diff options
Diffstat (limited to 'drivers/cpufreq/cpufreq_conservative.c')
-rw-r--r-- | drivers/cpufreq/cpufreq_conservative.c | 548 |
1 files changed, 159 insertions, 389 deletions
diff --git a/drivers/cpufreq/cpufreq_conservative.c b/drivers/cpufreq/cpufreq_conservative.c index 181abad..64ef737 100644 --- a/drivers/cpufreq/cpufreq_conservative.c +++ b/drivers/cpufreq/cpufreq_conservative.c @@ -11,83 +11,30 @@ * published by the Free Software Foundation. */ -#include <linux/kernel.h> -#include <linux/module.h> -#include <linux/init.h> #include <linux/cpufreq.h> -#include <linux/cpu.h> -#include <linux/jiffies.h> +#include <linux/init.h> +#include <linux/kernel.h> #include <linux/kernel_stat.h> +#include <linux/kobject.h> +#include <linux/module.h> #include <linux/mutex.h> -#include <linux/hrtimer.h> -#include <linux/tick.h> -#include <linux/ktime.h> -#include <linux/sched.h> +#include <linux/notifier.h> +#include <linux/percpu-defs.h> +#include <linux/sysfs.h> +#include <linux/types.h> -/* - * dbs is used in this file as a shortform for demandbased switching - * It helps to keep variable names smaller, simpler - */ +#include "cpufreq_governor.h" +/* Conservative governor macors */ #define DEF_FREQUENCY_UP_THRESHOLD (80) #define DEF_FREQUENCY_DOWN_THRESHOLD (20) - -/* - * The polling frequency of this governor depends on the capability of - * the processor. Default polling frequency is 1000 times the transition - * latency of the processor. The governor will work on any processor with - * transition latency <= 10mS, using appropriate sampling - * rate. - * For CPUs with transition latency > 10mS (mostly drivers with CPUFREQ_ETERNAL) - * this governor will not work. - * All times here are in uS. - */ -#define MIN_SAMPLING_RATE_RATIO (2) - -static unsigned int min_sampling_rate; - -#define LATENCY_MULTIPLIER (1000) -#define MIN_LATENCY_MULTIPLIER (100) #define DEF_SAMPLING_DOWN_FACTOR (1) #define MAX_SAMPLING_DOWN_FACTOR (10) -#define TRANSITION_LATENCY_LIMIT (10 * 1000 * 1000) - -static void do_dbs_timer(struct work_struct *work); - -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 delayed_work work; - unsigned int down_skip; - unsigned int requested_freq; - int cpu; - unsigned int enable:1; - /* - * percpu mutex that serializes governor limit change with - * do_dbs_timer invocation. We do not want do_dbs_timer to run - * when user is changing the governor or limits. - */ - struct mutex timer_mutex; -}; -static DEFINE_PER_CPU(struct cpu_dbs_info_s, cs_cpu_dbs_info); -static unsigned int dbs_enable; /* number of CPUs using this policy */ +static struct dbs_data cs_dbs_data; +static DEFINE_PER_CPU(struct cs_cpu_dbs_info_s, cs_cpu_dbs_info); -/* - * dbs_mutex protects dbs_enable in governor start/stop. - */ -static DEFINE_MUTEX(dbs_mutex); - -static struct dbs_tuners { - unsigned int sampling_rate; - unsigned int sampling_down_factor; - unsigned int up_threshold; - unsigned int down_threshold; - unsigned int ignore_nice; - unsigned int freq_step; -} dbs_tuners_ins = { +static struct cs_dbs_tuners cs_tuners = { .up_threshold = DEF_FREQUENCY_UP_THRESHOLD, .down_threshold = DEF_FREQUENCY_DOWN_THRESHOLD, .sampling_down_factor = DEF_SAMPLING_DOWN_FACTOR, @@ -95,61 +42,121 @@ static struct dbs_tuners { .freq_step = 5, }; -/* keep track of frequency transitions */ -static int -dbs_cpufreq_notifier(struct notifier_block *nb, unsigned long val, - void *data) +/* + * Every sampling_rate, we check, if current idle time is less than 20% + * (default), then we try to increase frequency Every sampling_rate * + * sampling_down_factor, we check, if current idle time is more than 80%, then + * we try to decrease frequency + * + * Any frequency increase takes it to the maximum frequency. Frequency reduction + * happens at minimum steps of 5% (default) of maximum frequency + */ +static void cs_check_cpu(int cpu, unsigned int load) { - struct cpufreq_freqs *freq = data; - struct cpu_dbs_info_s *this_dbs_info = &per_cpu(cs_cpu_dbs_info, - freq->cpu); + struct cs_cpu_dbs_info_s *dbs_info = &per_cpu(cs_cpu_dbs_info, cpu); + struct cpufreq_policy *policy = dbs_info->cdbs.cur_policy; + unsigned int freq_target; + + /* + * break out if we 'cannot' reduce the speed as the user might + * want freq_step to be zero + */ + if (cs_tuners.freq_step == 0) + return; + + /* Check for frequency increase */ + if (load > cs_tuners.up_threshold) { + dbs_info->down_skip = 0; + + /* if we are already at full speed then break out early */ + if (dbs_info->requested_freq == policy->max) + return; + + freq_target = (cs_tuners.freq_step * policy->max) / 100; + + /* max freq cannot be less than 100. But who knows.... */ + if (unlikely(freq_target == 0)) + freq_target = 5; + + dbs_info->requested_freq += freq_target; + if (dbs_info->requested_freq > policy->max) + dbs_info->requested_freq = policy->max; + __cpufreq_driver_target(policy, dbs_info->requested_freq, + CPUFREQ_RELATION_H); + return; + } + + /* + * The optimal frequency is the frequency that is the lowest that can + * support the current CPU usage without triggering the up policy. To be + * safe, we focus 10 points under the threshold. + */ + if (load < (cs_tuners.down_threshold - 10)) { + freq_target = (cs_tuners.freq_step * policy->max) / 100; + + dbs_info->requested_freq -= freq_target; + if (dbs_info->requested_freq < policy->min) + dbs_info->requested_freq = policy->min; + + /* + * if we cannot reduce the frequency anymore, break out early + */ + if (policy->cur == policy->min) + return; + + __cpufreq_driver_target(policy, dbs_info->requested_freq, + CPUFREQ_RELATION_H); + return; + } +} + +static void cs_dbs_timer(struct work_struct *work) +{ + struct cs_cpu_dbs_info_s *dbs_info = container_of(work, + struct cs_cpu_dbs_info_s, cdbs.work.work); + unsigned int cpu = dbs_info->cdbs.cpu; + int delay = delay_for_sampling_rate(cs_tuners.sampling_rate); + + mutex_lock(&dbs_info->cdbs.timer_mutex); + + dbs_check_cpu(&cs_dbs_data, cpu); + + schedule_delayed_work_on(cpu, &dbs_info->cdbs.work, delay); + mutex_unlock(&dbs_info->cdbs.timer_mutex); +} + +static int dbs_cpufreq_notifier(struct notifier_block *nb, unsigned long val, + void *data) +{ + struct cpufreq_freqs *freq = data; + struct cs_cpu_dbs_info_s *dbs_info = + &per_cpu(cs_cpu_dbs_info, freq->cpu); struct cpufreq_policy *policy; - if (!this_dbs_info->enable) + if (!dbs_info->enable) return 0; - policy = this_dbs_info->cur_policy; + policy = dbs_info->cdbs.cur_policy; /* - * we only care if our internally tracked freq moves outside - * the 'valid' ranges of freqency available to us otherwise - * we do not change it + * we only care if our