diff options
Diffstat (limited to 'drivers/cpufreq/cpufreq_ondemand.c')
| -rw-r--r-- | drivers/cpufreq/cpufreq_ondemand.c | 985 |
1 files changed, 371 insertions, 614 deletions
diff --git a/drivers/cpufreq/cpufreq_ondemand.c b/drivers/cpufreq/cpufreq_ondemand.c index c631f27a3dc..18d40918909 100644 --- a/drivers/cpufreq/cpufreq_ondemand.c +++ b/drivers/cpufreq/cpufreq_ondemand.c @@ -10,159 +10,62 @@ * published by the Free Software Foundation. */ -#include <linux/kernel.h> -#include <linux/module.h> -#include <linux/init.h> -#include <linux/cpufreq.h> +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + #include <linux/cpu.h> -#include <linux/jiffies.h> -#include <linux/kernel_stat.h> -#include <linux/mutex.h> -#include <linux/hrtimer.h> +#include <linux/percpu-defs.h> +#include <linux/slab.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 - */ +#include "cpufreq_governor.h" -#define DEF_FREQUENCY_DOWN_DIFFERENTIAL (10) +/* On-demand governor macros */ #define DEF_FREQUENCY_UP_THRESHOLD (80) #define DEF_SAMPLING_DOWN_FACTOR (1) #define MAX_SAMPLING_DOWN_FACTOR (100000) -#define MICRO_FREQUENCY_DOWN_DIFFERENTIAL (3) #define MICRO_FREQUENCY_UP_THRESHOLD (95) #define MICRO_FREQUENCY_MIN_SAMPLE_RATE (10000) #define MIN_FREQUENCY_UP_THRESHOLD (11) #define MAX_FREQUENCY_UP_THRESHOLD (100) -/* - * 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 TRANSITION_LATENCY_LIMIT (10 * 1000 * 1000) +static DEFINE_PER_CPU(struct od_cpu_dbs_info_s, od_cpu_dbs_info); -static void do_dbs_timer(struct work_struct *work); -static int cpufreq_governor_dbs(struct cpufreq_policy *policy, - unsigned int event); +static struct od_ops od_ops; #ifndef CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND -static +static struct cpufreq_governor cpufreq_gov_ondemand; #endif -struct cpufreq_governor cpufreq_gov_ondemand = { - .name = "ondemand", - .governor = cpufreq_governor_dbs, - .max_transition_latency = TRANSITION_LATENCY_LIMIT, - .owner = THIS_MODULE, -}; - -/* Sampling types */ -enum {DBS_NORMAL_SAMPLE, DBS_SUB_SAMPLE}; - -struct cpu_dbs_info_s { - cputime64_t prev_cpu_idle; - cputime64_t prev_cpu_iowait; - cputime64_t prev_cpu_wall; - cputime64_t prev_cpu_nice; - struct cpufreq_policy *cur_policy; - struct delayed_work work; - struct cpufreq_frequency_table *freq_table; - unsigned int freq_lo; - unsigned int freq_lo_jiffies; - unsigned int freq_hi_jiffies; - unsigned int rate_mult; - int cpu; - unsigned int sample_type: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, od_cpu_dbs_info); -static unsigned int dbs_enable; /* number of CPUs using this policy */ +static unsigned int default_powersave_bias; -/* - * dbs_mutex protects data in dbs_tuners_ins from concurrent changes on - * different CPUs. It protects dbs_enable in governor start/stop. - */ -static DEFINE_MUTEX(dbs_mutex); - -static struct workqueue_struct *kondemand_wq; - -static struct dbs_tuners { - unsigned int sampling_rate; - unsigned int up_threshold; - unsigned int down_differential; - unsigned int ignore_nice; - unsigned int sampling_down_factor; - unsigned int powersave_bias; - unsigned int io_is_busy; -} dbs_tuners_ins = { - .up_threshold = DEF_FREQUENCY_UP_THRESHOLD, - .sampling_down_factor = DEF_SAMPLING_DOWN_FACTOR, - .down_differential = DEF_FREQUENCY_DOWN_DIFFERENTIAL, - .ignore_nice = 0, - .powersave_bias = 0, -}; - -static inline cputime64_t get_cpu_idle_time_jiffy(unsigned int cpu, - cputime64_t *wall) -{ - 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); - - 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); - - idle_time = cputime64_sub(cur_wall_time, busy_time); - if (wall) - *wall = (cputime64_t)jiffies_to_usecs(cur_wall_time); - - return (cputime64_t)jiffies_to_usecs(idle_time); -} - -static inline cputime64_t get_cpu_idle_time(unsigned int cpu, cputime64_t *wall) +static void ondemand_powersave_bias_init_cpu(int cpu) { - u64 idle_time = get_cpu_idle_time_us(cpu, wall); + struct od_cpu_dbs_info_s *dbs_info = &per_cpu(od_cpu_dbs_info, cpu); - if (idle_time == -1ULL) - return get_cpu_idle_time_jiffy(cpu, wall); - - return idle_time; + dbs_info->freq_table = cpufreq_frequency_get_table(cpu); + dbs_info->freq_lo = 0; } -static inline cputime64_t get_cpu_iowait_time(unsigned int cpu, cputime64_t *wall) +/* + * Not all CPUs want IO time to be accounted as busy; this depends on how + * efficient idling at a higher frequency/voltage is. + * Pavel Machek says this is not so for various generations of AMD and old + * Intel systems. + * Mike Chan (android.com) claims this is also not true for ARM. + * Because of this, whitelist specific known (series) of CPUs by default, and + * leave all others up to the user. + */ +static int should_io_be_busy(void) { - u64 iowait_time = get_cpu_iowait_time_us(cpu, wall); - - if (iowait_time == -1ULL) - return 0; - - return iowait_time; +#if defined(CONFIG_X86) + /* + * For Intel, Core 2 (model 15) and later have an efficient idle. + */ + if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL && + boot_cpu_data.x86 == 6 && + boot_cpu_data.x86_model >= 15) + return 1; +#endif + return 0; } /* @@ -170,16 +73,17 @@ static inline cputime64_t get_cpu_iowait_time(unsigned int cpu, cputime64_t *wal * Returns the freq_hi to be used right now and will set freq_hi_jiffies, * freq_lo, and freq_lo_jiffies in percpu area for averaging freqs. */ -static unsigned int powersave_bias_target(struct cpufreq_policy *policy, - unsigned int freq_next, - unsigned int relation) +static unsigned int generic_powersave_bias_target(struct cpufreq_policy *policy, + unsigned int freq_next, unsigned int relation) { unsigned int freq_req, freq_reduc, freq_avg; unsigned int freq_hi, freq_lo; unsigned int index = 0; unsigned int jiffies_total, jiffies_hi, jiffies_lo; - struct cpu_dbs_info_s *dbs_info = &per_cpu(od_cpu_dbs_info, + struct od_cpu_dbs_info_s *dbs_info = &per_cpu(od_cpu_dbs_info, policy->cpu); + struct dbs_data *dbs_data = policy->governor_data; + struct od_dbs_tuners *od_tuners = dbs_data->tuners; if (!dbs_info->freq_table) { dbs_info->freq_lo = 0; @@ -190,7 +94,7 @@ static unsigned int powersave_bias_target(struct cpufreq_policy *policy, cpufreq_frequency_table_target(policy, dbs_info->freq_table, freq_next, relation, &index); freq_req = dbs_info->freq_table[index].frequency; - freq_reduc = freq_req * dbs_tuners_ins.powersave_bias / 1000; + freq_reduc = freq_req * od_tuners->powersave_bias / 1000; freq_avg = freq_req - freq_reduc; /* Find freq bounds for freq_avg in freq_table */ @@ -209,7 +113,7 @@ static unsigned int powersave_bias_target(struct cpufreq_policy *policy, dbs_info->freq_lo_jiffies = 0; return freq_lo; } - jiffies_total = usecs_to_jiffies(dbs_tuners_ins.sampling_rate); + jiffies_total = usecs_to_jiffies(od_tuners->sampling_rate); jiffies_hi = (freq_avg - freq_lo) * jiffies_total; jiffies_hi += ((freq_hi - freq_lo) / 2); jiffies_hi /= (freq_hi - freq_lo); @@ -220,13 +124,6 @@ static unsigned int powersave_bias_target(struct cpufreq_policy *policy, return freq_hi; } -static void ondemand_powersave_bias_init_cpu(int cpu) -{ - struct cpu_dbs_info_s *dbs_info = &per_cpu(od_cpu_dbs_info, cpu); - dbs_info->freq_table = cpufreq_frequency_get_table(cpu); - dbs_info->freq_lo = 0; -} - static void ondemand_powersave_bias_init(void) { int i; @@ -235,67 +132,170 @@ static void ondemand_powersave_bias_init(void) } } -/************************** sysfs interface ************************/ - -static ssize_t show_sampling_rate_max(struct kobject *kobj, - struct attribute *attr, char *buf) +static void dbs_freq_increase(struct cpufreq_policy *policy, unsigned int freq) { - printk_once(KERN_INFO "CPUFREQ: ondemand sampling_rate_max " - "sysfs file is deprecated - used by: %s\n", current->comm); - return sprintf(buf, "%u\n", -1U); + struct dbs_data *dbs_data = policy->governor_data; + struct od_dbs_tuners *od_tuners = dbs_data->tuners; + + if (od_tuners->powersave_bias) + freq = od_ops.powersave_bias_target(policy, freq, + CPUFREQ_RELATION_H); + else if (policy->cur == policy->max) + return; + + __cpufreq_driver_target(policy, freq, od_tuners->powersave_bias ? + CPUFREQ_RELATION_L : CPUFREQ_RELATION_H); } -static ssize_t show_sampling_rate_min(struct kobject *kobj, - struct attribute *attr, char *buf) +/* + * Every sampling_rate, we check, if current idle time is less than 20% + * (default), then we try to increase frequency. Else, we adjust the frequency + * proportional to load. + */ +static void od_check_cpu(int cpu, unsigned int load) { - return sprintf(buf, "%u\n", min_sampling_rate); -} + struct od_cpu_dbs_info_s *dbs_info = &per_cpu(od_cpu_dbs_info, cpu); + struct cpufreq_policy *policy = dbs_info->cdbs.cur_policy; + struct dbs_data *dbs_data = policy->governor_data; + struct od_dbs_tuners *od_tuners = dbs_data->tuners; -define_one_global_ro(sampling_rate_max); -define_one_global_ro(sampling_rate_min); + dbs_info->freq_lo = 0; + + /* Check for frequency increase */ + if (load > od_tuners->up_threshold) { + /* If switching to max speed, apply sampling_down_factor */ + if (policy->cur < policy->max) + dbs_info->rate_mult = + od_tuners->sampling_down_factor; + dbs_freq_increase(policy, policy->max); + } else { + /* Calculate the next frequency proportional to load */ + unsigned int freq_next; + freq_next = load * policy->cpuinfo.max_freq / 100; -/* cpufreq_ondemand 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); \ + /* No longer fully busy, reset rate_mult */ + dbs_info->rate_mult = 1; + + if (!od_tuners->powersave_bias) { + __cpufreq_driver_target(policy, freq_next, + CPUFREQ_RELATION_L); + return; + } + + freq_next = od_ops.powersave_bias_target(policy, freq_next, + CPUFREQ_RELATION_L); + __cpufreq_driver_target(policy, freq_next, CPUFREQ_RELATION_L); + } } -show_one(sampling_rate, sampling_rate); -show_one(io_is_busy, io_is_busy); -show_one(up_threshold, up_threshold); -show_one(sampling_down_factor, sampling_down_factor); -show_one(ignore_nice_load, ignore_nice); -show_one(powersave_bias, powersave_bias); - -/*** delete after deprecation time ***/ - -#define DEPRECATION_MSG(file_name) \ - printk_once(KERN_INFO "CPUFREQ: Per core ondemand sysfs " \ - "interface is deprecated - " #file_name "\n"); - -#define show_one_old(file_name) \ -static ssize_t show_##file_name##_old \ -(struct cpufreq_policy *unused, char *buf) \ -{ \ - printk_once(KERN_INFO "CPUFREQ: Per core ondemand sysfs " \ - "interface is deprecated - " #file_name "\n"); \ - return show_##file_name(NULL, NULL, buf); \ + +static void od_dbs_timer(struct work_struct *work) +{ + struct od_cpu_dbs_info_s *dbs_info = + container_of(work, struct od_cpu_dbs_info_s, cdbs.work.work); + unsigned int cpu = dbs_info->cdbs.cur_policy->cpu; + struct od_cpu_dbs_info_s *core_dbs_info = &per_cpu(od_cpu_dbs_info, + cpu); + struct dbs_data *dbs_data = dbs_info->cdbs.cur_policy->governor_data; + struct od_dbs_tuners *od_tuners = dbs_data->tuners; + int delay = 0, sample_type = core_dbs_info->sample_type; + bool modify_all = true; + + mutex_lock(&core_dbs_info->cdbs.timer_mutex); + if (!need_load_eval(&core_dbs_info->cdbs, od_tuners->sampling_rate)) { + modify_all = false; + goto max_delay; + } + + /* Common NORMAL_SAMPLE setup */ + core_dbs_info->sample_type = OD_NORMAL_SAMPLE; + if (sample_type == OD_SUB_SAMPLE) { + delay = core_dbs_info->freq_lo_jiffies; + __cpufreq_driver_target(core_dbs_info->cdbs.cur_policy, + core_dbs_info->freq_lo, CPUFREQ_RELATION_H); + } else { + dbs_check_cpu(dbs_data, cpu); + if (core_dbs_info->freq_lo) { + /* Setup timer for SUB_SAMPLE */ + core_dbs_info->sample_type = OD_SUB_SAMPLE; + delay = core_dbs_info->freq_hi_jiffies; + } + } + +max_delay: + if (!delay) + delay = delay_for_sampling_rate(od_tuners->sampling_rate + * core_dbs_info->rate_mult); + + gov_queue_work(dbs_data, dbs_info->cdbs.cur_policy, delay, modify_all); + mutex_unlock(&core_dbs_info->cdbs.timer_mutex); } -show_one_old(sampling_rate); -show_one_old(up_threshold); -show_one_old(ignore_nice_load); -show_one_old(powersave_bias); -show_one_old(sampling_rate_min); -show_one_old(sampling_rate_max); -cpufreq_freq_attr_ro_old(sampling_rate_min); -cpufreq_freq_attr_ro_old(sampling_rate_max); +/************************** sysfs interface ************************/ +static struct common_dbs_data od_dbs_cdata; -/*** delete after deprecation time ***/ +/** + * update_sampling_rate - update sampling rate effective immediately if needed. + * @new_rate: new sampling rate + * + * If new rate is smaller than the old, simply updating + * dbs_tuners_int.sampling_rate might not be appropriate. For example, if the + * original sampling_rate was 1 second and the requested new sampling rate is 10 + * ms because the user needs immediate reaction from ondemand governor, but not + * sure if higher frequency will be required or not, then, the governor may + * change the sampling rate too late; up to 1 second later. Thus, if we are + * reducing the sampling rate, we need to make the new value effective + * immediately. + */ +static void update_sampling_rate(struct dbs_data *dbs_data, + unsigned int new_rate) +{ + struct od_dbs_tuners *od_tuners = dbs_data->tuners; + int cpu; + + od_tuners->sampling_rate = new_rate = max(new_rate, + dbs_data->min_sampling_rate); + + for_each_online_cpu(cpu) { + struct cpufreq_policy *policy; + struct od_cpu_dbs_info_s *dbs_info; + unsigned long next_sampling, appointed_at; + + policy = cpufreq_cpu_get(cpu); + if (!policy) + continue; + if (policy->governor != &cpufreq_gov_ondemand) { + cpufreq_cpu_put(policy); + continue; + } + dbs_info = &per_cpu(od_cpu_dbs_info, cpu); + cpufreq_cpu_put(policy); + + mutex_lock(&dbs_info->cdbs.timer_mutex); -static ssize_t store_sampling_rate(struct kobject *a, struct attribute *b, - const char *buf, size_t count) + if (!delayed_work_pending(&dbs_info->cdbs.work)) { + mutex_unlock(&dbs_info->cdbs.timer_mutex); + continue; + } + + next_sampling = jiffies + usecs_to_jiffies(new_rate); + appointed_at = dbs_info->cdbs.work.timer.expires; + + if (time_before(next_sampling, appointed_at)) { + + mutex_unlock(&dbs_info->cdbs.timer_mutex); + cancel_delayed_work_sync(&dbs_info->cdbs.work); + mutex_lock(&dbs_info->cdbs.timer_mutex); + + gov_queue_work(dbs_data, dbs_info->cdbs.cur_policy, + usecs_to_jiffies(new_rate), true); + + } + mutex_unlock(&dbs_info->cdbs.timer_mutex); + } +} + +static ssize_t store_sampling_rate(struct dbs_data *dbs_data, const char *buf, + size_t count) { unsigned int input; int ret; @@ -303,33 +303,37 @@ static ssize_t store_sampling_rate(struct kobject *a, struct attribute *b, if (ret != 1) return -EINVAL; - mutex_lock(&dbs_mutex); - dbs_tuners_ins.sampling_rate = max(input, min_sampling_rate); - mutex_unlock(&dbs_mutex); - + update_sampling_rate(dbs_data, input); return count; } -static ssize_t store_io_is_busy(struct kobject *a, struct attribute *b, - const char *buf, size_t count) +static ssize_t store_io_is_busy(struct dbs_data *dbs_data, const char *buf, + size_t count) { + struct od_dbs_tuners *od_tuners = dbs_data->tuners; unsigned int input; int ret; + unsigned int j; ret = sscanf(buf, "%u", &input); if (ret != 1) return -EINVAL; + od_tuners->io_is_busy = !!input; - mutex_lock(&dbs_mutex); - dbs_tuners_ins.io_is_busy = !!input; - mutex_unlock(&dbs_mutex); - + /* we need to re-evaluate prev_cpu_idle */ + for_each_online_cpu(j) { + struct od_cpu_dbs_info_s *dbs_info = &per_cpu(od_cpu_dbs_info, + j); + dbs_info->cdbs.prev_cpu_idle = get_cpu_idle_time(j, + &dbs_info->cdbs.prev_cpu_wall, od_tuners->io_is_busy); + } return count; } -static ssize_t store_up_threshold(struct kobject *a, struct attribute *b, - const char *buf, size_t count) +static ssize_t store_up_threshold(struct dbs_data *dbs_data, const char *buf, + size_t count) { + struct od_dbs_tuners *od_tuners = dbs_data->tuners; unsigned int input; int ret; ret = sscanf(buf, "%u", &input); @@ -339,39 +343,35 @@ static ssize_t store_up_threshold(struct kobject *a, struct attribute *b, return -EINVAL; } - mutex_lock(&dbs_mutex); - dbs_tuners_ins.up_threshold = input; - mutex_unlock(&dbs_mutex); - + od_tuners->up_threshold = input; return count; } -static ssize_t store_sampling_down_factor(struct kobject *a, - struct attribute *b, const char *buf, size_t count) +static ssize_t store_sampling_down_factor(struct dbs_data *dbs_data, + const char *buf, size_t count) { + struct od_dbs_tuners *od_tuners = dbs_data->tuners; unsigned int input, j; int ret; ret = sscanf(buf, "%u", &input); if (ret != 1 || input > MAX_SAMPLING_DOWN_FACTOR || input < 1) return -EINVAL; - mutex_lock(&dbs_mutex); - dbs_tuners_ins.sampling_down_factor = input; + od_tuners->sampling_down_factor = input; /* Reset down sampling multiplier in case it was active */ for_each_online_cpu(j) { - struct cpu_dbs_info_s *dbs_info; - dbs_info = &per_cpu(od_cpu_dbs_info, j); + struct od_cpu_dbs_info_s *dbs_info = &per_cpu(od_cpu_dbs_info, + j); dbs_info->rate_mult = 1; } - mutex_unlock(&dbs_mutex); - return count; } -static ssize_t store_ignore_nice_load(struct kobject *a, struct attribute *b, - const char *buf, size_t count) +static ssize_t store_ignore_nice_load(struct dbs_data *dbs_data, + const char *buf, size_t count) { + struct od_dbs_tuners *od_tuners = dbs_data->tuners; unsigned int input; int ret; @@ -384,31 +384,29 @@ static ssize_t store_ignore_nice_load(struct kobject *a, struct attribute *b, if (input > 1) input = 1; - mutex_lock(&dbs_mutex); - if (input == dbs_tuners_ins.ignore_nice) { /* nothing to do */ - mutex_unlock(&dbs_mutex); + if (input == od_tuners->ignore_nice_load) { /* nothing to do */ return count; } - dbs_tuners_ins.ignore_nice = input; + od_tuners->ignore_nice_load = input; /* we need to re-evaluate prev_cpu_idle */ for_each_online_cpu(j) { - struct cpu_dbs_info_s *dbs_info; + struct od_cpu_dbs_info_s *dbs_info; dbs_info = &per_cpu(od_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 = kstat_cpu(j).cpustat.nice; + dbs_info->cdbs.prev_cpu_idle = get_cpu_idle_time(j, + &dbs_info->cdbs.prev_cpu_wall, od_tuners->io_is_busy); + if (od_tuners->ignore_nice_load) + dbs_info->cdbs.prev_cpu_nice = + kcpustat_cpu(j).cpustat[CPUTIME_NICE]; } - mutex_unlock(&dbs_mutex); - return count; } -static ssize_t store_powersave_bias(struct kobject *a, struct attribute *b, - const char *buf, size_t count) +static ssize_t store_powersave_bias(struct dbs_data *dbs_data, const char *buf, + size_t count) { + struct od_dbs_tuners *od_tuners = dbs_data->tuners; unsigned int input; int ret; ret = sscanf(buf, "%u", &input); @@ -419,444 +417,203 @@ static ssize_t store_powersave_bias(struct kobject *a, struct attribute *b, if (input > 1000) input = 1000; - mutex_lock(&dbs_mutex); - dbs_tuners_ins.powersave_bias = input; + od_tuners->powersave_bias = input; ondemand_powersave_bias_init(); - mutex_unlock(&dbs_mutex); - return count; } -define_one_global_rw(sampling_rate); -define_one_global_rw(io_is_busy); -define_one_global_rw(up_threshold); -define_one_global_rw(sampling_down_factor); -define_one_global_rw(ignore_nice_load); -define_one_global_rw(powersave_bias); - -static struct attribute *dbs_attributes[] = { - &sampling_rate_max.attr, - &sampling_rate_min.attr, - &sampling_rate.attr, - &up_threshold.attr, - &sampling_down_factor.attr, - &ignore_nice_load.attr, - &powersave_bias.attr, - &io_is_busy.attr, +show_store_one(od, sampling_rate); +show_store_one(od, io_is_busy); +show_store_one(od, up_threshold); +show_store_one(od, sampling_down_factor); +show_store_one(od, ignore_nice_load); +show_store_one(od, powersave_bias); +declare_show_sampling_rate_min(od); + +gov_sys_pol_attr_rw(sampling_rate); +gov_sys_pol_attr_rw(io_is_busy); +gov_sys_pol_attr_rw(up_threshold); +gov_sys_pol_attr_rw(sampling_down_factor); +gov_sys_pol_attr_rw(ignore_nice_load); +gov_sys_pol_attr_rw(powersave_bias); +gov_sys_pol_attr_ro(sampling_rate_min); + +static struct attribute *dbs_attributes_gov_sys[] = { + &sampling_rate_min_gov_sys.attr, + &sampling_rate_gov_sys.attr, + &up_threshold_gov_sys.attr, + &sampling_down_factor_gov_sys.attr, + &ignore_nice_load_gov_sys.attr, + &powersave_bias_gov_sys.attr, + &io_is_busy_gov_sys.attr, NULL }; -static struct attribute_group dbs_attr_group = { - .attrs = dbs_attributes, +static struct attribute_group od_attr_group_gov_sys = { + .attrs = dbs_attributes_gov_sys, .name = "ondemand", }; -/*** delete after deprecation time ***/ - -#define write_one_old(file_name) \ -static ssize_t store_##file_name##_old \ -(struct cpufreq_policy *unused, const char *buf, size_t count) \ -{ \ - printk_once(KERN_INFO "CPUFREQ: Per core ondemand sysfs " \ - "interface