X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=drivers%2Fcpufreq%2Fcpufreq_conservative.c;h=e2657837d954a459a731639811fcf3e4b7d91399;hb=05a9bd46e49a9cbb09a0c61c901642a9911bf56e;hp=7498f2506adeaa2cffe379509786a24fd8049714;hpb=08a28e2e98aa821cf6f15f8a267beb2f33377bb9;p=safe%2Fjmp%2Flinux-2.6 diff --git a/drivers/cpufreq/cpufreq_conservative.c b/drivers/cpufreq/cpufreq_conservative.c index 7498f25..e265783 100644 --- a/drivers/cpufreq/cpufreq_conservative.c +++ b/drivers/cpufreq/cpufreq_conservative.c @@ -22,7 +22,7 @@ #include #include #include -#include +#include #include #include #include @@ -37,66 +37,106 @@ #define DEF_FREQUENCY_UP_THRESHOLD (80) #define DEF_FREQUENCY_DOWN_THRESHOLD (20) -/* - * The polling frequency of this governor depends on the capability of +/* + * 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 + * 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. + * For CPUs with transition latency > 10mS (mostly drivers + * with CPUFREQ_ETERNAL), this governor will not work. * All times here are in uS. */ -static unsigned int def_sampling_rate; +static unsigned int def_sampling_rate; #define MIN_SAMPLING_RATE_RATIO (2) /* for correct statistics, we need at least 10 ticks between each measure */ -#define MIN_STAT_SAMPLING_RATE (MIN_SAMPLING_RATE_RATIO * jiffies_to_usecs(10)) -#define MIN_SAMPLING_RATE (def_sampling_rate / MIN_SAMPLING_RATE_RATIO) +#define MIN_STAT_SAMPLING_RATE \ + (MIN_SAMPLING_RATE_RATIO * jiffies_to_usecs(10)) +#define MIN_SAMPLING_RATE \ + (def_sampling_rate / MIN_SAMPLING_RATE_RATIO) #define MAX_SAMPLING_RATE (500 * def_sampling_rate) #define DEF_SAMPLING_RATE_LATENCY_MULTIPLIER (1000) #define DEF_SAMPLING_DOWN_FACTOR (1) #define MAX_SAMPLING_DOWN_FACTOR (10) -#define TRANSITION_LATENCY_LIMIT (10 * 1000) +#define TRANSITION_LATENCY_LIMIT (10 * 1000 * 1000) -static void do_dbs_timer(void *data); +static void do_dbs_timer(struct work_struct *work); struct cpu_dbs_info_s { - struct cpufreq_policy *cur_policy; - unsigned int prev_cpu_idle_up; - unsigned int prev_cpu_idle_down; - unsigned int enable; + struct cpufreq_policy *cur_policy; + unsigned int prev_cpu_idle_up; + unsigned int prev_cpu_idle_down; + unsigned int enable; + unsigned int down_skip; + unsigned int requested_freq; }; static DEFINE_PER_CPU(struct cpu_dbs_info_s, cpu_dbs_info); static unsigned int dbs_enable; /* number of CPUs using this policy */ -static DEFINE_MUTEX (dbs_mutex); -static DECLARE_WORK (dbs_work, do_dbs_timer, NULL); +/* + * DEADLOCK ALERT! There is a ordering requirement between cpu_hotplug + * lock and dbs_mutex. cpu_hotplug lock should always be held before + * dbs_mutex. If any function that can potentially take cpu_hotplug lock + * (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 + */ +static DEFINE_MUTEX (dbs_mutex); +static DECLARE_DELAYED_WORK(dbs_work, do_dbs_timer); 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; + 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; }; static struct dbs_tuners dbs_tuners_ins = { - .up_threshold = DEF_FREQUENCY_UP_THRESHOLD, - .down_threshold = DEF_FREQUENCY_DOWN_THRESHOLD, - .sampling_down_factor = DEF_SAMPLING_DOWN_FACTOR, + .up_threshold = DEF_FREQUENCY_UP_THRESHOLD, + .down_threshold = DEF_FREQUENCY_DOWN_THRESHOLD, + .sampling_down_factor = DEF_SAMPLING_DOWN_FACTOR, + .ignore_nice = 0, + .freq_step = 5, }; static inline unsigned int get_cpu_idle_time(unsigned int cpu) { - return kstat_cpu(cpu).cpustat.idle + + unsigned int add_nice = 0, ret; + + if (dbs_tuners_ins.ignore_nice) + add_nice = kstat_cpu(cpu).cpustat.nice; + + ret = kstat_cpu(cpu).cpustat.idle + kstat_cpu(cpu).cpustat.iowait + - ( dbs_tuners_ins.ignore_nice ? - kstat_cpu(cpu).cpustat.nice : - 0); + add_nice; + + return ret; } +/* keep track of frequency transitions */ +static int +dbs_cpufreq_notifier(struct notifier_block *nb, unsigned long val, + void *data) +{ + struct cpufreq_freqs *freq = data; + struct cpu_dbs_info_s *this_dbs_info = &per_cpu(cpu_dbs_info, + freq->cpu); + + if (!this_dbs_info->enable) + return 0; + + this_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_max(struct cpufreq_policy *policy, char *buf) { @@ -108,8 +148,8 @@ static ssize_t show_sampling_rate_min(struct cpufreq_policy *policy, char *buf) return sprintf (buf, "%u\n", MIN_SAMPLING_RATE); } -#define define_one_ro(_name) \ -static struct freq_attr _name = \ +#define define_one_ro(_name) \ +static struct freq_attr _name = \ __ATTR(_name, 0444, show_##_name, NULL) define_one_ro(sampling_rate_max); @@ -129,7 +169,7 @@ 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 cpufreq_policy *unused, +static ssize_t store_sampling_down_factor(struct cpufreq_policy *unused, const char *buf, size_t count) { unsigned int input; @@ -145,7 +185,7 @@ static ssize_t store_sampling_down_factor(struct cpufreq_policy *unused, return count; } -static ssize_t store_sampling_rate(struct cpufreq_policy *unused, +static ssize_t store_sampling_rate(struct cpufreq_policy *unused, const char *buf, size_t count) { unsigned int input; @@ -164,7 +204,7 @@ static ssize_t store_sampling_rate(struct cpufreq_policy *unused, return count; } -static ssize_t store_up_threshold(struct cpufreq_policy *unused, +static ssize_t store_up_threshold(struct cpufreq_policy *unused, const char *buf, size_t count) { unsigned int input; @@ -172,8 +212,7 @@ static ssize_t store_up_threshold(struct cpufreq_policy *unused, ret = sscanf (buf, "%u", &input); mutex_lock(&dbs_mutex); - if (ret != 1 || input > 100 || input < 0 || - input <= dbs_tuners_ins.down_threshold) { + if (ret != 1 || input > 100 || input <= dbs_tuners_ins.down_threshold) { mutex_unlock(&dbs_mutex); return -EINVAL; } @@ -184,7 +223,7 @@ static ssize_t store_up_threshold(struct cpufreq_policy *unused, return count; } -static ssize_t store_down_threshold(struct cpufreq_policy *unused, +static ssize_t store_down_threshold(struct cpufreq_policy *unused, const char *buf, size_t count) { unsigned int input; @@ -192,8 +231,7 @@ static ssize_t store_down_threshold(struct cpufreq_policy *unused, ret = sscanf (buf, "%u", &input); mutex_lock(&dbs_mutex); - if (ret != 1 || input > 100 || input < 0 || - input >= dbs_tuners_ins.up_threshold) { + if (ret != 1 || input > 100 || input >= dbs_tuners_ins.up_threshold) { mutex_unlock(&dbs_mutex); return -EINVAL; } @@ -211,16 +249,16 @@ static ssize_t store_ignore_nice_load(struct cpufreq_policy *policy, int ret; unsigned int j; - - ret = sscanf (buf, "%u", &input); - if ( ret != 1 ) + + ret = sscanf(buf, "%u", &input); + 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; } @@ -244,14 +282,14 @@ static ssize_t store_freq_step(struct cpufreq_policy *policy, unsigned int input; int ret; - ret = sscanf (buf, "%u", &input); + ret = sscanf(buf, "%u", &input); - if ( ret != 1 ) + if (ret != 1) return -EINVAL; - if ( input > 100 ) + 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 :) */ mutex_lock(&dbs_mutex); @@ -295,51 +333,28 @@ static void dbs_check_cpu(int cpu) { unsigned int idle_ticks, up_idle_ticks, down_idle_ticks; unsigned int tmp_idle_ticks, total_idle_ticks; - unsigned int freq_step; + unsigned int freq_target; unsigned int freq_down_sampling_rate; - static unsigned short down_skip[NR_CPUS]; - static unsigned int requested_freq[NR_CPUS]; - static unsigned int init_flag = NR_CPUS; struct cpu_dbs_info_s *this_dbs_info = &per_cpu(cpu_dbs_info, cpu); struct