Merge tag 'devfreq-fixes-for-5.19-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/chanwoo/linux

Pull devfreq fixes for 5.19-rc5 from Chanwoo Choi:

"1. Fix devfreq passive governor issue when cpufreq policies are not
    ready during kernel boot because some CPUs turn on after kernel
    booting or others.

    - Re-initialize the vairables of struct devfreq_passive_data when
      PROBE_DEFER happens when cpufreq_get() returns NULL.

    - Use dev_err_probe to mute warning when PROBE_DEFER.

    - Fix cpufreq passive unregister erroring on PROBE_DEFER
      by using the allocated parent_cpu_data list to free resouce
      instead of for_each_possible_cpu().

    - Remove duplicate cpufreq passive unregister and warning when
      PROBE_DEFER.

    - Use HZ_PER_KZH macro in units.h.

    - Fix wrong indentation in SPDX-License line.

 2. Fix reference count leak in exynos-ppmu.c by using of_node_put().

 3. Rework freq_table to be local to devfreq struct

    - struct devfreq_dev_profile includes freq_table array to store
      the supported frequencies. If devfreq driver doesn't initialize
      the freq_table, devfreq core allocates the memory and initializes
      the freq_table.

      On a devfreq PROBE_DEFER, the freq_table in the driver profile
      struct is never reset and may be left in an undefined state. To fix
      this and correctly handle PROBE_DEFER, use a local freq_table and
      max_state in the devfreq struct."

* tag 'devfreq-fixes-for-5.19-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/chanwoo/linux:
  PM / devfreq: passive: revert an editing accident in SPDX-License line
  PM / devfreq: Fix kernel warning with cpufreq passive register fail
  PM / devfreq: Rework freq_table to be local to devfreq struct
  PM / devfreq: exynos-ppmu: Fix refcount leak in of_get_devfreq_events
  PM / devfreq: passive: Use HZ_PER_KHZ macro in units.h
  PM / devfreq: Fix cpufreq passive unregister erroring on PROBE_DEFER
  PM / devfreq: Mute warning on governor PROBE_DEFER
  PM / devfreq: Fix kernel panic with cpu based scaling to passive gov

Author: rafaeljw

drivers/devfreq/devfreq.c

@@ -123,7 +123,7 @@ void devfreq_get_freq_range(struct devfreq *devfreq,
 			    unsigned long *min_freq,
 			    unsigned long *max_freq)
 {
-	unsigned long *freq_table = devfreq->profile->freq_table;
+	unsigned long *freq_table = devfreq->freq_table;
 	s32 qos_min_freq, qos_max_freq;
 
 	lockdep_assert_held(&devfreq->lock);
@@ -133,11 +133,11 @@ void devfreq_get_freq_range(struct devfreq *devfreq,
 	 * The devfreq drivers can initialize this in either ascending or
 	 * descending order and devfreq core supports both.
 	 */
-	if (freq_table[0] < freq_table[devfreq->profile->max_state - 1]) {
+	if (freq_table[0] < freq_table[devfreq->max_state - 1]) {
 		*min_freq = freq_table[0];
-		*max_freq = freq_table[devfreq->profile->max_state - 1];
+		*max_freq = freq_table[devfreq->max_state - 1];
 	} else {
-		*min_freq = freq_table[devfreq->profile->max_state - 1];
+		*min_freq = freq_table[devfreq->max_state - 1];
 		*max_freq = freq_table[0];
 	}
 
@@ -169,16 +169,15 @@ static int devfreq_get_freq_level(struct devfreq *devfreq, unsigned long freq)
 {
 	int lev;
 
-	for (lev = 0; lev < devfreq->profile->max_state; lev++)
-		if (freq == devfreq->profile->freq_table[lev])
+	for (lev = 0; lev < devfreq->max_state; lev++)
+		if (freq == devfreq->freq_table[lev])
 			return lev;
 
 	return -EINVAL;
 }
 
 static int set_freq_table(struct devfreq *devfreq)
 {
-	struct devfreq_dev_profile *profile = devfreq->profile;
 	struct dev_pm_opp *opp;
 	unsigned long freq;
 	int i, count;
@@ -188,25 +187,22 @@ static int set_freq_table(struct devfreq *devfreq)
 	if (count <= 0)
 		return -EINVAL;
 