is deprecated - " #file_name "\n"); \ - return store_##file_name(NULL, NULL, buf, count); \ -} -write_one_old(sampling_rate); -write_one_old(up_threshold); -write_one_old(ignore_nice_load); -write_one_old(powersave_bias); - -cpufreq_freq_attr_rw_old(sampling_rate); -cpufreq_freq_attr_rw_old(up_threshold); -cpufreq_freq_attr_rw_old(ignore_nice_load); -cpufreq_freq_attr_rw_old(powersave_bias); - -static struct attribute *dbs_attributes_old[] = { - &sampling_rate_max_old.attr, - &sampling_rate_min_old.attr, - &sampling_rate_old.attr, - &up_threshold_old.attr, - &ignore_nice_load_old.attr, - &powersave_bias_old.attr, - NULL +static struct attribute *dbs_attributes_gov_pol[] = { + &sampling_rate_min_gov_pol.attr, + &sampling_rate_gov_pol.attr, + &up_threshold_gov_pol.attr, + &sampling_down_factor_gov_pol.attr, + &ignore_nice_load_gov_pol.attr, + &powersave_bias_gov_pol.attr, + &io_is_busy_gov_pol.attr, + NULL }; -static struct attribute_group dbs_attr_group_old = { - .attrs = dbs_attributes_old, - .name = "ondemand", +static struct attribute_group od_attr_group_gov_pol = { + .attrs = dbs_attributes_gov_pol, + .name = "ondemand", }; -/*** delete after deprecation time ***/ - /************************** sysfs end ************************/ -static void dbs_freq_increase(struct cpufreq_policy *p, unsigned int freq) -{ - if (dbs_tuners_ins.powersave_bias) - freq = powersave_bias_target(p, freq, CPUFREQ_RELATION_H); - else if (p->cur == p->max) - return; - - __cpufreq_driver_target(p, freq, dbs_tuners_ins.powersave_bias ? - CPUFREQ_RELATION_L : CPUFREQ_RELATION_H); -} - -static void dbs_check_cpu(struct cpu_dbs_info_s *this_dbs_info) +static int od_init(struct dbs_data *dbs_data) { - unsigned int max_load_freq; - - struct cpufreq_policy *policy; - unsigned int j; - - this_dbs_info->freq_lo = 0; - 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, we look for a the lowest - * frequency which can sustain the load while keeping idle time over - * 30%. If such a frequency exist, we try to decrease to this frequency. - * - * Any frequency increase takes it to the maximum frequency. - * Frequency reduction happens at minimum steps of - * 5% (default) of current frequency - */ - - /* Get Absolute Load - in terms of freq */ - max_load_freq = 0; + struct od_dbs_tuners *tuners; + u64 idle_time; + int cpu; - for_each_cpu(j, policy->cpus) { - struct cpu_dbs_info_s *j_dbs_info; - cputime64_t cur_wall_time, cur_idle_time, cur_iowait_time; - unsigned int idle_time, wall_time, iowait_time; - unsigned int load, load_freq; - int freq_avg; + tuners = kzalloc(sizeof(*tuners), GFP_KERNEL); + if (!tuners) { + pr_err("%s: kzalloc failed\n", __func__); + return -ENOMEM; + } - j_dbs_info = &per_cpu(od_cpu_dbs_info, j); + cpu = get_cpu(); + idle_time = get_cpu_idle_time_us(cpu, NULL); + put_cpu(); + if (idle_time != -1ULL) { + /* Idle micro accounting is supported. Use finer thresholds */ + tuners->up_threshold = MICRO_FREQUENCY_UP_THRESHOLD; + /* + * In nohz/micro accounting case we set the minimum frequency + * not depending on HZ, but fixed (very low). The deferred + * timer might skip some samples if idle/sleeping as needed. + */ + dbs_data->min_sampling_rate = MICRO_FREQUENCY_MIN_SAMPLE_RATE; + } else { + tuners->up_threshold = DEF_FREQUENCY_UP_THRESHOLD; - cur_idle_time = get_cpu_idle_time(j, &cur_wall_time); - cur_iowait_time = get_cpu_iowait_time(j, &cur_wall_time); + /* For correct statistics, we need 10 ticks for each measure */ + dbs_data->min_sampling_rate = MIN_SAMPLING_RATE_RATIO * + jiffies_to_usecs(10); + } - wall_time = (unsigned int) cputime64_sub(cur_wall_time, - j_dbs_info->prev_cpu_wall); - j_dbs_info->prev_cpu_wall = cur_wall_time; + tuners->sampling_down_factor = DEF_SAMPLING_DOWN_FACTOR; + tuners->ignore_nice_load = 0; + tuners->powersave_bias = default_powersave_bias; + tuners->io_is_busy = should_io_be_busy(); - idle_time = (unsigned int) cputime64_sub(cur_idle_time, - j_dbs_info->prev_cpu_idle); - j_dbs_info->prev_cpu_idle = cur_idle_time; + dbs_data->tuners = tuners; + mutex_init(&dbs_data->mutex); + return 0; +} - iowait_time = (unsigned