpu_cooling.h
+++ b/include/linux/cpu_cooling.h
@@ -19,6 +19,23 @@
 
 struct cpufreq_policy;
 
+/**
+ * struct cpufreq_cooling_plat_ops - platfom cpu cooling policy ops
+ *
+ * @cpufreq_plat_cpu_ctrl: this function provides a further core control
+ * policy when the current policies cannot cool down to an expected
+ * temperature value.
+ *
+ * @cpufreq_plat_min_freq_limit: set cpu frequency limit, cooling devices
+ * are not allowed to adjust cpu frequency to out of that limit.
+ */
+struct cpufreq_cooling_plat_ops {
+	int (*cpufreq_plat_cpu_ctrl)(struct cpufreq_policy *policy,
+				u32 load, u32 normalised_power, u32 freq_power);
+	void (*cpufreq_plat_min_freq_limit)(struct cpufreq_policy *policy,
+						u32 *target_freq);
+};
+
 #ifdef CONFIG_CPU_FREQ_THERMAL
 /**
  * cpufreq_cooling_register - function to create cpufreq cooling device.
@@ -40,6 +57,19 @@ void cpufreq_cooling_unregister(struct thermal_cooling_device *cdev);
 struct thermal_cooling_device *
 of_cpufreq_cooling_register(struct cpufreq_policy *policy);
 
+/**
+ * cpufreq_cooling_plat_ops_register - register platform callback function.
+ * @cdev: thermal cooling device pointer.
+ * @plat_ops: platform callback function pointer.
+ */
+int cpufreq_cooling_plat_ops_register(struct thermal_cooling_device *cdev,
+			struct cpufreq_cooling_plat_ops *plat_ops);
+/**
+ * cpufreq_cooling_plat_ops_unregister - unregister platform callback function.
+ * @cdev: thermal cooling device pointer.
+ */
+int  cpufreq_cooling_plat_ops_unregister(struct thermal_cooling_device *cdev);
+
 #else /* !CONFIG_CPU_FREQ_THERMAL */
 static inline struct thermal_cooling_device *
 cpufreq_cooling_register(struct cpufreq_policy *policy)
-- 
2.20.1

[RFC PATCH v1 2/2] thermal/drivers/sprd_cpu_cooling: Add platform mitigation thermal driver

[zhang.lyra]

From: Jeson Gao <jeson.gao@unisoc.com>

This driver provides a further core ctrl policy to reduce temperature for
unisoc platform.

CPU unplug or isolation usually would be used for core ctrl policy.

This dirver is showing an example of using core ctrl policy.

Signed-off-by: Jeson Gao <jeson.gao@unisoc.com>
Signed-off-by: Chunyan Zhang <zhang.lyra@gmail.com>
---
 drivers/thermal/Kconfig            |   8 +
 drivers/thermal/Makefile           |   1 +
 drivers/thermal/sprd_cpu_cooling.c | 340 +++++++++++++++++++++++++++++
 3 files changed, 349 insertions(+)
 create mode 100644 drivers/thermal/sprd_cpu_cooling.c

