#include <linux/jiffies.h>
#include <linux/kernel_stat.h>
#include <linux/mutex.h>
+#include <linux/hrtimer.h>
+#include <linux/tick.h>
+#include <linux/ktime.h>
+#include <linux/sched.h>
/*
* dbs is used in this file as a shortform for demandbased switching
* It helps to keep variable names smaller, simpler
*/
+#define DEF_FREQUENCY_DOWN_DIFFERENTIAL (10)
#define DEF_FREQUENCY_UP_THRESHOLD (80)
+#define MICRO_FREQUENCY_DOWN_DIFFERENTIAL (3)
+#define MICRO_FREQUENCY_UP_THRESHOLD (95)
#define MIN_FREQUENCY_UP_THRESHOLD (11)
#define MAX_FREQUENCY_UP_THRESHOLD (100)
(MIN_SAMPLING_RATE_RATIO * jiffies_to_usecs(10))
#define MIN_SAMPLING_RATE \
(def_sampling_rate / MIN_SAMPLING_RATE_RATIO)
+/* Above MIN_SAMPLING_RATE will vanish with its sysfs file soon
+ * Define the minimal settable sampling rate to the greater of:
+ * - "HW transition latency" * 100 (same as default sampling / 10)
+ * - MIN_STAT_SAMPLING_RATE
+ * To avoid that userspace shoots itself.
+*/
+static unsigned int minimum_sampling_rate(void)
+{
+ return max(def_sampling_rate / 10, MIN_STAT_SAMPLING_RATE);
+}
+
+/* This will also vanish soon with removing sampling_rate_max */
#define MAX_SAMPLING_RATE (500 * def_sampling_rate)
-#define DEF_SAMPLING_RATE_LATENCY_MULTIPLIER (1000)
-#define TRANSITION_LATENCY_LIMIT (10 * 1000)
+#define LATENCY_MULTIPLIER (1000)
+#define TRANSITION_LATENCY_LIMIT (10 * 1000 * 1000)
-static void do_dbs_timer(void *data);
+static void do_dbs_timer(struct work_struct *work);
+
+/* Sampling types */
+enum {DBS_NORMAL_SAMPLE, DBS_SUB_SAMPLE};
struct cpu_dbs_info_s {
cputime64_t prev_cpu_idle;
cputime64_t prev_cpu_wall;
+ cputime64_t prev_cpu_nice;
struct cpufreq_policy *cur_policy;
- struct work_struct work;
- unsigned int enable;
+ struct delayed_work work;
struct cpufreq_frequency_table *freq_table;
unsigned int freq_lo;
unsigned int freq_lo_jiffies;
unsigned int freq_hi_jiffies;
+ int cpu;
+ unsigned int enable:1,
+ sample_type:1;
};
static DEFINE_PER_CPU(struct cpu_dbs_info_s, cpu_dbs_info);
* (like __cpufreq_driver_target()) is being called with dbs_mutex taken, then
* cpu_hotplug lock should be taken before that. Note that cpu_hotplug lock
* is recursive for the same process. -Venki
+ * DEADLOCK ALERT! (2) : do_dbs_timer() must not take the dbs_mutex, because it
+ * would deadlock with cancel_delayed_work_sync(), which is needed for proper
+ * raceless workqueue teardown.