-	profile->max_state = count;
-	profile->freq_table = devm_kcalloc(devfreq->dev.parent,
-					profile->max_state,
-					sizeof(*profile->freq_table),
-					GFP_KERNEL);
-	if (!profile->freq_table) {
-		profile->max_state = 0;
+	devfreq->max_state = count;
+	devfreq->freq_table = devm_kcalloc(devfreq->dev.parent,
+					   devfreq->max_state,
+					   sizeof(*devfreq->freq_table),
+					   GFP_KERNEL);
+	if (!devfreq->freq_table)
 		return -ENOMEM;
-	}
 
-	for (i = 0, freq = 0; i < profile->max_state; i++, freq++) {
+	for (i = 0, freq = 0; i < devfreq->max_state; i++, freq++) {
 		opp = dev_pm_opp_find_freq_ceil(devfreq->dev.parent, &freq);
 		if (IS_ERR(opp)) {
-			devm_kfree(devfreq->dev.parent, profile->freq_table);
-			profile->max_state = 0;
+			devm_kfree(devfreq->dev.parent, devfreq->freq_table);
 			return PTR_ERR(opp);
 		}
 		dev_pm_opp_put(opp);
-		profile->freq_table[i] = freq;
+		devfreq->freq_table[i] = freq;
 	}
 
 	return 0;
@@ -246,7 +242,7 @@ int devfreq_update_status(struct devfreq *devfreq, unsigned long freq)
 
 	if (lev != prev_lev) {
 		devfreq->stats.trans_table[
-			(prev_lev * devfreq->profile->max_state) + lev]++;
+			(prev_lev * devfreq->max_state) + lev]++;
 		devfreq->stats.total_trans++;
 	}
 
@@ -835,6 +831,9 @@ struct devfreq *devfreq_add_device(struct device *dev,
 		if (err < 0)
 			goto err_dev;
 		mutex_lock(&devfreq->lock);
+	} else {
+		devfreq->freq_table = devfreq->profile->freq_table;
+		devfreq->max_state = devfreq->profile->max_state;
 	}
 
 	devfreq->scaling_min_freq = find_available_min_freq(devfreq);
@@ -870,8 +869,8 @@ struct devfreq *devfreq_add_device(struct device *dev,
 
 	devfreq->stats.trans_table = devm_kzalloc(&devfreq->dev,
 			array3_size(sizeof(unsigned int),
-				    devfreq->profile->max_state,
-				    devfreq->profile->max_state),
+				    devfreq->max_state,
+				    devfreq->max_state),
 			GFP_KERNEL);
 	if (!devfreq->stats.trans_table) {
 		mutex_unlock(&devfreq->lock);
@@ -880,7 +879,7 @@ struct devfreq *devfreq_add_device(struct device *dev,
 	}
 
 	devfreq->stats.time_in_state = devm_kcalloc(&devfreq->dev,
-			devfreq->profile->max_state,
+			devfreq->max_state,
 			sizeof(*devfreq->stats.time_in_state),
 			GFP_KERNEL);
 	if (!devfreq->stats.time_in_state) {
@@ -932,8 +931,9 @@ struct devfreq *devfreq_add_device(struct device *dev,
 	err = devfreq->governor->event_handler(devfreq, DEVFREQ_GOV_START,
 						NULL);
 	if (err) {
-		dev_err(dev, "%s: Unable to start governor for the device\n",
-			__func__);
+		dev_err_probe(dev, err,
+			"%s: Unable to start governor for the device\n",
+			 __func__);
 		goto err_init;
 	}
 	create_sysfs_files(devfreq, devfreq->governor);
@@ -1665,9 +1665,9 @@ static ssize_t available_frequencies_show(struct device *d,
 
 	mutex_lock(&df->lock);
 