diff --git a/drivers/thermal/Kconfig b/drivers/thermal/Kconfig
index 91af271e9bb0..41a57b0a0d57 100644
--- a/drivers/thermal/Kconfig
+++ b/drivers/thermal/Kconfig
@@ -490,4 +490,12 @@ config SPRD_THERMAL
 	help
 	  Support for the Spreadtrum thermal sensor driver in the Linux thermal
 	  framework.
+
+config SPRD_CPU_COOLING
+	tristate "sprd cpu cooling support"
+	depends on CPU_FREQ_THERMAL
+	help
+	  This implements the sprd cpu cooling mechanism to mitigate temperature
+	  rising.
+
 endif
diff --git a/drivers/thermal/Makefile b/drivers/thermal/Makefile
index 8c8ed7b79915..403184299485 100644
--- a/drivers/thermal/Makefile
+++ b/drivers/thermal/Makefile
@@ -60,3 +60,4 @@ obj-$(CONFIG_ZX2967_THERMAL)	+= zx2967_thermal.o
 obj-$(CONFIG_UNIPHIER_THERMAL)	+= uniphier_thermal.o
 obj-$(CONFIG_AMLOGIC_THERMAL)     += amlogic_thermal.o
 obj-$(CONFIG_SPRD_THERMAL)	+= sprd_thermal.o
+obj-$(CONFIG_SPRD_CPU_COOLING)	+= sprd_cpu_cooling.o
diff --git a/drivers/thermal/sprd_cpu_cooling.c b/drivers/thermal/sprd_cpu_cooling.c
new file mode 100644
index 000000000000..18ce41e2bd5a
--- /dev/null
+++ b/drivers/thermal/sprd_cpu_cooling.c
@@ -0,0 +1,340 @@
+// SPDX-License-Identifier: GPL-2.0
+// Copyright (C) 2020 Spreadtrum Communications Inc.
+
+#include <linux/thermal.h>
+#include <linux/module.h>
+#include <linux/err.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/cpu.h>
+#include <linux/sched.h>
+#include <linux/platform_device.h>
+#include <linux/printk.h>
+#include <linux/cpufreq.h>
+
+#include <linux/cpu_cooling.h>
+
+struct sprd_cooling_cluster {
+	int id;
+	int cpu;
+	int cpus;
+	u32 min_freq;
+	struct cpufreq_policy policy;
+};
+
+struct sprd_cooling_data {
+	int nr_clusters;
+	struct sprd_cooling_cluster *cluster;
+};
+
+static void sprd_update_target_cpus(struct cpufreq_policy *policy,
+				    u32 target_cpus)
+{
+	int ret, cpu, first;
+	u32 curr_online_cpus, ncpus;
+
+	ncpus = cpumask_weight(policy->related_cpus);
+	curr_online_cpus = cpumask_weight(policy->cpus);
+	first = cpumask_first(policy->related_cpus);
+
+	if (target_cpus > curr_online_cpus) {
+		cpu = first;
+		for_each_cpu(cpu, policy->related_cpus) {
+			if (curr_online_cpus == target_cpus)
+				break;
+			if ((target_cpus > curr_online_cpus) &&
+			    !cpu_online(cpu)) {
+				ret = add_cpu(cpu);
+				if (!ret && cpu_online(cpu))
+					curr_online_cpus++;
+			}
+		}
+	} else if (target_cpus < curr_online_cpus) {
+		for (cpu = (first + ncpus - 1); cpu >= first; cpu--) {
+			if (curr_online_cpus == target_cpus)
+				break;
+			if ((target_cpus < curr_online_cpus) &&
+			    cpu_online(cpu)) {
+				ret = remove_cpu(cpu);
+				if (!ret && !cpu_online(cpu))
+					curr_online_cpus--;
+			}
+		}
+	}
+}
+
+static void sprd_estimate_down_cpus(struct cpufreq_policy *policy,
+				    u32 load, u32 normalised_power,
+				    u32 freq_power)
+{
+	int i = 0;
+	u32 target_cpus, online_cpus;
+	u32 avg_load, estimate_power =  0;
+
+	if (normalised_power) {
+		online_cpus = cpumask_weight(policy->cpus);
+		if (!online_cpus)
+			return;
+
+		avg_load = (load / online_cpus) ?: 1;