*/
static DEFINE_MUTEX(dbs_mutex);
static struct dbs_tuners {
unsigned int sampling_rate;
unsigned int up_threshold;
+ unsigned int down_differential;
unsigned int ignore_nice;
unsigned int powersave_bias;
} dbs_tuners_ins = {
.up_threshold = DEF_FREQUENCY_UP_THRESHOLD,
+ .down_differential = DEF_FREQUENCY_DOWN_DIFFERENTIAL,
.ignore_nice = 0,
.powersave_bias = 0,
};
-static inline cputime64_t get_cpu_idle_time(unsigned int cpu)
+static inline cputime64_t get_cpu_idle_time_jiffy(unsigned int cpu,
+ cputime64_t *wall)
{
- cputime64_t retval;
+ cputime64_t idle_time;
+ cputime64_t cur_wall_time;
+ cputime64_t busy_time;
+
+ cur_wall_time = jiffies64_to_cputime64(get_jiffies_64());
+ busy_time = cputime64_add(kstat_cpu(cpu).cpustat.user,
+ kstat_cpu(cpu).cpustat.system);
- retval = cputime64_add(kstat_cpu(cpu).cpustat.idle,
- kstat_cpu(cpu).cpustat.iowait);
+ busy_time = cputime64_add(busy_time, kstat_cpu(cpu).cpustat.irq);
+ busy_time = cputime64_add(busy_time, kstat_cpu(cpu).cpustat.softirq);
+ busy_time = cputime64_add(busy_time, kstat_cpu(cpu).cpustat.steal);
+ busy_time = cputime64_add(busy_time, kstat_cpu(cpu).cpustat.nice);
- if (dbs_tuners_ins.ignore_nice)
- retval = cputime64_add(retval, kstat_cpu(cpu).cpustat.nice);
+ idle_time = cputime64_sub(cur_wall_time, busy_time);
+ if (wall)
+ *wall = cur_wall_time;
- return retval;
+ return idle_time;
+}
+
+static inline cputime64_t get_cpu_idle_time(unsigned int cpu, cputime64_t *wall)
+{
+ u64 idle_time = get_cpu_idle_time_us(cpu, wall);
+
+ if (idle_time == -1ULL)
+ return get_cpu_idle_time_jiffy(cpu, wall);
+
+ return idle_time;
}
/*
/************************** sysfs interface ************************/
static ssize_t show_sampling_rate_max(struct cpufreq_policy *policy, char *buf)
{
- return sprintf (buf, "%u\n", MAX_SAMPLING_RATE);
+ static int print_once;
+
+ if (!print_once) {
+ printk(KERN_INFO "CPUFREQ: ondemand sampling_rate_max "
+ "sysfs file is deprecated - used by: %s\n",
+ current->comm);
+ print_once = 1;
+ }
+ return sprintf(buf, "%u\n", MAX_SAMPLING_RATE);
}
static ssize_t show_sampling_rate_min(struct cpufreq_policy *policy, char *buf)
{
- return sprintf (buf, "%u\n", MIN_SAMPLING_RATE);
+ static int print_once;
+
+ if (!print_once) {
+ printk(KERN_INFO "CPUFREQ: ondemand sampling_rate_min "
+ "sysfs file is deprecated - used by: %s\n",
+ current->comm);
+ print_once = 1;
+ }
+ return sprintf(buf, "%u\n", MIN_SAMPLING_RATE);
}
#define define_one_ro(_name) \
ret = sscanf(buf, "%u", &input);
mutex_lock(&dbs_mutex);
- if (ret != 1 || input > MAX_SAMPLING_RATE
- || input < MIN_SAMPLING_RATE) {
+ if (ret != 1) {
mutex_unlock(&dbs_mutex);
return -EINVAL;
}
-
- dbs_tuners_ins.sampling_rate = input;
+ dbs_tuners_ins.sampling_rate = max(input, minimum_sampling_rate());
mutex_unlock(&dbs_mutex);
return count;
unsigned int j;
ret = sscanf(buf, "%u", &input);
- if ( ret != 1 )
+ if (ret != 1)
return -EINVAL;
- if ( input > 1 )
+ if (input > 1)
input = 1;
mutex_lock(&dbs_mutex);
- if ( input == dbs_tuners_ins.ignore_nice ) { /* nothing to do */
+ if (input == dbs_tuners_ins.ignore_nice) { /* nothing to do */
mutex_unlock(&dbs_mutex);
return count;
}
for_each_online_cpu(j) {
struct cpu_dbs_info_s *dbs_info;
dbs_info = &per_cpu(cpu_dbs_info, j);
- dbs_info->prev_cpu_idle = get_cpu_idle_time(j);
- dbs_info->prev_cpu_wall = get_jiffies_64();
