diff options
Diffstat (limited to 'drivers/cpufreq')
| -rw-r--r-- | drivers/cpufreq/Kconfig.arm | 7 | ||||
| -rw-r--r-- | drivers/cpufreq/Makefile | 5 | ||||
| -rw-r--r-- | drivers/cpufreq/cpufreq-cpu0.c | 2 | ||||
| -rw-r--r-- | drivers/cpufreq/cpufreq.c | 37 | ||||
| -rw-r--r-- | drivers/cpufreq/cpufreq_conservative.c | 558 | ||||
| -rw-r--r-- | drivers/cpufreq/cpufreq_governor.c | 318 | ||||
| -rw-r--r-- | drivers/cpufreq/cpufreq_governor.h | 176 | ||||
| -rw-r--r-- | drivers/cpufreq/cpufreq_ondemand.c | 731 | ||||
| -rw-r--r-- | drivers/cpufreq/cpufreq_performance.c | 2 | ||||
| -rw-r--r-- | drivers/cpufreq/cpufreq_powersave.c | 2 | ||||
| -rw-r--r-- | drivers/cpufreq/cpufreq_stats.c | 4 | ||||
| -rw-r--r-- | drivers/cpufreq/cpufreq_userspace.c | 2 | ||||
| -rw-r--r-- | drivers/cpufreq/exynos-cpufreq.c | 11 | ||||
| -rw-r--r-- | drivers/cpufreq/freq_table.c | 2 | ||||
| -rw-r--r-- | drivers/cpufreq/longhaul.c | 4 | ||||
| -rw-r--r-- | drivers/cpufreq/powernow-k8.c | 4 | ||||
| -rw-r--r-- | drivers/cpufreq/spear-cpufreq.c | 291 |
17 files changed, 1219 insertions, 937 deletions
diff --git a/drivers/cpufreq/Kconfig.arm b/drivers/cpufreq/Kconfig.arm index 5961e64..a0b3661 100644 --- a/drivers/cpufreq/Kconfig.arm +++ b/drivers/cpufreq/Kconfig.arm @@ -76,3 +76,10 @@ config ARM_EXYNOS5250_CPUFREQ help This adds the CPUFreq driver for Samsung EXYNOS5250 SoC. + +config ARM_SPEAR_CPUFREQ + bool "SPEAr CPUFreq support" + depends on PLAT_SPEAR + default y + help + This adds the CPUFreq driver support for SPEAr SOCs. diff --git a/drivers/cpufreq/Makefile b/drivers/cpufreq/Makefile index 1bc90e1..1f254ec0 100644 --- a/drivers/cpufreq/Makefile +++ b/drivers/cpufreq/Makefile @@ -7,8 +7,8 @@ obj-$(CONFIG_CPU_FREQ_STAT) += cpufreq_stats.o obj-$(CONFIG_CPU_FREQ_GOV_PERFORMANCE) += cpufreq_performance.o obj-$(CONFIG_CPU_FREQ_GOV_POWERSAVE) += cpufreq_powersave.o obj-$(CONFIG_CPU_FREQ_GOV_USERSPACE) += cpufreq_userspace.o -obj-$(CONFIG_CPU_FREQ_GOV_ONDEMAND) += cpufreq_ondemand.o -obj-$(CONFIG_CPU_FREQ_GOV_CONSERVATIVE) += cpufreq_conservative.o +obj-$(CONFIG_CPU_FREQ_GOV_ONDEMAND) += cpufreq_ondemand.o cpufreq_governor.o +obj-$(CONFIG_CPU_FREQ_GOV_CONSERVATIVE) += cpufreq_conservative.o cpufreq_governor.o # CPUfreq cross-arch helpers obj-$(CONFIG_CPU_FREQ_TABLE) += freq_table.o @@ -50,6 +50,7 @@ obj-$(CONFIG_ARM_EXYNOS4210_CPUFREQ) += exynos4210-cpufreq.o obj-$(CONFIG_ARM_EXYNOS4X12_CPUFREQ) += exynos4x12-cpufreq.o obj-$(CONFIG_ARM_EXYNOS5250_CPUFREQ) += exynos5250-cpufreq.o obj-$(CONFIG_ARM_OMAP2PLUS_CPUFREQ) += omap-cpufreq.o +obj-$(CONFIG_ARM_SPEAR_CPUFREQ) += spear-cpufreq.o ################################################################################## # PowerPC platform drivers diff --git a/drivers/cpufreq/cpufreq-cpu0.c b/drivers/cpufreq/cpufreq-cpu0.c index e915827..52bf36d 100644 --- a/drivers/cpufreq/cpufreq-cpu0.c +++ b/drivers/cpufreq/cpufreq-cpu0.c @@ -174,7 +174,7 @@ static struct cpufreq_driver cpu0_cpufreq_driver = { .attr = cpu0_cpufreq_attr, }; -static int __devinit cpu0_cpufreq_driver_init(void) +static int cpu0_cpufreq_driver_init(void) { struct device_node *np; int ret; diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index fb8a527..1f93dbd 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -15,6 +15,8 @@ * */ +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + #include <linux/kernel.h> #include <linux/module.h> #include <linux/init.h> @@ -127,7 +129,7 @@ static int __init init_cpufreq_transition_notifier_list(void) pure_initcall(init_cpufreq_transition_notifier_list); static int off __read_mostly; -int cpufreq_disabled(void) +static int cpufreq_disabled(void) { return off; } @@ -402,7 +404,7 @@ static int __cpufreq_set_policy(struct cpufreq_policy *data, static ssize_t store_##file_name \ (struct cpufreq_policy *policy, const char *buf, size_t count) \ { \ - unsigned int ret = -EINVAL; \ + unsigned int ret; \ struct cpufreq_policy new_policy; \ \ ret = cpufreq_get_policy(&new_policy, policy->cpu); \ @@ -445,7 +447,7 @@ static ssize_t show_scaling_governor(struct cpufreq_policy *policy, char *buf) else if (policy->policy == CPUFREQ_POLICY_PERFORMANCE) return sprintf(buf, "performance\n"); else if (policy->governor) - return scnprintf(buf, CPUFREQ_NAME_LEN, "%s\n", + return scnprintf(buf, CPUFREQ_NAME_PLEN, "%s\n", policy->governor->name); return -EINVAL; } @@ -457,7 +459,7 @@ static ssize_t show_scaling_governor(struct cpufreq_policy *policy, char *buf) static ssize_t store_scaling_governor(struct cpufreq_policy *policy, const char *buf, size_t count) { - unsigned int ret = -EINVAL; + unsigned int ret; char str_governor[16]; struct cpufreq_policy new_policy; @@ -491,7 +493,7 @@ static ssize_t store_scaling_governor(struct cpufreq_policy *policy, */ static ssize_t show_scaling_driver(struct cpufreq_policy *policy, char *buf) { - return scnprintf(buf, CPUFREQ_NAME_LEN, "%s\n", cpufreq_driver->name); + return scnprintf(buf, CPUFREQ_NAME_PLEN, "%s\n", cpufreq_driver->name); } /** @@ -512,7 +514,7 @@ static ssize_t show_scaling_available_governors(struct cpufreq_policy *policy, if (i >= (ssize_t) ((PAGE_SIZE / sizeof(char)) - (CPUFREQ_NAME_LEN + 2))) goto out; - i += scnprintf(&buf[i], CPUFREQ_NAME_LEN, "%s ", t->name); + i += scnprintf(&buf[i], CPUFREQ_NAME_PLEN, "%s ", t->name); } out: i += sprintf(&buf[i], "\n"); @@ -581,7 +583,7 @@ static ssize_t show_scaling_setspeed(struct cpufreq_policy *policy, char *buf) } /** - * show_scaling_driver - show the current cpufreq HW/BIOS limitation + * show_bios_limit - show the current cpufreq HW/BIOS limitation */ static ssize_t show_bios_limit(struct cpufreq_policy *policy, char *buf) { @@ -1468,12 +1470,23 @@ int __cpufreq_driver_target(struct cpufreq_policy *policy, unsigned int relation) { int retval = -EINVAL; + unsigned int old_target_freq = target_freq; if (cpufreq_disabled()) return -ENODEV; - pr_debug("target for CPU %u: %u kHz, relation %u\n", policy->cpu, - target_freq, relation); + /* Make sure that target_freq is within supported range */ + if (target_freq > policy->max) + target_freq = policy->max; + if (target_freq < policy->min) + target_freq = policy->min; + + pr_debug("target for CPU %u: %u kHz, relation %u, requested %u kHz\n", + policy->cpu, target_freq, relation, old_target_freq); + + if (target_freq == policy->cur) + return 0; + if (cpu_online(policy->cpu) && cpufreq_driver->target) retval = cpufreq_driver->target(policy, target_freq, relation); @@ -1509,12 +1522,14 @@ int __cpufreq_driver_getavg(struct cpufreq_policy *policy, unsigned int cpu) { int ret = 0; + if (!