+		do {
+			estimate_power += (freq_power * avg_load) / 100;
+			if (estimate_power > normalised_power)
+				break;
+			i++;
+		} while (i < online_cpus);
+
+		target_cpus = max(i, 0);
+	} else
+		target_cpus = 0;
+
+	sprd_update_target_cpus(policy, target_cpus);
+}
+
+static void sprd_estimate_up_cpus(struct cpufreq_policy *policy,
+				  u32 load, u32 normalised_power,
+				  u32 freq_power)
+{
+	int i = 0, ncpus;
+	u32 target_cpus;
+	u32 avg_load, estimate_power =  0;
+
+	ncpus = cpumask_weight(policy->related_cpus);
+	avg_load = (load / ncpus) ?: 1;
+
+	do {
+		estimate_power += (freq_power * avg_load) / 100;
+		if (estimate_power > normalised_power)
+			break;
+		i++;
+	} while (i < ncpus);
+
+	target_cpus = min(i, ncpus);
+	sprd_update_target_cpus(policy, target_cpus);
+}
+
+static void sprd_keep_cpus(struct cpufreq_policy *policy)
+{
+	u32 online_cpus;
+
+	online_cpus = cpumask_weight(policy->cpus);
+	sprd_update_target_cpus(policy, online_cpus);
+}
+
+static int sprd_cpufreq_cpu_ctrl(struct cpufreq_policy *policy,
+				 u32 load, u32 normalised_power,
+				 u32 freq_power)
+{
+	unsigned int ncpus, online_cpus;
+
+	ncpus = cpumask_weight(policy->related_cpus);
+	online_cpus = cpumask_weight(policy->cpus);
+
+	if (normalised_power <  freq_power)
+		sprd_estimate_down_cpus(policy, load,
+					normalised_power, freq_power);
+	else if (online_cpus < ncpus)
+		sprd_estimate_up_cpus(policy, load,
+				      normalised_power, freq_power);
+	else
+		sprd_keep_cpus(policy);
+
+	return 0;
+}
+
+static void sprd_cpufreq_min_freq_limit(struct cpufreq_policy *policy,
+					u32 *target_freq)
+{
+	struct sprd_cooling_cluster *cluster =
+		container_of(policy, struct sprd_cooling_cluster, policy);
+
+	if (*target_freq < cluster->min_freq)
+		*target_freq = cluster->min_freq;
+}
+
+struct cpufreq_cooling_plat_ops plat_ops = {
+	.cpufreq_plat_cpu_ctrl = sprd_cpufreq_cpu_ctrl,
+	.cpufreq_plat_min_freq_limit = sprd_cpufreq_min_freq_limit,
+};
+
+static int sprd_get_cluster_counts(void)
+{
+	int cpu = 0, core_num = -1;
+	int cluster_num = 0;
+
+	do {
+		core_num = cpumask_weight(topology_core_cpumask(cpu));
+		if (core_num > 0) {
+			cpu = cpu + core_num;
+			cluster_num++;
+		} else
+			break;
+	} while (cpu > 0);
+
+	return cluster_num;
+}
+
+static int sprd_cpu_cooing_ops_register(struct platform_device *pdev)
+{
+	int id, ret = 0;
+	struct thermal_cooling_device *cdev;
+	struct sprd_cooling_cluster *cluster;
+	struct sprd_cooling_data *data = platform_get_drvdata(pdev);
+
+	for (id = 0; id < data->nr_clusters; id++) {
+		cluster = &data->cluster[id];
+		cdev = cluster->policy.cdev;
+		if (cdev) {
+			ret = cpufreq_cooling_plat_ops_register(cdev,
+								&plat_ops);
+			if (ret < 0) {
+				dev_err(&pdev->dev,
+					"CPU%d: failed to register platform function\n",
+					cluster->cpu);
+				break;
+			}
+		} else {
+			ret = -ENODEV;
+			dev_err(&pdev->dev,
+				"CPU%d: failed to get thermal device\n",
+				cluster->cpu);
+			break;
+		}
+	}
+
+	return ret;
+}
+
+static int sprd_cpu_cooing_ops_unregister(struct platform_device *pdev)
