1 files changed, 314 insertions, 74 deletions

diff --git a/drivers/cpufreq/scmi-cpufreq.c b/drivers/cpufreq/scmi-cpufreq.c
index 242c3370544e..d2a110079f5f 100644
--- a/drivers/cpufreq/scmi-cpufreq.c
+++ b/drivers/cpufreq/scmi-cpufreq.c
@@ -2,40 +2,54 @@
 /*
  * System Control and Power Interface (SCMI) based CPUFreq Interface driver
  *
- * Copyright (C) 2018 ARM Ltd.
+ * Copyright (C) 2018-2021 ARM Ltd.
  * Sudeep Holla <sudeep.holla@arm.com>
  */
 
 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
+#include <linux/clk-provider.h>
 #include <linux/cpu.h>
 #include <linux/cpufreq.h>
 #include <linux/cpumask.h>
-#include <linux/cpu_cooling.h>
+#include <linux/energy_model.h>
 #include <linux/export.h>
 #include <linux/module.h>
+#include <linux/of.h>
 #include <linux/pm_opp.h>
+#include <linux/pm_qos.h>
 #include <linux/slab.h>
 #include <linux/scmi_protocol.h>
 #include <linux/types.h>
+#include <linux/units.h>
 
 struct scmi_data {
 	int domain_id;
+	int nr_opp;
 	struct device *cpu_dev;
-	struct thermal_cooling_device *cdev;
+	cpumask_var_t opp_shared_cpus;
+	struct notifier_block limit_notify_nb;
+	struct freq_qos_request	limits_freq_req;
 };
 
-static const struct scmi_handle *handle;
+static struct scmi_protocol_handle *ph;
+static const struct scmi_perf_proto_ops *perf_ops;
+static struct cpufreq_driver scmi_cpufreq_driver;
 
 static unsigned int scmi_cpufreq_get_rate(unsigned int cpu)
 {
-	struct cpufreq_policy *policy = cpufreq_cpu_get_raw(cpu);
-	struct scmi_perf_ops *perf_ops = handle->perf_ops;
-	struct scmi_data *priv = policy->driver_data;
+	struct cpufreq_policy *policy;
+	struct scmi_data *priv;
 	unsigned long rate;
 	int ret;
 
-	ret = perf_ops->freq_get(handle, priv->domain_id, &rate, false);
+	policy = cpufreq_cpu_get_raw(cpu);
+	if (unlikely(!policy))
+		return 0;
+
+	priv = policy->driver_data;
+
+	ret = perf_ops->freq_get(ph, priv->domain_id, &rate, false);
 	if (ret)
 		return 0;
 	return rate / 1000;
@@ -49,45 +63,55 @@ static unsigned int scmi_cpufreq_get_rate(unsigned int cpu)
 static int
 scmi_cpufreq_set_target(struct cpufreq_policy *policy, unsigned int index)
 {
-	int ret;
 	struct scmi_data *priv = policy->driver_data;
-	struct scmi_perf_ops *perf_ops = handle->perf_ops;
 	u64 freq = policy->freq_table[index].frequency;
 
-	ret = perf_ops->freq_set(handle, priv->domain_id, freq * 1000, false);
-	if (!ret)
-		arch_set_freq_scale(policy->related_cpus, freq,
-				    policy->cpuinfo.max_freq);
-	return ret;
+	return perf_ops->freq_set(ph, priv->domain_id, freq * 1000, false);
 }
 
 static unsigned int scmi_cpufreq_fast_switch(struct cpufreq_policy *policy,
 					     unsigned int target_freq)
 {
 	struct scmi_data *priv = policy->driver_data;
-	struct scmi_perf_ops *perf_ops = handle->perf_ops;
+	unsigned long freq = target_freq;
 