internally tracked freq moves outside the 'valid' + * ranges of freqency available to us otherwise we do not change it */ - if (this_dbs_info->requested_freq > policy->max - || this_dbs_info->requested_freq < policy->min) - this_dbs_info->requested_freq = freq->new; + if (dbs_info->requested_freq > policy->max + || dbs_info->requested_freq < policy->min) + dbs_info->requested_freq = freq->new; return 0; } -static struct notifier_block dbs_cpufreq_notifier_block = { - .notifier_call = dbs_cpufreq_notifier -}; - /************************** sysfs interface ************************/ static ssize_t show_sampling_rate_min(struct kobject *kobj, struct attribute *attr, char *buf) { - return sprintf(buf, "%u\n", min_sampling_rate); + return sprintf(buf, "%u\n", cs_dbs_data.min_sampling_rate); } -define_one_global_ro(sampling_rate_min); - -/* cpufreq_conservative Governor Tunables */ -#define show_one(file_name, object) \ -static ssize_t show_##file_name \ -(struct kobject *kobj, struct attribute *attr, char *buf) \ -{ \ - return sprintf(buf, "%u\n", dbs_tuners_ins.object); \ -} -show_one(sampling_rate, sampling_rate); -show_one(sampling_down_factor, sampling_down_factor); -show_one(up_threshold, up_threshold); -show_one(down_threshold, down_threshold); -show_one(ignore_nice_load, ignore_nice); -show_one(freq_step, freq_step); - static ssize_t store_sampling_down_factor(struct kobject *a, struct attribute *b, const char *buf, size_t count) @@ -161,7 +168,7 @@ static ssize_t store_sampling_down_factor(struct kobject *a, if (ret != 1 || input > MAX_SAMPLING_DOWN_FACTOR || input < 1) return -EINVAL; - dbs_tuners_ins.sampling_down_factor = input; + cs_tuners.sampling_down_factor = input; return count; } @@ -175,7 +182,7 @@ static ssize_t store_sampling_rate(struct kobject *a, struct attribute *b, if (ret != 1) return -EINVAL; - dbs_tuners_ins.sampling_rate = max(input, min_sampling_rate); + cs_tuners.sampling_rate = max(input, cs_dbs_data.min_sampling_rate); return count; } @@ -186,11 +193,10 @@ static ssize_t store_up_threshold(struct kobject *a, struct attribute *b, int ret; ret = sscanf(buf, "%u", &input); - if (ret != 1 || input > 100 || - input <= dbs_tuners_ins.down_threshold) + if (ret != 1 || input > 100 || input <= cs_tuners.down_threshold) return -EINVAL; - dbs_tuners_ins.up_threshold = input; + cs_tuners.up_threshold = input; return count; } @@ -203,21 +209,19 @@ static ssize_t store_down_threshold(struct kobject *a, struct attribute *b, /* cannot be lower than 11 otherwise freq will not fall */ if (ret != 1 || input < 11 || input > 100 || - input >= dbs_tuners_ins.up_threshold) + input >= cs_tuners.up_threshold) return -EINVAL; - dbs_tuners_ins.down_threshold = input; + cs_tuners.down_threshold = input; return count; } static ssize_t store_ignore_nice_load(struct kobject *a, struct attribute *b, const char *buf, size_t count) { - unsigned int input; + unsigned int input, j; int ret; - unsigned int j; - ret = sscanf(buf, "%u", &input); if (ret != 1) return -EINVAL; @@ -225,19 +229,20 @@ static ssize_t store_ignore_nice_load(struct kobject *a, struct attribute *b, if (input > 1) input = 1; - if (input == dbs_tuners_ins.ignore_nice) /* nothing to do */ + if (input == cs_tuners.ignore_nice) /* nothing to do */ return count; - dbs_tuners_ins.ignore_nice = input; + cs_tuners.ignore_nice = input; /* we need to re-evaluate prev_cpu_idle */ for_each_online_cpu(j) { - struct cpu_dbs_info_s *dbs_info; + struct cs_cpu_dbs_info_s *dbs_info; dbs_info = &per_cpu(cs_cpu_dbs_info, j); - 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 = kcpustat_cpu(j).cpustat[CPUTIME_NICE]; + dbs_info->cdbs.prev_cpu_idle = get_cpu_idle_time(j, + &dbs_info->cdbs.prev_cpu_wall); + if (cs_tuners.ignore_nice) + dbs_info->cdbs.prev_cpu_nice = + kcpustat_cpu(j).cpustat[CPUTIME_NICE]; } return count; } @@ -255,18 +260,28 @@ static ssize_t store_freq_step(struct kobject *a, struct attribute *b, if (input > 100) input = 100; - /* no need to test here if freq_step is zero as the user might actually - * want this, they would be crazy though :) */ - dbs_tuners_ins.freq_step = input; + /* + * no need to test here if freq_step is zero as the user might actually + * want this, they would be crazy though :) + */ + cs_tuners.freq_step = input; return count; } +show_one(cs, sampling_rate, sampling_rate); +show_one(cs, sampling_down_factor, sampling_down_factor); +show_one(cs, up_threshold, up_threshold); +show_one(cs, down_threshold, down_threshold); +show_one(cs, ignore_nice_load, ignore_nice); +show_one(cs, freq_step, freq_step); + define_one_global_rw(sampling_rate); define_one_global_rw(sampling_down_factor); define_one_global_rw(up_threshold); define_one_global_rw(down_threshold); define_one_global_rw(ignore_nice_load); define_one_global_rw(freq_step); +define_one_global_ro(sampling_rate_min); static struct attribute *dbs_attributes[] = { &sampling_rate_min.attr, @@ -279,283 +294,38 @@ static struct attribute *dbs_attributes[] = { NULL }; -static struct attribute_group dbs_attr_group = { +static struct attribute_group cs_attr_group = { .attrs = dbs_attributes, .name = "conservative", }; /************************** sysfs end ************************/ -static void dbs_check_cpu(struct cpu_dbs_info_s *this_dbs_info) -{ - unsigned int load = 0; - unsigned int max_load = 0; - unsigned int freq_target; - - struct cpufreq_policy *policy; - unsigned int j; - - policy = this_dbs_info->cur_policy; - - /* - * Every sampling_rate, we check, if current idle time is less - * than 20% (default), then we try to increase frequency - * Every sampling_rate*sampling_down_factor, we check, if current - * idle time is more than 80%, then we try to decrease frequency - * - * Any frequency increase takes it to the maximum frequency. - * Frequency reduction happens at minimum steps of - * 5% (default) of maximum frequency - */ - - /* Get Absolute Load */ - 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; - - j_dbs_info = &per_cpu(cs_cpu_dbs_info, j); - - cur_idle_time = get_cpu_idle_time(j, &cur_wall_time); - - wall_time = (unsigned int) - (cur_wall_time - j_dbs_info->prev_cpu_wall); - j_dbs_info->prev_cpu_wall = cur_wall_time; - - idle_time = (unsigned int) - (cur_idle_time - j_dbs_info->prev_cpu_idle); - j_dbs_info->prev_cpu_idle = cur_idle_time; - - if (dbs_tuners_ins.ignore_nice) { - u64 cur_nice; - unsigned long cur_nice_jiffies; - - cur_nice = kcpustat_cpu(j).cpustat[CPUTIME_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 = kcpustat_cpu(j).cpustat[CPUTIME_NICE]; - idle_time += jiffies_to_usecs(cur_nice_jiffies); - } - - if (unlikely(!wall_time || wall_time < idle_time)) - continue; - - load = 100 * (wall_time - idle_time) / wall_time; - - if (load > max_load) - max_load = load; - } +define_get_cpu_dbs_routines(cs_cpu_dbs_info); - /* - * break out if we 'cannot' reduce the speed as the user might - * want freq_step to be zero - */ - if (dbs_tuners_ins.freq_step == 0) - return; - - /* Check for frequency increase */ - if (max_load > dbs_tuners_ins.up_threshold) { - this_dbs_info->down_skip = 0; - - /* if we are already at full speed then break out early */ - if (this_dbs_info->requested_freq == policy->max) - return; - - freq_target = (dbs_tuners_ins.freq_step * policy->max) / 100; - - /* max freq cannot be less than 100. But who knows.... */ - if (unlikely(freq_target == 0)) - freq_target = 5; - - this_dbs_info->requested_freq += freq_target; - if (this_dbs_info->requested_freq > policy->max) - this_dbs_info->requested_freq = policy->max; - - __cpufreq_driver_target(policy, this_dbs_info->requested_freq, - CPUFREQ_RELATION_H); - return; - } - - /* - * The optimal frequency is the frequency that is the lowest that - * can support the current CPU usage without triggering the up - * policy. To be safe, we focus 10 points under the threshold. - */ - if (max_load < (dbs_tuners_ins.down_threshold - 10)) { - freq_target = (dbs_tuners_ins.freq_step * policy->max) / 100; - - this_dbs_info->requested_freq -= freq_target; - if (this_dbs_info->requested_freq < policy->min) - this_dbs_info->requested_freq = policy->min; - - /* - * if we cannot reduce the frequency anymore, break out early - */ - if (policy->cur == policy->min) - return; - - __cpufreq_driver_target(policy, this_dbs_info->requested_freq, - CPUFREQ_RELATION_H); - return; - } -} - -static void do_dbs_timer(struct work_struct *work) -{ - struct cpu_dbs_info_s *dbs_info = - container_of(work, struct cpu_dbs_info_s, work.work); - unsigned int 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; - - mutex_lock(&dbs_info->timer_mutex); - - dbs_check_cpu(dbs_info); - - schedule_delayed_work_on(cpu, &dbs_info->work, delay); - mutex_unlock(&dbs_info->timer_mutex); -} - -static inline void dbs_timer_init(struct cpu_dbs_info_s *dbs_info) -{ - /* We want all CPUs to do sampling nearly on same jiffy */ - int delay = usecs_to_jiffies(dbs_tuners_ins.sampling_rate); - delay -= jiffies % delay; +static struct notifier_block cs_cpufreq_notifier_block = { + .notifier_call = dbs_cpufreq_notifier, +}; - dbs_info->enable = 1; - INIT_DEFERRABLE_WORK(&dbs_info->work, do_dbs_timer); - schedule_delayed_work_on(dbs_info->cpu, &dbs_info->work, delay); -} +static struct cs_ops cs_ops = { + .notifier_block = &cs_cpufreq_notifier_block, +}; -static inline void dbs_timer_exit(struct cpu_dbs_info_s *dbs_info) -{ - dbs_info->enable = 0; - cancel_delayed_work_sync(&dbs_info->work); -} +static struct dbs_data cs_dbs_data = { + .governor = GOV_CONSERVATIVE, + .attr_group = &cs_attr_group, + .tuners = &cs_tuners, + .get_cpu_cdbs = get_cpu_cdbs, + .get_cpu_dbs_info_s = get_cpu_dbs_info_s, + .gov_dbs_timer = cs_dbs_timer, + .gov_check_cpu = cs_check_cpu, + .gov_ops = &cs_ops, +}; -static int cpufreq_governor_dbs(struct cpufreq_policy *policy, +static int cs_cpufreq_governor_dbs(struct cpufreq_policy *policy, unsigned int event) { - unsigned int cpu = policy->cpu; - struct cpu_dbs_info_s *this_dbs_info; - unsigned