+{
+	int id, ret = 0;
+	struct thermal_cooling_device *cdev;
+	struct sprd_cooling_cluster *cluster;
+	struct sprd_cooling_data *data = platform_get_drvdata(pdev);
+
+	for (id = 0; id < data->nr_clusters; id++) {
+		cluster = &data->cluster[id];
+		cdev = cluster->policy.cdev;
+		if (cdev) {
+			ret = cpufreq_cooling_plat_ops_unregister(cdev);
+			if (ret < 0) {
+				dev_err(&pdev->dev,
+					"cpu%d: failed to unregister platform function\n",
+					cluster->cpu);
+				break;
+			}
+		} else {
+			dev_err(&pdev->dev,
+				"cpu%d: failed to get thermal device\n",
+				cluster->cpu);
+			ret = -ENODEV;
+			break;
+		}
+	}
+
+	return ret;
+}
+
+static int sprd_init_cooling_cluster(struct platform_device *pdev)
+{
+	int cpu = 0, id, core_num, ret = 0;
+	struct sprd_cooling_cluster *cluster;
+	struct sprd_cooling_data *data = platform_get_drvdata(pdev);
+
+	do {
+		core_num = cpumask_weight(topology_core_cpumask(cpu));
+		id = topology_physical_package_id((cpu));
+		if (core_num > 0 && id >= 0) {
+			cluster = &data->cluster[id];
+			cluster->id = id;
+			cluster->cpu = cpu;
+			cluster->cpus = core_num;
+			ret = cpufreq_get_policy(&(cluster->policy), cpu);
+			if (ret < 0) {
+				dev_err(&pdev->dev,
+					"CPU%d failed to get policy\n",
+					cpu);
+				break;
+			}
+
+			cpu = cpu + core_num;
+		} else
+			break;
+
+	} while (cpu > 0);
+
+	return ret;
+}
+
+static int sprd_cpu_cooling_probe(struct platform_device *pdev)
+{
+	int ret = -1;
+	int counts = 0;
+	struct sprd_cooling_data *data;
+	struct device *dev = &pdev->dev;
+
+	data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
+	if (!data)
+		return -ENOMEM;
+
+	counts = sprd_get_cluster_counts();
+	data->nr_clusters = counts;
+	data->cluster = devm_kzalloc(dev,
+				     sizeof(*data->cluster) * data->nr_clusters,
+				     GFP_KERNEL);
+	if (!data->cluster)
+		return -ENOMEM;
+
+	platform_set_drvdata(pdev, data);
+
+	ret = sprd_init_cooling_cluster(pdev);
+	if (ret < 0) {
+		dev_err(dev, "Failed to init cooling cluster\n");
+		return ret;
+	}
+
+	ret = sprd_cpu_cooing_ops_register(pdev);
+	if (ret < 0) {
+		dev_err(dev, "Failed to register cooling callback function\n");
+		return ret;
+	}
+
+	return 0;
+}
+
+static int sprd_cpu_cooling_remove(struct platform_device *pdev)
+{
+	int ret;
+
+	ret = sprd_cpu_cooing_ops_unregister(pdev);
+	if (ret < 0) {
+		dev_err(&pdev->dev, "Failed to unregister cooling callback function\n");
+		return ret;
+	}
+
+	return 0;
+}
+
+static struct platform_driver sprd_cpu_cooling_driver = {
+	.probe = sprd_cpu_cooling_probe,
+	.remove = sprd_cpu_cooling_remove,
+	.driver = {
+		   .owner = THIS_MODULE,
+		   .name = "sprd_cpu_cooling",
+		   },
+};
+static int __init sprd_cpu_cooling_init(void)
+{
+	return platform_driver_register(&sprd_cpu_cooling_driver);
+}
+
+static void __exit sprd_cpu_cooling_exit(void)
+{
+	platform_driver_unregister(&sprd_cpu_cooling_driver);
+}
+
+module_init(sprd_cpu_cooling_init)
+module_exit(sprd_cpu_cooling_exit);
+
+MODULE_DESCRIPTION("sprd cpu cooling");
+MODULE_LICENSE("GPL");
-- 
2.20.1