-	if (!perf_ops->freq_set(handle, priv->domain_id,
-				target_freq * 1000, true)) {
-		arch_set_freq_scale(policy->related_cpus, target_freq,
-				    policy->cpuinfo.max_freq);
+	if (!perf_ops->freq_set(ph, priv->domain_id, freq * 1000, true))
 		return target_freq;
-	}
 
 	return 0;
 }
 
+static int scmi_cpu_domain_id(struct device *cpu_dev)
+{
+	struct device_node *np = cpu_dev->of_node;
+	struct of_phandle_args domain_id;
+	int index;
+
+	if (of_parse_phandle_with_args(np, "clocks", "#clock-cells", 0,
+				       &domain_id)) {
+		/* Find the corresponding index for power-domain "perf". */
+		index = of_property_match_string(np, "power-domain-names",
+						 "perf");
+		if (index < 0)
+			return -EINVAL;
+
+		if (of_parse_phandle_with_args(np, "power-domains",
+					       "#power-domain-cells", index,
+					       &domain_id))
+			return -EINVAL;
+	}
+
+	return domain_id.args[0];
+}
+
 static int
-scmi_get_sharing_cpus(struct device *cpu_dev, struct cpumask *cpumask)
+scmi_get_sharing_cpus(struct device *cpu_dev, int domain,
+		      struct cpumask *cpumask)
 {
-	int cpu, domain, tdomain;
+	int cpu, tdomain;
 	struct device *tcpu_dev;
 
-	domain = handle->perf_ops->device_domain_id(cpu_dev);
-	if (domain < 0)
-		return domain;
-
-	for_each_possible_cpu(cpu) {
+	for_each_present_cpu(cpu) {
 		if (cpu == cpu_dev->id)
 			continue;
 
@@ -95,7 +119,7 @@ scmi_get_sharing_cpus(struct device *cpu_dev, struct cpumask *cpumask)
 		if (!tcpu_dev)
 			continue;
 
-		tdomain = handle->perf_ops->device_domain_id(tcpu_dev);
+		tdomain = scmi_cpu_domain_id(tcpu_dev);
 		if (tdomain == domain)
 			cpumask_set_cpu(cpu, cpumask);
 	}
@@ -103,13 +127,81 @@ scmi_get_sharing_cpus(struct device *cpu_dev, struct cpumask *cpumask)
 	return 0;
 }
 
-static int scmi_cpufreq_init(struct cpufreq_policy *policy)
+static int __maybe_unused
+scmi_get_cpu_power(struct device *cpu_dev, unsigned long *power,
+		   unsigned long *KHz)
 {
+	enum scmi_power_scale power_scale = perf_ops->power_scale_get(ph);
+	unsigned long Hz;
+	int ret, domain;
+
+	domain = scmi_cpu_domain_id(cpu_dev);
+	if (domain < 0)
+		return domain;
+
+	/* Get the power cost of the performance domain. */
+	Hz = *KHz * 1000;
+	ret = perf_ops->est_power_get(ph, domain, &Hz, power);
+	if (ret)
+		return ret;
+
+	/* Convert the power to uW if it is mW (ignore bogoW) */
+	if (power_scale == SCMI_POWER_MILLIWATTS)
+		*power *= MICROWATT_PER_MILLIWATT;
+
+	/* The EM framework specifies the frequency in KHz. */
+	*KHz = Hz / 1000;
+
+	return 0;
+}
+
+static int
+scmi_get_rate_limit(u32 domain, bool has_fast_switch)
+{
+	int ret, rate_limit;
+
+	if (has_fast_switch) {
+		/*
+		 * Fast channels are used whenever available,
+		 * so use their rate_limit value if populated.
+		 */
+		ret = perf_ops->fast_switch_rate_limit(ph, domain,
+						       &rate_limit);
+		if (!ret && rate_limit)
+			return rate_limit;
+	}
+
+	ret = perf_ops->rate_limit_get(ph, domain, &rate_limit);
+	if (ret)
+		return 0;
+
+	return rate_limit;
+}
+
+static int scmi_limit_notify_cb(struct notifier_block *nb, unsigned long event, void *data)
+{
+	struct scmi_data *priv = container_of(nb, struct scmi_data, limit_notify_nb);
+	struct scmi_perf_limits_report *limit_notify = data;
+	unsigned int limit_freq_khz;
 	int ret;
+
+	limit_freq_khz = limit_notify->range_max_freq / HZ_PER_KHZ;
+
+	ret = freq_qos_update_request(&priv->limits_freq_req, limit_freq_khz);
+	if (ret < 0)
+		pr_warn("failed to update freq constraint: %d\n", ret);
+
+	return NOTIFY_OK;
+}
+
+static int scmi_cpufreq_init(struct cpufreq_policy *policy)
+{
+	int ret, nr_opp, domain;
 	unsigned int latency;
 	struct device *cpu_dev;
 	struct scmi_data *priv;
 	struct cpufreq_frequency_table *freq_table;
+	struct scmi_device *sdev = cpufreq_get_driver_data();
 