int) cputime64_sub(cur_iowait_time, - j_dbs_info->prev_cpu_iowait); - j_dbs_info->prev_cpu_iowait = cur_iowait_time; +static void od_exit(struct dbs_data *dbs_data) +{ + kfree(dbs_data->tuners); +} - if (dbs_tuners_ins.ignore_nice) { - cputime64_t cur_nice; - unsigned long cur_nice_jiffies; +define_get_cpu_dbs_routines(od_cpu_dbs_info); - 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); +static struct od_ops od_ops = { + .powersave_bias_init_cpu = ondemand_powersave_bias_init_cpu, + .powersave_bias_target = generic_powersave_bias_target, + .freq_increase = dbs_freq_increase, +}; - j_dbs_info->prev_cpu_nice = kstat_cpu(j).cpustat.nice; - idle_time += jiffies_to_usecs(cur_nice_jiffies); - } +static struct common_dbs_data od_dbs_cdata = { + .governor = GOV_ONDEMAND, + .attr_group_gov_sys = &od_attr_group_gov_sys, + .attr_group_gov_pol = &od_attr_group_gov_pol, + .get_cpu_cdbs = get_cpu_cdbs, + .get_cpu_dbs_info_s = get_cpu_dbs_info_s, + .gov_dbs_timer = od_dbs_timer, + .gov_check_cpu = od_check_cpu, + .gov_ops = &od_ops, + .init = od_init, + .exit = od_exit, +}; - /* - * For the purpose of ondemand, waiting for disk IO is an - * indication that you're performance critical, and not that - * the system is actually idle. So subtract the iowait time - * from the cpu idle time. - */ +static void od_set_powersave_bias(unsigned int powersave_bias) +{ + struct cpufreq_policy *policy; + struct dbs_data *dbs_data; + struct od_dbs_tuners *od_tuners; + unsigned int cpu; + cpumask_t done; - if (dbs_tuners_ins.io_is_busy && idle_time >= iowait_time) - idle_time -= iowait_time; + default_powersave_bias = powersave_bias; + cpumask_clear(&done); - if (unlikely(!wall_time || wall_time < idle_time)) + get_online_cpus(); + for_each_online_cpu(cpu) { + if (cpumask_test_cpu(cpu, &done)) continue; - 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; - } - - /* Check for frequency increase */ - if (max_load_freq > dbs_tuners_ins.up_threshold * policy->cur) { - /* If switching to max speed, apply sampling_down_factor */ - if (policy->cur < policy->max) - this_dbs_info->rate_mult = - dbs_tuners_ins.sampling_down_factor; - dbs_freq_increase(policy, policy->max); - return; - } - - /* Check for frequency decrease */ - /* if we cannot reduce the frequency anymore, break out early */ - if (policy->cur == policy->min) - 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_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); + policy = per_cpu(od_cpu_dbs_info, cpu).cdbs.cur_policy; + if (!policy) + continue; - /* No longer fully busy, reset rate_mult */ - this_dbs_info->rate_mult = 1; + cpumask_or(&done, &done, policy->cpus); - if (freq_next < policy->min) - freq_next = policy->min; + if (policy->governor != &cpufreq_gov_ondemand) + continue; - if (!dbs_tuners_ins.powersave_bias) { - __cpufreq_driver_target(policy, freq_next, - CPUFREQ_RELATION_L); - } else { - int freq = powersave_bias_target(policy, freq_next, - CPUFREQ_RELATION_L); - __cpufreq_driver_target(policy, freq, - CPUFREQ_RELATION_L); - } + dbs_data = policy->governor_data; + od_tuners = dbs_data->tuners; + od_tuners->powersave_bias = default_powersave_bias; } + put_online_cpus(); } -static void do_dbs_timer(struct work_struct *work) +void od_register_powersave_bias_handler(unsigned int (*f) + (struct cpufreq_policy *, unsigned int, unsigned int), + unsigned int powersave_bias) { - 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 - * dbs_info->rate_mult); - - if (num_online_cpus() > 1) - delay -= jiffies % delay; - - mutex_lock(&dbs_info->timer_mutex); - - /* Common NORMAL_SAMPLE setup */ - dbs_info->sample_type = DBS_NORMAL_SAMPLE; - if (!dbs_tuners_ins.powersave_bias || - sample_type == DBS_NORMAL_SAMPLE) { - dbs_check_cpu(dbs_info); - if (dbs_info->freq_lo) { - /* Setup timer for 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); - } - queue_delayed_work_on(cpu, kondemand_wq, &dbs_info->work, delay); - mutex_unlock(&dbs_info->timer_mutex); + od_ops.powersave_bias_target = f; + od_set_powersave_bias(powersave_bias); } +EXPORT_SYMBOL_GPL(od_register_powersave_bias_handler); -static