-	for (i = 0; i < df->profile->max_state; i++)
+	for (i = 0; i < df->max_state; i++)
 		count += scnprintf(&buf[count], (PAGE_SIZE - count - 2),
-				"%lu ", df->profile->freq_table[i]);
+				"%lu ", df->freq_table[i]);
 
 	mutex_unlock(&df->lock);
 	/* Truncate the trailing space */
@@ -1690,7 +1690,7 @@ static ssize_t trans_stat_show(struct device *dev,
 
 	if (!df->profile)
 		return -EINVAL;
-	max_state = df->profile->max_state;
+	max_state = df->max_state;
 
 	if (max_state == 0)
 		return sprintf(buf, "Not Supported.\n");
@@ -1707,19 +1707,17 @@ static ssize_t trans_stat_show(struct device *dev,
 	len += sprintf(buf + len, "           :");
 	for (i = 0; i < max_state; i++)
 		len += sprintf(buf + len, "%10lu",
-				df->profile->freq_table[i]);
+				df->freq_table[i]);
 
 	len += sprintf(buf + len, "   time(ms)\n");
 
 	for (i = 0; i < max_state; i++) {
-		if (df->profile->freq_table[i]
-					== df->previous_freq) {
+		if (df->freq_table[i] == df->previous_freq)
 			len += sprintf(buf + len, "*");
-		} else {
+		else
 			len += sprintf(buf + len, " ");
-		}
-		len += sprintf(buf + len, "%10lu:",
-				df->profile->freq_table[i]);
+
+		len += sprintf(buf + len, "%10lu:", df->freq_table[i]);
 		for (j = 0; j < max_state; j++)
 			len += sprintf(buf + len, "%10u",
 				df->stats.trans_table[(i * max_state) + j]);
@@ -1743,19 +1741,19 @@ static ssize_t trans_stat_store(struct device *dev,
 	if (!df->profile)
 		return -EINVAL;
 
-	if (df->profile->max_state == 0)
+	if (df->max_state == 0)
 		return count;
 
 	err = kstrtoint(buf, 10, &value);
 	if (err || value != 0)
 		return -EINVAL;
 
 	mutex_lock(&df->lock);
-	memset(df->stats.time_in_state, 0, (df->profile->max_state *
+	memset(df->stats.time_in_state, 0, (df->max_state *
 					sizeof(*df->stats.time_in_state)));
 	memset(df->stats.trans_table, 0, array3_size(sizeof(unsigned int),
-					df->profile->max_state,
-					df->profile->max_state));
+					df->max_state,
+					df->max_state));
 	df->stats.total_trans = 0;
 	df->stats.last_update = get_jiffies_64();
 	mutex_unlock(&df->lock);

drivers/devfreq/event/exynos-ppmu.c

@@ -519,15 +519,19 @@ static int of_get_devfreq_events(struct device_node *np,
 
 	count = of_get_child_count(events_np);
 	desc = devm_kcalloc(dev, count, sizeof(*desc), GFP_KERNEL);
-	if (!desc)
+	if (!desc) {
+		of_node_put(events_np);
 		return -ENOMEM;
+	}
 	info->num_events = count;
 
 	of_id = of_match_device(exynos_ppmu_id_match, dev);
 	if (of_id)
 		info->ppmu_type = (enum exynos_ppmu_type)of_id->data;
-	else
+	else {
+		of_node_put(events_np);
 		return -EINVAL;
+	}
 
 	j = 0;
 	for_each_child_of_node(events_np, node) {

drivers/devfreq/governor_passive.c

@@ -1,4 +1,4 @@
-	// SPDX-License-Identifier: GPL-2.0-only
+// SPDX-License-Identifier: GPL-2.0-only
 /*
  * linux/drivers/devfreq/governor_passive.c
  *
@@ -14,10 +14,9 @@
 #include <linux/slab.h>
 #include <linux/device.h>
 #include <linux/devfreq.h>
+#include <linux/units.h>
 #include "governor.h"
 
-#define HZ_PER_KHZ	1000
-
 static struct devfreq_cpu_data *
 get_parent_cpu_data(struct devfreq_passive_data *p_data,
 		    struct cpufreq_policy *policy)
@@ -34,6 +33,20 @@ get_parent_cpu_data(struct devfreq_passive_data *p_data,
 	return NULL;
 }
 