 	cpu_dev = get_cpu_device(policy->cpu);
 	if (!cpu_dev) {
@@ -117,46 +209,82 @@ static int scmi_cpufreq_init(struct cpufreq_policy *policy)
 		return -ENODEV;
 	}
 
-	ret = handle->perf_ops->device_opps_add(handle, cpu_dev);
-	if (ret) {
-		dev_warn(cpu_dev, "failed to add opps to the device\n");
-		return ret;
-	}
+	domain = scmi_cpu_domain_id(cpu_dev);
+	if (domain < 0)
+		return domain;
 
-	ret = scmi_get_sharing_cpus(cpu_dev, policy->cpus);
-	if (ret) {
-		dev_warn(cpu_dev, "failed to get sharing cpumask\n");
-		return ret;
+	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;
 	}
 
-	ret = dev_pm_opp_set_sharing_cpus(cpu_dev, policy->cpus);
+	/* Obtain CPUs that share SCMI performance controls */
+	ret = scmi_get_sharing_cpus(cpu_dev, domain, policy->cpus);
 	if (ret) {
-		dev_err(cpu_dev, "%s: failed to mark OPPs as shared: %d\n",
-			__func__, ret);
-		return ret;
+		dev_warn(cpu_dev, "failed to get sharing cpumask\n");
+		goto out_free_cpumask;
 	}
 
-	ret = dev_pm_opp_get_opp_count(cpu_dev);
-	if (ret <= 0) {
-		dev_dbg(cpu_dev, "OPP table is not ready, deferring probe\n");
-		ret = -EPROBE_DEFER;
-		goto out_free_opp;
+	/*
+	 * Obtain CPUs that share performance levels.
+	 * 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, priv->opp_shared_cpus);
+	if (ret || cpumask_empty(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(priv->opp_shared_cpus, policy->cpus);
 	}
 
-	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
-	if (!priv) {
-		ret = -ENOMEM;
-		goto out_free_opp;
+	 /*
+	  * A previous CPU may have marked OPPs as shared for a few CPUs, based on
+	  * what OPP core provided. If the current CPU is part of those few, then
+	  * there is no need to add OPPs again.
+	  */
+	nr_opp = dev_pm_opp_get_opp_count(cpu_dev);
+	if (nr_opp <= 0) {
+		ret = perf_ops->device_opps_add(ph, cpu_dev, domain);
+		if (ret) {
+			dev_warn(cpu_dev, "failed to add opps to the device\n");
+			goto out_free_cpumask;
+		}
+
+		nr_opp = dev_pm_opp_get_opp_count(cpu_dev);
+		if (nr_opp <= 0) {
+			dev_err(cpu_dev, "%s: No OPPs for this device: %d\n",
+				__func__, nr_opp);
+
+			ret = -ENODEV;
+			goto out_free_opp;
+		}
+
+		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);
+
+			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;
-	priv->domain_id = handle->perf_ops->device_domain_id(cpu_dev);
+	priv->domain_id = domain;
 
 	policy->driver_data = priv;
 	policy->freq_table = freq_table;
@@ -164,68 +292,180 @@ static int scmi_cpufreq_init(struct cpufreq_policy *policy)
 	/* SCMI allows DVFS request for any domain from any CPU */
 	policy->dvfs_possible_from_any_cpu = true;
 
-	latency = handle->perf_ops->transition_latency_get(handle, cpu_dev);
+	latency = perf_ops->transition_latency_get(ph, domain);
 	if (!latency)
-		latency = CPUFREQ_ETERNAL;
+		latency = CPUFREQ_DEFAULT_TRANSITION_LATENCY_NS;
 
 	policy->cpuinfo.transition_latency = latency;
 