+ dbs_info->prev_cpu_idle = get_cpu_idle_time(j,
+ &dbs_info->prev_cpu_wall);
+ if (dbs_tuners_ins.ignore_nice)
+ dbs_info->prev_cpu_nice = kstat_cpu(j).cpustat.nice;
+
}
mutex_unlock(&dbs_mutex);
define_one_rw(ignore_nice_load);
define_one_rw(powersave_bias);
-static struct attribute * dbs_attributes[] = {
+static struct attribute *dbs_attributes[] = {
&sampling_rate_max.attr,
&sampling_rate_min.attr,
&sampling_rate.attr,
static void dbs_check_cpu(struct cpu_dbs_info_s *this_dbs_info)
{
- unsigned int idle_ticks, total_ticks;
- unsigned int load;
- cputime64_t cur_jiffies;
+ unsigned int max_load_freq;
struct cpufreq_policy *policy;
unsigned int j;
this_dbs_info->freq_lo = 0;
policy = this_dbs_info->cur_policy;
- cur_jiffies = jiffies64_to_cputime64(get_jiffies_64());
- total_ticks = (unsigned int) cputime64_sub(cur_jiffies,
- this_dbs_info->prev_cpu_wall);
- this_dbs_info->prev_cpu_wall = cur_jiffies;
- if (!total_ticks)
- return;
+
/*
* Every sampling_rate, we check, if current idle time is less
* than 20% (default), then we try to increase frequency
* 5% (default) of current frequency
*/
- /* Get Idle Time */
- idle_ticks = UINT_MAX;
- for_each_cpu_mask(j, policy->cpus) {
- cputime64_t total_idle_ticks;
- unsigned int tmp_idle_ticks;
+ /* Get Absolute Load - in terms of freq */
+ max_load_freq = 0;
+
+ for_each_cpu(j, policy->cpus) {
struct cpu_dbs_info_s *j_dbs_info;
+ cputime64_t cur_wall_time, cur_idle_time;
+ unsigned int idle_time, wall_time;
+ unsigned int load, load_freq;
+ int freq_avg;
j_dbs_info = &per_cpu(cpu_dbs_info, j);
- total_idle_ticks = get_cpu_idle_time(j);
- tmp_idle_ticks = (unsigned int) cputime64_sub(total_idle_ticks,
+
+ cur_idle_time = get_cpu_idle_time(j, &cur_wall_time);
+
+ wall_time = (unsigned int) cputime64_sub(cur_wall_time,
+ j_dbs_info->prev_cpu_wall);
+ j_dbs_info->prev_cpu_wall = cur_wall_time;
+
+ idle_time = (unsigned int) cputime64_sub(cur_idle_time,
j_dbs_info->prev_cpu_idle);
- j_dbs_info->prev_cpu_idle = total_idle_ticks;
+ j_dbs_info->prev_cpu_idle = cur_idle_time;
+
+ if (dbs_tuners_ins.ignore_nice) {
+ cputime64_t cur_nice;
+ unsigned long cur_nice_jiffies;
+
+ cur_nice = cputime64_sub(kstat_cpu(j).cpustat.nice,
+ j_dbs_info->prev_cpu_nice);
+ /*
+ * Assumption: nice time between sampling periods will
+ * be less than 2^32 jiffies for 32 bit sys
+ */
+ cur_nice_jiffies = (unsigned long)
+ cputime64_to_jiffies64(cur_nice);
+
+ j_dbs_info->prev_cpu_nice = kstat_cpu(j).cpustat.nice;
+ idle_time += jiffies_to_usecs(cur_nice_jiffies);
+ }
+
+ if (unlikely(!wall_time || wall_time < idle_time))
+ continue;
- if (tmp_idle_ticks < idle_ticks)
- idle_ticks = tmp_idle_ticks;
+ load = 100 * (wall_time - idle_time) / wall_time;
+
+ freq_avg = __cpufreq_driver_getavg(policy, j);
+ if (freq_avg <= 0)
+ freq_avg = policy->cur;
+
+ load_freq = load * freq_avg;
+ if (load_freq > max_load_freq)
+ max_load_freq = load_freq;
}
- load = (100 * (total_ticks - idle_ticks)) / total_ticks;
/* Check for frequency increase */
- if (load > dbs_tuners_ins.up_threshold) {
+ if (max_load_freq > dbs_tuners_ins.up_threshold * policy->cur) {
/* if we are already at full speed then break out early */
if (!dbs_tuners_ins.powersave_bias) {
if (policy->cur == policy->max)
* can support the current CPU usage without triggering the up
* policy. To be safe, we focus 10 points under the threshold.