int j; - int rc; - - this_dbs_info = &per_cpu(cs_cpu_dbs_info, cpu); - - switch (event) { - case CPUFREQ_GOV_START: - if ((!cpu_online(cpu)) || (!policy->cur)) - return -EINVAL; - - mutex_lock(&dbs_mutex); - - for_each_cpu(j, policy->cpus) { - struct cpu_dbs_info_s *j_dbs_info; - j_dbs_info = &per_cpu(cs_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); - if (dbs_tuners_ins.ignore_nice) - j_dbs_info->prev_cpu_nice = - kcpustat_cpu(j).cpustat[CPUTIME_NICE]; - } - this_dbs_info->cpu = cpu; - this_dbs_info->down_skip = 0; - this_dbs_info->requested_freq = policy->cur; - - mutex_init(&this_dbs_info->timer_mutex); - dbs_enable++; - /* - * Start the timerschedule work, when this governor - * is used for first time - */ - if (dbs_enable == 1) { - unsigned int latency; - /* policy latency is in nS. Convert it to uS first */ - latency = policy->cpuinfo.transition_latency / 1000; - if (latency == 0) - latency = 1; - - rc = sysfs_create_group(cpufreq_global_kobject, - &dbs_attr_group); - if (rc) { - mutex_unlock(&dbs_mutex); - return rc; - } - - /* - * conservative does not implement micro like ondemand - * governor, thus we are bound to jiffes/HZ - */ - min_sampling_rate = - MIN_SAMPLING_RATE_RATIO * jiffies_to_usecs(10); - /* Bring kernel and HW constraints together */ - min_sampling_rate = max(min_sampling_rate, - MIN_LATENCY_MULTIPLIER * latency); - dbs_tuners_ins.sampling_rate = - max(min_sampling_rate, - latency * LATENCY_MULTIPLIER); - - cpufreq_register_notifier( - &dbs_cpufreq_notifier_block, - CPUFREQ_TRANSITION_NOTIFIER); - } - mutex_unlock(&dbs_mutex); - - dbs_timer_init(this_dbs_info); - - break; - - case CPUFREQ_GOV_STOP: - dbs_timer_exit(this_dbs_info); - - mutex_lock(&dbs_mutex); - dbs_enable--; - mutex_destroy(&this_dbs_info->timer_mutex); - - /* - * Stop the timerschedule work, when this governor - * is used for first time - */ - if (dbs_enable == 0) - cpufreq_unregister_notifier( - &dbs_cpufreq_notifier_block, - CPUFREQ_TRANSITION_NOTIFIER); - - mutex_unlock(&dbs_mutex); - if (!dbs_enable) - sysfs_remove_group(cpufreq_global_kobject, - &dbs_attr_group); - - break; - - case CPUFREQ_GOV_LIMITS: - mutex_lock(&this_dbs_info->timer_mutex); - if (policy->max < this_dbs_info->cur_policy->cur) - __cpufreq_driver_target( - this_dbs_info->cur_policy, - 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); - dbs_check_cpu(this_dbs_info); - mutex_unlock(&this_dbs_info->timer_mutex); - - break; - } - return 0; + return cpufreq_governor_dbs(&cs_dbs_data, policy, event); } #ifndef CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE @@ -563,13 +333,14 @@ static #endif struct cpufreq_governor cpufreq_gov_conservative = { .name = "conservative", - .governor = cpufreq_governor_dbs, + .governor = cs_cpufreq_governor_dbs, .max_transition_latency = TRANSITION_LATENCY_LIMIT, .owner = THIS_MODULE, }; static int __init cpufreq_gov_dbs_init(void) { + mutex_init(&cs_dbs_data.mutex); return cpufreq_register_governor(&cpufreq_gov_conservative); } @@ -578,7 +349,6 @@ static void __exit cpufreq_gov_dbs_exit(void) cpufreq_unregister_governor(&cpufreq_gov_conservative); } - MODULE_AUTHOR("Alexander Clouter <alex@digriz.org.uk>"); MODULE_DESCRIPTION("'cpufreq_conservative' - A dynamic cpufreq governor for " "Low Latency Frequency Transition capable processors " |