-	policy->fast_switch_possible = true;
+	policy->fast_switch_possible =
+		perf_ops->fast_switch_possible(ph, domain);
+
+	policy->transition_delay_us =
+		scmi_get_rate_limit(domain, policy->fast_switch_possible);
+
+	ret = freq_qos_add_request(&policy->constraints, &priv->limits_freq_req, FREQ_QOS_MAX,
+				   FREQ_QOS_MAX_DEFAULT_VALUE);
+	if (ret < 0) {
+		dev_err(cpu_dev, "failed to add qos limits request: %d\n", ret);
+		goto out_free_table;
+	}
+
+	priv->limit_notify_nb.notifier_call = scmi_limit_notify_cb;
+	ret = sdev->handle->notify_ops->event_notifier_register(sdev->handle, SCMI_PROTOCOL_PERF,
+							SCMI_EVENT_PERFORMANCE_LIMITS_CHANGED,
+							&priv->domain_id,
+							&priv->limit_notify_nb);
+	if (ret)
+		dev_warn(&sdev->dev,
+			 "failed to register for limits change notifier for domain %d\n",
+			 priv->domain_id);
+
 	return 0;
 
-out_free_priv:
-	kfree(priv);
+out_free_table:
+	dev_pm_opp_free_cpufreq_table(cpu_dev, &freq_table);
 out_free_opp:
 	dev_pm_opp_remove_all_dynamic(cpu_dev);
 
+out_free_cpumask:
+	free_cpumask_var(priv->opp_shared_cpus);
+
+out_free_priv:
+	kfree(priv);
+
 	return ret;
 }
 
-static int scmi_cpufreq_exit(struct cpufreq_policy *policy)
+static void scmi_cpufreq_exit(struct cpufreq_policy *policy)
 {
 	struct scmi_data *priv = policy->driver_data;
+	struct scmi_device *sdev = cpufreq_get_driver_data();
 
-	cpufreq_cooling_unregister(priv->cdev);
+	sdev->handle->notify_ops->event_notifier_unregister(sdev->handle, SCMI_PROTOCOL_PERF,
+							    SCMI_EVENT_PERFORMANCE_LIMITS_CHANGED,
+							    &priv->domain_id,
+							    &priv->limit_notify_nb);
+	freq_qos_remove_request(&priv->limits_freq_req);
 	dev_pm_opp_free_cpufreq_table(priv->cpu_dev, &policy->freq_table);
-	kfree(priv);
 	dev_pm_opp_remove_all_dynamic(priv->cpu_dev);
-
-	return 0;
+	free_cpumask_var(priv->opp_shared_cpus);
+	kfree(priv);
 }
 
-static void scmi_cpufreq_ready(struct cpufreq_policy *policy)
+static void scmi_cpufreq_register_em(struct cpufreq_policy *policy)
 {
+	struct em_data_callback em_cb = EM_DATA_CB(scmi_get_cpu_power);
+	enum scmi_power_scale power_scale = perf_ops->power_scale_get(ph);
 	struct scmi_data *priv = policy->driver_data;
-
-	priv->cdev = of_cpufreq_cooling_register(policy);
+	bool em_power_scale = false;
+
+	/*
+	 * This callback will be called for each policy, but we don't need to
+	 * register with EM every time. Despite not being part of the same
+	 * policy, some CPUs may still share their perf-domains, and a CPU from
+	 * another policy may already have registered with EM on behalf of CPUs
+	 * of this policy.
+	 */
+	if (!priv->nr_opp)
+		return;
+
+	if (power_scale == SCMI_POWER_MILLIWATTS
+	    || power_scale == SCMI_POWER_MICROWATTS)
+		em_power_scale = true;
+
+	em_dev_register_perf_domain(get_cpu_device(policy->cpu), priv->nr_opp,
+				    &em_cb, priv->opp_shared_cpus,
+				    em_power_scale);
 }
 