*/
- if (load < (dbs_tuners_ins.up_threshold - 10)) {
- unsigned int freq_next, freq_cur;
-
- freq_cur = cpufreq_driver_getavg(policy);
- if (!freq_cur)
- freq_cur = policy->cur;
-
- freq_next = (freq_cur * load) /
- (dbs_tuners_ins.up_threshold - 10);
+ if (max_load_freq <
+ (dbs_tuners_ins.up_threshold - dbs_tuners_ins.down_differential) *
+ policy->cur) {
+ unsigned int freq_next;
+ freq_next = max_load_freq /
+ (dbs_tuners_ins.up_threshold -
+ dbs_tuners_ins.down_differential);
if (!dbs_tuners_ins.powersave_bias) {
__cpufreq_driver_target(policy, freq_next,
}
}
-/* Sampling types */
-enum {DBS_NORMAL_SAMPLE, DBS_SUB_SAMPLE};
-
-static void do_dbs_timer(void *data)
+static void do_dbs_timer(struct work_struct *work)
{
- unsigned int cpu = smp_processor_id();
- struct cpu_dbs_info_s *dbs_info = &per_cpu(cpu_dbs_info, cpu);
+ struct cpu_dbs_info_s *dbs_info =
+ container_of(work, struct cpu_dbs_info_s, work.work);
+ unsigned int cpu = dbs_info->cpu;
+ int sample_type = dbs_info->sample_type;
+
/* We want all CPUs to do sampling nearly on same jiffy */
int delay = usecs_to_jiffies(dbs_tuners_ins.sampling_rate);
+
delay -= jiffies % delay;
- if (!dbs_info->enable)
+ if (lock_policy_rwsem_write(cpu) < 0)
+ return;
+
+ if (!dbs_info->enable) {
+ unlock_policy_rwsem_write(cpu);
return;
+ }
+
/* Common NORMAL_SAMPLE setup */
- INIT_WORK(&dbs_info->work, do_dbs_timer, (void *)DBS_NORMAL_SAMPLE);
+ dbs_info->sample_type = DBS_NORMAL_SAMPLE;
if (!dbs_tuners_ins.powersave_bias ||
- (unsigned long) data == DBS_NORMAL_SAMPLE) {
- lock_cpu_hotplug();
+ sample_type == DBS_NORMAL_SAMPLE) {
dbs_check_cpu(dbs_info);
- unlock_cpu_hotplug();
if (dbs_info->freq_lo) {
/* Setup timer for SUB_SAMPLE */
- INIT_WORK(&dbs_info->work, do_dbs_timer,
- (void *)DBS_SUB_SAMPLE);
+ dbs_info->sample_type = DBS_SUB_SAMPLE;
delay = dbs_info->freq_hi_jiffies;
}
} else {
__cpufreq_driver_target(dbs_info->cur_policy,
- dbs_info->freq_lo,
- CPUFREQ_RELATION_H);
+ dbs_info->freq_lo, CPUFREQ_RELATION_H);
}
queue_delayed_work_on(cpu, kondemand_wq, &dbs_info->work, delay);
+ unlock_policy_rwsem_write(cpu);
}
-static inline void dbs_timer_init(unsigned int cpu)
+static inline void dbs_timer_init(struct cpu_dbs_info_s *dbs_info)
{
- struct cpu_dbs_info_s *dbs_info = &per_cpu(cpu_dbs_info, cpu);
/* We want all CPUs to do sampling nearly on same jiffy */
int delay = usecs_to_jiffies(dbs_tuners_ins.sampling_rate);
delay -= jiffies % delay;
+ dbs_info->enable = 1;
ondemand_powersave_bias_init();
- INIT_WORK(&dbs_info->work, do_dbs_timer, NULL);
- queue_delayed_work_on(cpu, kondemand_wq, &dbs_info->work, delay);
+ dbs_info->sample_type = DBS_NORMAL_SAMPLE;
+ INIT_DELAYED_WORK_DEFERRABLE(&dbs_info->work, do_dbs_timer);
+ queue_delayed_work_on(dbs_info->cpu, kondemand_wq, &dbs_info->work,
+ delay);
}
static inline void dbs_timer_exit(struct cpu_dbs_info_s *dbs_info)
{
dbs_info->enable = 0;
- cancel_delayed_work(&dbs_info->work);
- flush_workqueue(kondemand_wq);
+ cancel_delayed_work_sync(&dbs_info->work);
}
static int cpufreq_governor_dbs(struct cpufreq_policy *policy,
if ((!cpu_online(cpu)) || (!policy->cur))
return -EINVAL;
- if (policy->cpuinfo.transition_latency >
- (TRANSITION_LATENCY_LIMIT * 1000)) {
- printk(KERN_WARNING "ondemand governor failed to load "
- "due to too long transition latency\n");
- return -EINVAL;
- }
if (this_dbs_info->enable) /* Already enabled */
break;
mutex_lock(&dbs_mutex);
dbs_enable++;
- if (dbs_enable == 1) {
- kondemand_wq = create_workqueue("kondemand");
- if (!kondemand_wq) {
- printk(KERN_ERR
- "Creation of kondemand failed\n");
- dbs_enable--;
- mutex_unlock(&dbs_mutex);
- return -ENOSPC;
- }
- }