cpufreq_policy *policy; if (!this_dbs_info->enable) return; - if ( init_flag != 0 ) { - for_each_cpu(init_flag) { - down_skip[init_flag] = 0; - /* I doubt a CPU exists with a freq of 0hz :) */ - requested_freq[init_flag] = 0; - } - init_flag = 0; - } - - /* - * If its a freshly initialised cpu we setup requested_freq. This - * check could be avoided if we did not care about a first time - * stunted increase in CPU speed when there is a load. I feel we - * should be initialising this to something. The removal of a CPU - * is not a problem, after a short time the CPU should settle down - * to a 'natural' frequency. - */ - if (requested_freq[cpu] == 0) - requested_freq[cpu] = this_dbs_info->cur_policy->cur; - policy = this_dbs_info->cur_policy; - /* - * The default safe range is 20% to 80% + /* + * The default safe range is 20% to 80% * Every sampling_rate, we check - * - If current idle time is less than 20%, then we try to - * increase frequency + * - If current idle time is less than 20%, 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 + * - 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 max_frequency + * Any frequency increase takes it to the maximum frequency. + * Frequency reduction happens at minimum steps of + * 5% (default) of max_frequency */ /* Check for frequency increase */ @@ -360,32 +375,32 @@ static void dbs_check_cpu(int cpu) usecs_to_jiffies(dbs_tuners_ins.sampling_rate); if (idle_ticks < up_idle_ticks) { - down_skip[cpu] = 0; + this_dbs_info->down_skip = 0; this_dbs_info->prev_cpu_idle_down = this_dbs_info->prev_cpu_idle_up; /* if we are already at full speed then break out early */ - if (requested_freq[cpu] == policy->max) + if (this_dbs_info->requested_freq == policy->max) return; - - freq_step = (dbs_tuners_ins.freq_step * policy->max) / 100; + + freq_target = (dbs_tuners_ins.freq_step * policy->max) / 100; /* max freq cannot be less than 100. But who knows.... */ - if (unlikely(freq_step == 0)) - freq_step = 5; - - requested_freq[cpu] += freq_step; - if (requested_freq[cpu] > policy->max) - requested_freq[cpu] = policy->max; - - __cpufreq_driver_target(policy, requested_freq[cpu], + 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; } /* Check for frequency decrease */ - down_skip[cpu]++; - if (down_skip[cpu] < dbs_tuners_ins.sampling_down_factor) + this_dbs_info->down_skip++; + if (this_dbs_info->down_skip < dbs_tuners_ins.sampling_down_factor) return; /* Check for frequency decrease */ @@ -399,7 +414,7 @@ static void dbs_check_cpu(int cpu) /* Scale idle ticks by 100 and compare with up and down ticks */ idle_ticks *= 100; - down_skip[cpu] = 0; + this_dbs_info->down_skip = 0; freq_down_sampling_rate = dbs_tuners_ins.sampling_rate * dbs_tuners_ins.sampling_down_factor; @@ -410,42 +425,42 @@ static void dbs_check_cpu(int cpu) /* * if we are already at the lowest speed then break out early * or if we 'cannot' reduce the speed as the user might want - * freq_step to be zero + * freq_target to be zero */ - if (requested_freq[cpu] == policy->min + if (this_dbs_info->requested_freq == policy->min || dbs_tuners_ins.freq_step == 0) return; - freq_step = (dbs_tuners_ins.freq_step * policy->max) / 100; + freq_target = (dbs_tuners_ins.freq_step * policy->max) / 100; /* max freq cannot be less than 100. But who knows.... */ - if (unlikely(freq_step == 0)) - freq_step = 5; + if (unlikely(freq_target == 0)) + freq_target = 5; - requested_freq[cpu] -= freq_step; - if (requested_freq[cpu] < policy->min) - requested_freq[cpu] = policy->min; + this_dbs_info->requested_freq -= freq_target; + if (this_dbs_info->requested_freq < policy->min) + this_dbs_info->requested_freq = policy->min; - __cpufreq_driver_target(policy, requested_freq[cpu], + __cpufreq_driver_target(policy, this_dbs_info->requested_freq, CPUFREQ_RELATION_H); return; } } -static void do_dbs_timer(void *data) -{ +static void do_dbs_timer(struct work_struct *work) +{ int i; mutex_lock(&dbs_mutex); for_each_online_cpu(i) dbs_check_cpu(i); - schedule_delayed_work(&dbs_work, + schedule_delayed_work(&dbs_work, usecs_to_jiffies(dbs_tuners_ins.sampling_rate)); mutex_unlock(&dbs_mutex); -} +} static inline void dbs_timer_init(void) { - INIT_WORK(&dbs_work, do_dbs_timer, NULL); + init_timer_deferrable(&dbs_work.timer); schedule_delayed_work(&dbs_work, usecs_to_jiffies(dbs_tuners_ins.sampling_rate)); return; @@ -463,33 +478,39 @@ static int cpufreq_governor_dbs(struct cpufreq_policy *policy, unsigned int cpu = policy->cpu; struct cpu_dbs_info_s *this_dbs_info; unsigned int j; + int rc; this_dbs_info = &per_cpu(cpu_dbs_info, cpu); switch (event) { case CPUFREQ_GOV_START: - if ((!cpu_online(cpu)) || - (!policy->cur)) + if ((!cpu_online(cpu)) || (!policy->cur)) return -EINVAL; - if (policy->cpuinfo.transition_latency > - (TRANSITION_LATENCY_LIMIT * 1000)) - return -EINVAL; if (this_dbs_info->enable) /* Already enabled */ break; - + mutex_lock(&dbs_mutex); - for_each_cpu_mask(j, policy->cpus) { + + rc = sysfs_create_group(&policy->kobj, &dbs_attr_group); + if (rc) { + mutex_unlock(&dbs_mutex); + return rc; + } + + for_each_cpu_mask_nr(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_up = get_cpu_idle_time(cpu); j_dbs_info->prev_cpu_idle_down = j_dbs_info->prev_cpu_idle_up; } this_dbs_info->enable = 1; - sysfs_create_group(&policy->kobj, &dbs_attr_group); + this_dbs_info->down_skip = 0; + this_dbs_info->requested_freq = policy->cur; + dbs_enable++; /* * Start the timerschedule work, when this governor @@ -509,12 +530,13 @@ static int cpufreq_governor_dbs(struct cpufreq_policy *policy, def_sampling_rate = MIN_STAT_SAMPLING_RATE; dbs_tuners_ins.sampling_rate = def_sampling_rate; - dbs_tuners_ins.ignore_nice = 0; - dbs_tuners_ins.freq_step = 5; dbs_timer_init(); + cpufreq_register_notifier( + &dbs_cpufreq_notifier_block, + CPUFREQ_TRANSITION_NOTIFIER); } - + mutex_unlock(&dbs_mutex); break; @@ -527,9 +549,13 @@ static int cpufreq_governor_dbs(struct cpufreq_policy *policy, * Stop the timerschedule work, when this governor * is used for first time */ - if (dbs_enable == 0) + if (dbs_enable == 0) { dbs_timer_exit(); - + cpufreq_unregister_notifier( + &dbs_cpufreq_notifier_block, + CPUFREQ_TRANSITION_NOTIFIER); + } + mutex_unlock(&dbs_mutex); break; @@ -539,26 +565,30 @@ static int cpufreq_governor_dbs(struct cpufreq_policy *policy, 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 = "conservative", - .governor = cpufreq_governor_dbs, - .owner = THIS_MODULE, +#ifndef CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE +static +#endif +struct cpufreq_governor cpufreq_gov_conservative = { + .name = "conservative", + .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); + return cpufreq_register_governor(&cpufreq_gov_conservative); } static void __exit cpufreq_gov_dbs_exit(void) @@ -566,7 +596,7 @@ static void __exit cpufreq_gov_dbs_exit(void) /* Make sure that the scheduled work is indeed not running */ flush_scheduled_work(); - cpufreq_unregister_governor(&cpufreq_gov_dbs); + cpufreq_unregister_governor(&cpufreq_gov_conservative); } @@ -576,5 +606,9 @@ MODULE_DESCRIPTION ("'cpufreq_conservative' - A dynamic cpufreq governor for " "optimised for use in a battery environment"); MODULE_LICENSE ("GPL"); +#ifdef CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE +fs_initcall(cpufreq_gov_dbs_init); +#else module_init(cpufreq_gov_dbs_init); +#endif module_exit(cpufreq_gov_dbs_exit);