 static struct cpufreq_driver scmi_cpufreq_driver = {
 	.name	= "scmi",
-	.flags	= CPUFREQ_STICKY | CPUFREQ_HAVE_GOVERNOR_PER_POLICY |
-		  CPUFREQ_NEED_INITIAL_FREQ_CHECK,
+	.flags	= CPUFREQ_HAVE_GOVERNOR_PER_POLICY |
+		  CPUFREQ_NEED_INITIAL_FREQ_CHECK |
+		  CPUFREQ_IS_COOLING_DEV,
 	.verify	= cpufreq_generic_frequency_table_verify,
-	.attr	= cpufreq_generic_attr,
 	.target_index	= scmi_cpufreq_set_target,
 	.fast_switch	= scmi_cpufreq_fast_switch,
 	.get	= scmi_cpufreq_get_rate,
 	.init	= scmi_cpufreq_init,
 	.exit	= scmi_cpufreq_exit,
-	.ready	= scmi_cpufreq_ready,
+	.register_em	= scmi_cpufreq_register_em,
+	.set_boost	= cpufreq_boost_set_sw,
 };
 
+static bool scmi_dev_used_by_cpus(struct device *scmi_dev)
+{
+	struct device_node *scmi_np = dev_of_node(scmi_dev);
+	struct device_node *cpu_np, *np;
+	struct device *cpu_dev;
+	int cpu, idx;
+
+	if (!scmi_np)
+		return false;
+
+	for_each_possible_cpu(cpu) {
+		cpu_dev = get_cpu_device(cpu);
+		if (!cpu_dev)
+			continue;
+
+		cpu_np = dev_of_node(cpu_dev);
+
+		np = of_parse_phandle(cpu_np, "clocks", 0);
+		of_node_put(np);
+
+		if (np == scmi_np)
+			return true;
+
+		idx = of_property_match_string(cpu_np, "power-domain-names", "perf");
+		np = of_parse_phandle(cpu_np, "power-domains", idx);
+		of_node_put(np);
+
+		if (np == scmi_np)
+			return true;
+	}
+
+	/*
+	 * Older Broadcom STB chips had a "clocks" property for CPU node(s)
+	 * that did not match the SCMI performance protocol node, if we got
+	 * there, it means we had such an older Device Tree, therefore return
+	 * true to preserve backwards compatibility.
+	 */
+	if (of_machine_is_compatible("brcm,brcmstb"))
+		return true;
+
+	return false;
+}
+
 static int scmi_cpufreq_probe(struct scmi_device *sdev)
 {
 	int ret;
+	struct device *dev = &sdev->dev;
+	const struct scmi_handle *handle;
 
 	handle = sdev->handle;
 
-	if (!handle || !handle->perf_ops)
+	if (!handle || !scmi_dev_used_by_cpus(dev))
 		return -ENODEV;
 
+	scmi_cpufreq_driver.driver_data = sdev;
+
+	perf_ops = handle->devm_protocol_get(sdev, SCMI_PROTOCOL_PERF, &ph);
+	if (IS_ERR(perf_ops))
+		return PTR_ERR(perf_ops);
+
+#ifdef CONFIG_COMMON_CLK
+	/* dummy clock provider as needed by OPP if clocks property is used */
+	if (of_property_present(dev->of_node, "#clock-cells")) {
+		ret = devm_of_clk_add_hw_provider(dev, of_clk_hw_simple_get, NULL);
+		if (ret)
+			return dev_err_probe(dev, ret, "%s: registering clock provider failed\n", __func__);
+	}
+#endif
+
 	ret = cpufreq_register_driver(&scmi_cpufreq_driver);
 	if (ret) {
-		dev_err(&sdev->dev, "%s: registering cpufreq failed, err: %d\n",
+		dev_err(dev, "%s: registering cpufreq failed, err: %d\n",
 			__func__, ret);
 	}
 
@@ -238,7 +478,7 @@ static void scmi_cpufreq_remove(struct scmi_device *sdev)
 }
 
 static const struct scmi_device_id scmi_id_table[] = {
-	{ SCMI_PROTOCOL_PERF },
+	{ SCMI_PROTOCOL_PERF, "cpufreq" },
 	{ },
 };
 MODULE_DEVICE_TABLE(scmi, scmi_id_table);