+static void delete_parent_cpu_data(struct devfreq_passive_data *p_data)
+{
+	struct devfreq_cpu_data *parent_cpu_data, *tmp;
+
+	list_for_each_entry_safe(parent_cpu_data, tmp, &p_data->cpu_data_list, node) {
+		list_del(&parent_cpu_data->node);
+
+		if (parent_cpu_data->opp_table)
+			dev_pm_opp_put_opp_table(parent_cpu_data->opp_table);
+
+		kfree(parent_cpu_data);
+	}
+}
+
 static unsigned long get_target_freq_by_required_opp(struct device *p_dev,
 						struct opp_table *p_opp_table,
 						struct opp_table *opp_table,
@@ -131,18 +144,18 @@ static int get_target_freq_with_devfreq(struct devfreq *devfreq,
 		goto out;
 
 	/* Use interpolation if required opps is not available */
-	for (i = 0; i < parent_devfreq->profile->max_state; i++)
-		if (parent_devfreq->profile->freq_table[i] == *freq)
+	for (i = 0; i < parent_devfreq->max_state; i++)
+		if (parent_devfreq->freq_table[i] == *freq)
 			break;
 
-	if (i == parent_devfreq->profile->max_state)
+	if (i == parent_devfreq->max_state)
 		return -EINVAL;
 
-	if (i < devfreq->profile->max_state) {
-		child_freq = devfreq->profile->freq_table[i];
+	if (i < devfreq->max_state) {
+		child_freq = devfreq->freq_table[i];
 	} else {
-		count = devfreq->profile->max_state;
-		child_freq = devfreq->profile->freq_table[count - 1];
+		count = devfreq->max_state;
+		child_freq = devfreq->freq_table[count - 1];
 	}
 
 out:
@@ -222,8 +235,7 @@ static int cpufreq_passive_unregister_notifier(struct devfreq *devfreq)
 {
 	struct devfreq_passive_data *p_data
 			= (struct devfreq_passive_data *)devfreq->data;
-	struct devfreq_cpu_data *parent_cpu_data;
-	int cpu, ret = 0;
+	int ret;
 
 	if (p_data->nb.notifier_call) {
 		ret = cpufreq_unregister_notifier(&p_data->nb,
@@ -232,27 +244,9 @@ static int cpufreq_passive_unregister_notifier(struct devfreq *devfreq)
 			return ret;
 	}
 
-	for_each_possible_cpu(cpu) {
-		struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
-		if (!policy) {
-			ret = -EINVAL;
-			continue;
-		}
-
-		parent_cpu_data = get_parent_cpu_data(p_data, policy);
-		if (!parent_cpu_data) {
-			cpufreq_cpu_put(policy);
-			continue;
-		}
+	delete_parent_cpu_data(p_data);
 
-		list_del(&parent_cpu_data->node);
-		if (parent_cpu_data->opp_table)
-			dev_pm_opp_put_opp_table(parent_cpu_data->opp_table);
-		kfree(parent_cpu_data);
-		cpufreq_cpu_put(policy);
-	}
-
-	return ret;
+	return 0;
 }
 
 static int cpufreq_passive_register_notifier(struct devfreq *devfreq)
@@ -336,7 +330,6 @@ static int cpufreq_passive_register_notifier(struct devfreq *devfreq)
 err_put_policy:
 	cpufreq_cpu_put(policy);
 err:
-	WARN_ON(cpufreq_passive_unregister_notifier(devfreq));
 
 	return ret;
 }
@@ -407,8 +400,7 @@ static int devfreq_passive_event_handler(struct devfreq *devfreq,
 	if (!p_data)
 		return -EINVAL;
 
-	if (!p_data->this)
-		p_data->this = devfreq;
+	p_data->this = devfreq;
 
 	switch (event) {
 	case DEVFREQ_GOV_START:

include/linux/devfreq.h

@@ -148,6 +148,8 @@ struct devfreq_stats {
  *		reevaluate operable frequencies. Devfreq users may use
  *		devfreq.nb to the corresponding register notifier call chain.
  * @work:	delayed work for load monitoring.
+ * @freq_table:		current frequency table used by the devfreq driver.
+ * @max_state:		count of entry present in the frequency table.
  * @previous_freq:	previously configured frequency value.
  * @last_status:	devfreq user device info, performance statistics
  * @data:	Private data of the governor. The devfreq framework does not
@@ -185,6 +187,9 @@ struct devfreq {
 	struct notifier_block nb;
 	struct delayed_work work;
 
+	unsigned long *freq_table;
+	unsigned int max_state;
+
 	unsigned long previous_freq;
 	struct devfreq_dev_status last_status;