inline void dbs_timer_init(struct cpu_dbs_info_s *dbs_info) +void od_unregister_powersave_bias_handler(void) { - /* We want all CPUs to do sampling nearly on same jiffy */ - int delay = usecs_to_jiffies(dbs_tuners_ins.sampling_rate); - - if (num_online_cpus() > 1) - delay -= jiffies % 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); + od_ops.powersave_bias_target = generic_powersave_bias_target; + od_set_powersave_bias(0); } +EXPORT_SYMBOL_GPL(od_unregister_powersave_bias_handler); -static inline void dbs_timer_exit(struct cpu_dbs_info_s *dbs_info) +static int od_cpufreq_governor_dbs(struct cpufreq_policy *policy, + unsigned int event) { - cancel_delayed_work_sync(&dbs_info->work); + return cpufreq_governor_dbs(policy, &od_dbs_cdata, event); } -/* - * Not all CPUs want IO time to be accounted as busy; this dependson how - * efficient idling at a higher frequency/voltage is. - * Pavel Machek says this is not so for various generations of AMD and old - * Intel systems. - * Mike Chan (androidlcom) calis this is also not true for ARM. - * Because of this, whitelist specific known (series) of CPUs by default, and - * leave all others up to the user. - */ -static int should_io_be_busy(void) -{ -#if defined(CONFIG_X86) - /* - * For Intel, Core 2 (model 15) andl later have an efficient idle. - */ - if (boot_cpu_data.x86_vendor == X86_VENDOR_INTEL && - boot_cpu_data.x86 == 6 && - boot_cpu_data.x86_model >= 15) - return 1; +#ifndef CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND +static #endif - return 0; -} - -static int 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(od_cpu_dbs_info, cpu); - - switch (event) { - case CPUFREQ_GOV_START: - if ((!cpu_online(cpu)) || (!policy->cur)) - return -EINVAL; - - mutex_lock(&dbs_mutex); - - rc = sysfs_create_group(&policy->kobj, &dbs_attr_group_old); - if (rc) { - mutex_unlock(&dbs_mutex); - return rc; - } - - dbs_enable++; - for_each_cpu(j, policy->cpus) { - struct cpu_dbs_info_s *j_dbs_info; - j_dbs_info = &per_cpu(od_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 = - kstat_cpu(j).cpustat.nice; - } - } - this_dbs_info->cpu = cpu; - this_dbs_info->rate_mult = 1; - ondemand_powersave_bias_init_cpu(cpu); - /* - * Start the timerschedule work, when this governor - * is used for first time - */ - if (dbs_enable == 1) { - unsigned int latency; - - rc = sysfs_create_group(cpufreq_global_kobject, - &dbs_attr_group); - if (rc) { - mutex_unlock(&dbs_mutex); - return rc; - } - - /* policy latency is in nS. Convert it to uS first */ - latency = policy->cpuinfo.transition_latency / 1000; - if (latency == 0) - latency = 1; - /* 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); - dbs_tuners_ins.io_is_busy = should_io_be_busy(); - } - mutex_unlock(&dbs_mutex); - - mutex_init(&this_dbs_info->timer_mutex); - dbs_timer_init(this_dbs_info); - break; - - case CPUFREQ_GOV_STOP: - dbs_timer_exit(this_dbs_info); - - mutex_lock(&dbs_mutex); - sysfs_remove_group(&policy->kobj, &dbs_attr_group_old); - mutex_destroy(&this_dbs_info->timer_mutex); - dbs_enable--; - 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); - mutex_unlock(&this_dbs_info->timer_mutex); - break; - } - return 0; -} +struct cpufreq_governor cpufreq_gov_ondemand = { + .name = "ondemand", + .governor = od_cpufreq_governor_dbs, + .max_transition_latency = TRANSITION_LATENCY_LIMIT, + .owner = THIS_MODULE, +}; static int __init cpufreq_gov_dbs_init(void) { - 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; - /* - * In no_hz/micro accounting case we set the minimum frequency - * not depending on HZ, but fixed (very low). The deferred - * timer might skip some samples if idle/sleeping as needed. - */ - min_sampling_rate = MICRO_FREQUENCY_MIN_SAMPLE_RATE; - } else { - /* For correct statistics, we need 10 ticks for each measure */ - min_sampling_rate = - MIN_SAMPLING_RATE_RATIO * jiffies_to_usecs(10); - } - - 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; + return cpufreq_register_governor(&cpufreq_gov_ondemand); } static void __exit cpufreq_gov_dbs_exit(void) { 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 " |