rc = sysfs_create_group(&policy->kobj, &dbs_attr_group);
if (rc) {
- if (dbs_enable == 1)
- destroy_workqueue(kondemand_wq);
dbs_enable--;
mutex_unlock(&dbs_mutex);
return rc;
}
- for_each_cpu_mask(j, policy->cpus) {
+ for_each_cpu(j, policy->cpus) {
struct cpu_dbs_info_s *j_dbs_info;
j_dbs_info = &per_cpu(cpu_dbs_info, j);
j_dbs_info->cur_policy = policy;
- j_dbs_info->prev_cpu_idle = get_cpu_idle_time(j);
- j_dbs_info->prev_cpu_wall = get_jiffies_64();
+ j_dbs_info->prev_cpu_idle = get_cpu_idle_time(j,
+ &j_dbs_info->prev_cpu_wall);
+ if (dbs_tuners_ins.ignore_nice) {
+ j_dbs_info->prev_cpu_nice =
+ kstat_cpu(j).cpustat.nice;
+ }
}
- this_dbs_info->enable = 1;
+ this_dbs_info->cpu = cpu;
/*
* Start the timerschedule work, when this governor
* is used for first time
if (latency == 0)
latency = 1;
- def_sampling_rate = latency *
- DEF_SAMPLING_RATE_LATENCY_MULTIPLIER;
-
- if (def_sampling_rate < MIN_STAT_SAMPLING_RATE)
- def_sampling_rate = MIN_STAT_SAMPLING_RATE;
+ def_sampling_rate =
+ max(latency * LATENCY_MULTIPLIER,
+ MIN_STAT_SAMPLING_RATE);
dbs_tuners_ins.sampling_rate = def_sampling_rate;
}
- dbs_timer_init(policy->cpu);
+ dbs_timer_init(this_dbs_info);
mutex_unlock(&dbs_mutex);
break;
dbs_timer_exit(this_dbs_info);
sysfs_remove_group(&policy->kobj, &dbs_attr_group);
dbs_enable--;
- if (dbs_enable == 0)
- destroy_workqueue(kondemand_wq);
-
mutex_unlock(&dbs_mutex);
break;
mutex_lock(&dbs_mutex);
if (policy->max < this_dbs_info->cur_policy->cur)
__cpufreq_driver_target(this_dbs_info->cur_policy,
- policy->max,
- CPUFREQ_RELATION_H);
+ policy->max, CPUFREQ_RELATION_H);
else if (policy->min > this_dbs_info->cur_policy->cur)
__cpufreq_driver_target(this_dbs_info->cur_policy,
- policy->min,
- CPUFREQ_RELATION_L);
+ policy->min, CPUFREQ_RELATION_L);
mutex_unlock(&dbs_mutex);
break;
}
return 0;
}
-static struct cpufreq_governor cpufreq_gov_dbs = {
- .name = "ondemand",
- .governor = cpufreq_governor_dbs,
- .owner = THIS_MODULE,
+#ifndef CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND
+static
+#endif
+struct cpufreq_governor cpufreq_gov_ondemand = {
+ .name = "ondemand",
+ .governor = cpufreq_governor_dbs,
+ .max_transition_latency = TRANSITION_LATENCY_LIMIT,
+ .owner = THIS_MODULE,
};
static int __init cpufreq_gov_dbs_init(void)
{
- return cpufreq_register_governor(&cpufreq_gov_dbs);
+ int err;
+ cputime64_t wall;
+ u64 idle_time;
+ int cpu = get_cpu();
+
+ idle_time = get_cpu_idle_time_us(cpu, &wall);
+ put_cpu();
+ if (idle_time != -1ULL) {
+ /* Idle micro accounting is supported. Use finer thresholds */
+ dbs_tuners_ins.up_threshold = MICRO_FREQUENCY_UP_THRESHOLD;
+ dbs_tuners_ins.down_differential =
+ MICRO_FREQUENCY_DOWN_DIFFERENTIAL;
+ }
+
+ kondemand_wq = create_workqueue("kondemand");
+ if (!kondemand_wq) {
+ printk(KERN_ERR "Creation of kondemand failed\n");
+ return -EFAULT;
+ }
+ err = cpufreq_register_governor(&cpufreq_gov_ondemand);
+ if (err)
+ destroy_workqueue(kondemand_wq);
+
+ return err;
}
static void __exit cpufreq_gov_dbs_exit(void)
{
- cpufreq_unregister_governor(&cpufreq_gov_dbs);
+ cpufreq_unregister_governor(&cpufreq_gov_ondemand);
+ destroy_workqueue(kondemand_wq);
}
MODULE_AUTHOR("Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>");
MODULE_AUTHOR("Alexey Starikovskiy <alexey.y.starikovskiy@intel.com>");
MODULE_DESCRIPTION("'cpufreq_ondemand' - A dynamic cpufreq governor for "
- "Low Latency Frequency Transition capable processors");
+ "Low Latency Frequency Transition capable processors");
MODULE_LICENSE("GPL");
+#ifdef CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND
+fs_initcall(cpufreq_gov_dbs_init);
+#else
module_init(cpufreq_gov_dbs_init);
+#endif
module_exit(cpufreq_gov_dbs_exit);