(cpu_online(cpu) && cpufreq_driver->getavg)) + return 0; + policy = cpufreq_cpu_get(policy->cpu); if (!policy) return -EINVAL; - if (cpu_online(cpu) && cpufreq_driver->getavg) - ret = cpufreq_driver->getavg(policy, cpu); + ret = cpufreq_driver->getavg(policy, cpu); cpufreq_cpu_put(policy); return ret; diff --git a/drivers/cpufreq/cpufreq_conservative.c b/drivers/cpufreq/cpufreq_conservative.c index a152af7..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, @@ -95,95 +42,121 @@ static struct dbs_tuners { .freq_step = 5, }; -static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall) +/* + * 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) { - u64 idle_time; - u64 cur_wall_time; - u64 busy_time; + 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; - cur_wall_time = jiffies64_to_cputime64(get_jiffies_64()); + /* Check for frequency increase */ + if (load > cs_tuners.up_threshold) { + dbs_info->down_skip = 0; - busy_time = kcpustat_cpu(cpu).cpustat[CPUTIME_USER]; - busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM]; - busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ]; - busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ]; - busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL]; - busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE]; + /* if we are already at full speed then break out early */ + if (dbs_info->requested_freq == policy->max) + return; - idle_time = cur_wall_time - busy_time; - if (wall) - *wall = jiffies_to_usecs(cur_wall_time); + freq_target = (cs_tuners.freq_step * policy->max) / 100; - return jiffies_to_usecs(idle_time); + /* 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 inline cputime64_t get_cpu_idle_time(unsigned int cpu, cputime64_t *wall) +static void cs_dbs_timer(struct work_struct *work) { - u64 idle_time = get_cpu_idle_time_us(cpu, NULL); + 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); - if (idle_time == -1ULL) - return get_cpu_idle_time_jiffy(cpu, wall); - else - idle_time += get_cpu_iowait_time_us(cpu, wall); + mutex_lock(&dbs_info->cdbs.timer_mutex); - return idle_time; + dbs_check_cpu(&cs_dbs_data, cpu); + + schedule_delayed_work_on(cpu, &dbs_info->cdbs.work, delay); + mutex_unlock(&dbs_info->cdbs.timer_mutex); } -/* keep track of frequency transitions */ -static int -dbs_cpufreq_notifier(struct notifier_block *nb, unsigned long val, - void *data) +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(cs_cpu_dbs_info, - freq->cpu); - + 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) @@ -195,7 +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; } @@ -209,7 +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; } @@ -220,11 +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; } @@ -237,21 +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; @@ -259,19 +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; } @@ -289,18 +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, @@ -313,283 +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); - } +define_get_cpu_dbs_routines(cs_cpu_dbs_info); - if (unlikely(!wall_time || wall_time < idle_time)) - continue; - - load = 100 * (wall_time - idle_time) / wall_time; - - if (load > max_load) - max_load = load; - } - - /* - * 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 @@ -597,13 +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); } @@ -612,7 +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 " diff --git a/drivers/cpufreq/cpufreq_governor.c b/drivers/cpufreq/cpufreq_governor.c new file mode 100644 index 0000000..6c5f1d3 --- /dev/null +++ b/drivers/cpufreq/cpufreq_governor.c @@ -0,0 +1,318 @@ +/* + * drivers/cpufreq/cpufreq_governor.c + * + * CPUFREQ governors common code + * + * Copyright (C) 2001 Russell King + * (C) 2003 Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>. + * (C) 2003 Jun Nakajima <jun.nakajima@intel.com> + * (C) 2009 Alexander Clouter <alex@digriz.org.uk> + * (c) 2012 Viresh Kumar <viresh.kumar@linaro.org> + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ + +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + +#include <asm/cputime.h> +#include <linux/cpufreq.h> +#include <linux/cpumask.h> +#include <linux/export.h> +#include <linux/kernel_stat.h> +#include <linux/mutex.h> +#include <linux/tick.h> +#include <linux/types.h> +#include <linux/workqueue.h> + +#include "cpufreq_governor.h" + +static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall) +{ + u64 idle_time; + u64 cur_wall_time; + u64 busy_time; + + cur_wall_time = jiffies64_to_cputime64(get_jiffies_64()); + + busy_time = kcpustat_cpu(cpu).cpustat[CPUTIME_USER]; + busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM]; + busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ]; + busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ]; + busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL]; + busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE]; + + idle_time = cur_wall_time - busy_time; + if (wall) + *wall = cputime_to_usecs(cur_wall_time); + + return cputime_to_usecs(idle_time); +} + +u64 get_cpu_idle_time(unsigned int cpu, u64 *wall) +{ + u64 idle_time = get_cpu_idle_time_us(cpu, NULL); + + if (idle_time == -1ULL) + return get_cpu_idle_time_jiffy(cpu, wall); + else + idle_time += get_cpu_iowait_time_us(cpu, wall); + + return idle_time; +} +EXPORT_SYMBOL_GPL(get_cpu_idle_time); + +void dbs_check_cpu(struct dbs_data *dbs_data, int cpu) +{ + struct cpu_dbs_common_info *cdbs = dbs_data->get_cpu_cdbs(cpu); + struct od_dbs_tuners *od_tuners = dbs_data->tuners; + struct cs_dbs_tuners *cs_tuners = dbs_data->tuners; + struct cpufreq_policy *policy; + unsigned int max_load = 0; + unsigned int ignore_nice; + unsigned int j; + + if (dbs_data->governor == GOV_ONDEMAND) + ignore_nice = od_tuners->ignore_nice; + else + ignore_nice = cs_tuners->ignore_nice; + + policy = cdbs->cur_policy; + + /* Get Absolute Load (in terms of freq for ondemand gov) */ + for_each_cpu(j, policy->cpus) { + struct cpu_dbs_common_info *j_cdbs; + u64 cur_wall_time, cur_idle_time, cur_iowait_time; + unsigned int idle_time, wall_time, iowait_time; + unsigned int load; + + j_cdbs = dbs_data->get_cpu_cdbs(j); + + cur_idle_time = get_cpu_idle_time(j, &cur_wall_time); + + wall_time = (unsigned int) + (cur_wall_time - j_cdbs->prev_cpu_wall); + j_cdbs->prev_cpu_wall = cur_wall_time; + + idle_time = (unsigned int) + (cur_idle_time - j_cdbs->prev_cpu_idle); + j_cdbs->prev_cpu_idle = cur_idle_time; + + if (ignore_nice) { + u64 cur_nice; + unsigned long cur_nice_jiffies; + + cur_nice = kcpustat_cpu(j).cpustat[CPUTIME_NICE] - + cdbs->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); + + cdbs->prev_cpu_nice = + kcpustat_cpu(j).cpustat[CPUTIME_NICE]; + idle_time += jiffies_to_usecs(cur_nice_jiffies); + } + + if (dbs_data->governor == GOV_ONDEMAND) { + struct od_cpu_dbs_info_s *od_j_dbs_info = + dbs_data->get_cpu_dbs_info_s(cpu); + + cur_iowait_time = get_cpu_iowait_time_us(j, + &cur_wall_time); + if (cur_iowait_time == -1ULL) + cur_iowait_time = 0; + + iowait_time = (unsigned int) (cur_iowait_time - + od_j_dbs_info->prev_cpu_iowait); + od_j_dbs_info->prev_cpu_iowait = cur_iowait_time; + + /* + * 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. + */ + if (od_tuners->io_is_busy && idle_time >= iowait_time) + idle_time -= iowait_time; + } + + if (unlikely(!wall_time || wall_time < idle_time)) + continue; + + load = 100 * (wall_time - idle_time) / wall_time; + + if (dbs_data->governor == GOV_ONDEMAND) { + int freq_avg = __cpufreq_driver_getavg(policy, j); + if (freq_avg <= 0) + freq_avg = policy->cur; + + load *= freq_avg; + } + + if (load > max_load) + max_load = load; + } + + dbs_data->gov_check_cpu(cpu, max_load); +} +EXPORT_SYMBOL_GPL(dbs_check_cpu); + +static inline void dbs_timer_init(struct dbs_data *dbs_data, + struct cpu_dbs_common_info *cdbs, unsigned int sampling_rate) +{ + int delay = delay_for_sampling_rate(sampling_rate); + + INIT_DEFERRABLE_WORK(&cdbs->work, dbs_data->gov_dbs_timer); + schedule_delayed_work_on(cdbs->cpu, &cdbs->work, delay); +} + +static inline void dbs_timer_exit(struct cpu_dbs_common_info *cdbs) +{ + cancel_delayed_work_sync(&cdbs->work); +} + +int cpufreq_governor_dbs(struct dbs_data *dbs_data, + struct cpufreq_policy *policy, unsigned int event) +{ + struct od_cpu_dbs_info_s *od_dbs_info = NULL; + struct cs_cpu_dbs_info_s *cs_dbs_info = NULL; + struct od_dbs_tuners *od_tuners = dbs_data->tuners; + struct cs_dbs_tuners *cs_tuners = dbs_data->tuners; + struct cpu_dbs_common_info *cpu_cdbs; + unsigned int *sampling_rate, latency, ignore_nice, j, cpu = policy->cpu; + int rc; + + cpu_cdbs = dbs_data->get_cpu_cdbs(cpu); + + if (dbs_data->governor == GOV_CONSERVATIVE) { + cs_dbs_info = dbs_data->get_cpu_dbs_info_s(cpu); + sampling_rate = &cs_tuners->sampling_rate; + ignore_nice = cs_tuners->ignore_nice; + } else { + od_dbs_info = dbs_data->get_cpu_dbs_info_s(cpu); + sampling_rate = &od_tuners->sampling_rate; + ignore_nice = od_tuners->ignore_nice; + } + + switch (event) { + case CPUFREQ_GOV_START: + if ((!cpu_online(cpu)) || (!policy->cur)) + return -EINVAL; + + mutex_lock(&dbs_data->mutex); + + dbs_data->enable++; + cpu_cdbs->cpu = cpu; + for_each_cpu(j, policy->cpus) { + struct cpu_dbs_common_info *j_cdbs; + j_cdbs = dbs_data->get_cpu_cdbs(j); + + j_cdbs->cur_policy = policy; + j_cdbs->prev_cpu_idle = get_cpu_idle_time(j, + &j_cdbs->prev_cpu_wall); + if (ignore_nice) + j_cdbs->prev_cpu_nice = + kcpustat_cpu(j).cpustat[CPUTIME_NICE]; + } + + /* + * Start the timerschedule work, when this governor is used for + * first time + */ + if (dbs_data->enable != 1) + goto second_time; + + rc = sysfs_create_group(cpufreq_global_kobject, + dbs_data->attr_group); + if (rc) { + mutex_unlock(&dbs_data->mutex); + return rc; + } + + /* policy latency is in nS. Convert it to uS first */ + latency = policy->cpuinfo.transition_latency / 1000; + if (latency == 0) + latency = 1; + + /* + * conservative does not implement micro like ondemand + * governor, thus we are bound to jiffes/HZ + */ + if (dbs_data->governor == GOV_CONSERVATIVE) { + struct cs_ops *ops = dbs_data->gov_ops; + + cpufreq_register_notifier(ops->notifier_block, + CPUFREQ_TRANSITION_NOTIFIER); + + dbs_data->min_sampling_rate = MIN_SAMPLING_RATE_RATIO * + jiffies_to_usecs(10); + } else { + struct od_ops *ops = dbs_data->gov_ops; + + od_tuners->io_is_busy = ops->io_busy(); + } + + /* Bring kernel and HW constraints together */ + dbs_data->min_sampling_rate = max(dbs_data->min_sampling_rate, + MIN_LATENCY_MULTIPLIER * latency); + *sampling_rate = max(dbs_data->min_sampling_rate, latency * + LATENCY_MULTIPLIER); + +second_time: + if (dbs_data->governor == GOV_CONSERVATIVE) { + cs_dbs_info->down_skip = 0; + cs_dbs_info->enable = 1; + cs_dbs_info->requested_freq = policy->cur; + } else { + struct od_ops *ops = dbs_data->gov_ops; + od_dbs_info->rate_mult = 1; + od_dbs_info->sample_type = OD_NORMAL_SAMPLE; + ops->powersave_bias_init_cpu(cpu); + } + mutex_unlock(&dbs_data->mutex); + + mutex_init(&cpu_cdbs->timer_mutex); + dbs_timer_init(dbs_data, cpu_cdbs, *sampling_rate); + break; + + case CPUFREQ_GOV_STOP: + if (dbs_data->governor == GOV_CONSERVATIVE) + cs_dbs_info->enable = 0; + + dbs_timer_exit(cpu_cdbs); + + mutex_lock(&dbs_data->mutex); + mutex_destroy(&cpu_cdbs->timer_mutex); + dbs_data->enable--; + if (!dbs_data->enable) { + struct cs_ops *ops = dbs_data->gov_ops; + + sysfs_remove_group(cpufreq_global_kobject, + dbs_data->attr_group); + if (dbs_data->governor == GOV_CONSERVATIVE) + cpufreq_unregister_notifier(ops->notifier_block, + CPUFREQ_TRANSITION_NOTIFIER); + } + mutex_unlock(&dbs_data->mutex); + + break; + + case CPUFREQ_GOV_LIMITS: + mutex_lock(&cpu_cdbs->timer_mutex); + if (policy->max < cpu_cdbs->cur_policy->cur) + __cpufreq_driver_target(cpu_cdbs->cur_policy, + policy->max, CPUFREQ_RELATION_H); + else if (policy->min > cpu_cdbs->cur_policy->cur) + __cpufreq_driver_target(cpu_cdbs->cur_policy, + policy->min, CPUFREQ_RELATION_L); + dbs_check_cpu(dbs_data, cpu); + mutex_unlock(&cpu_cdbs->timer_mutex); + break; + } + return 0; +} +EXPORT_SYMBOL_GPL(cpufreq_governor_dbs); diff --git a/drivers/cpufreq/cpufreq_governor.h b/drivers/cpufreq/cpufreq_governor.h new file mode 100644 index 0000000..f661654 --- /dev/null +++ b/drivers/cpufreq/cpufreq_governor.h @@ -0,0 +1,176 @@ +/* + * drivers/cpufreq/cpufreq_governor.h + * + * Header file for CPUFreq governors common code + * + * Copyright (C) 2001 Russell King + * (C) 2003 Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>. + * (C) 2003 Jun Nakajima <jun.nakajima@intel.com> + * (C) 2009 Alexander Clouter <alex@digriz.org.uk> + * (c) 2012 Viresh Kumar <viresh.kumar@linaro.org> + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ + +#ifndef _CPUFREQ_GOVERNER_H +#define _CPUFREQ_GOVERNER_H + +#include <linux/cpufreq.h> +#include <linux/kobject.h> +#include <linux/mutex.h> +#include <linux/workqueue.h> +#include <linux/sysfs.h> + +/* + * The polling frequency 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) +#define LATENCY_MULTIPLIER (1000) +#define MIN_LATENCY_MULTIPLIER (100) +#define TRANSITION_LATENCY_LIMIT (10 * 1000 * 1000) + +/* Ondemand Sampling types */ +enum {OD_NORMAL_SAMPLE, OD_SUB_SAMPLE}; + +/* Macro creating sysfs show routines */ +#define show_one(_gov, file_name, object) \ +static ssize_t show_##file_name \ +(struct kobject *kobj, struct attribute *attr, char *buf) \ +{ \ + return sprintf(buf, "%u\n", _gov##_tuners.object); \ +} + +#define define_get_cpu_dbs_routines(_dbs_info) \ +static struct cpu_dbs_common_info *get_cpu_cdbs(int cpu) \ +{ \ + return &per_cpu(_dbs_info, cpu).cdbs; \ +} \ + \ +static void *get_cpu_dbs_info_s(int cpu) \ +{ \ + return &per_cpu(_dbs_info, cpu); \ +} + +/* + * Abbreviations: + * dbs: used as a shortform for demand based switching It helps to keep variable + * names smaller, simpler + * cdbs: common dbs + * on_*: On-demand governor + * cs_*: Conservative governor + */ + +/* Per cpu structures */ +struct cpu_dbs_common_info { + int cpu; + u64 prev_cpu_idle; + u64 prev_cpu_wall; + u64 prev_cpu_nice; + struct cpufreq_policy *cur_policy; + struct delayed_work work; + /* + * percpu mutex that serializes governor limit change with gov_dbs_timer + * invocation. We do not want gov_dbs_timer to run when user is changing + * the governor or limits. + */ + struct mutex timer_mutex; +}; + +struct od_cpu_dbs_info_s { + struct cpu_dbs_common_info cdbs; + u64 prev_cpu_iowait; + struct cpufreq_frequency_table *freq_table; + unsigned int freq_lo; + unsigned int freq_lo_jiffies; + unsigned int freq_hi_jiffies; + unsigned int rate_mult; + unsigned int sample_type:1; +}; + +struct cs_cpu_dbs_info_s { + struct cpu_dbs_common_info cdbs; + unsigned int down_skip; + unsigned int requested_freq; + unsigned int enable:1; +}; + +/* Governers sysfs tunables */ +struct od_dbs_tuners { + unsigned int ignore_nice; + unsigned int sampling_rate; + unsigned int sampling_down_factor; + unsigned int up_threshold; + unsigned int down_differential; + unsigned int powersave_bias; + unsigned int io_is_busy; +}; + +struct cs_dbs_tuners { + unsigned int ignore_nice; + unsigned int sampling_rate; + unsigned int sampling_down_factor; + unsigned int up_threshold; + unsigned int down_threshold; + unsigned int freq_step; +}; + +/* Per Governer data */ +struct dbs_data { + /* Common across governors */ + #define GOV_ONDEMAND 0 + #define GOV_CONSERVATIVE 1 + int governor; + unsigned int min_sampling_rate; + unsigned int enable; /* number of CPUs using this policy */ + struct attribute_group *attr_group; + void *tuners; + + /* dbs_mutex protects dbs_enable in governor start/stop */ + struct mutex mutex; + + struct cpu_dbs_common_info *(*get_cpu_cdbs)(int cpu); + void *(*get_cpu_dbs_info_s)(int cpu); + void (*gov_dbs_timer)(struct work_struct *work); + void (*gov_check_cpu)(int cpu, unsigned int load); + + /* Governor specific ops, see below */ + void *gov_ops; +}; + +/* Governor specific ops, will be passed to dbs_data->gov_ops */ +struct od_ops { + int (*io_busy)(void); + void (*powersave_bias_init_cpu)(int cpu); + unsigned int (*powersave_bias_target)(struct cpufreq_policy *policy, + unsigned int freq_next, unsigned int relation); + void (*freq_increase)(struct cpufreq_policy *p, unsigned int freq); +}; + +struct cs_ops { + struct notifier_block *notifier_block; +}; + +static inline int delay_for_sampling_rate(unsigned int sampling_rate) +{ + int delay = usecs_to_jiffies(sampling_rate); + + /* We want all CPUs to do sampling nearly on same jiffy */ + if (num_online_cpus() > 1) + delay -= jiffies % delay; + + return delay; +} + +u64 get_cpu_idle_time(unsigned int cpu, u64 *wall); +void dbs_check_cpu(struct dbs_data *dbs_data, int cpu); +int cpufreq_governor_dbs(struct dbs_data *dbs_data, + struct cpufreq_policy *policy, unsigned int event); +#endif /* _CPUFREQ_GOVERNER_H */ diff --git a/drivers/cpufreq/cpufreq_ondemand.c b/drivers/cpufreq/cpufreq_ondemand.c index 396322f..7731f7c 100644 --- a/drivers/cpufreq/cpufreq_ondemand.c +++ b/drivers/cpufreq/cpufreq_ondemand.c @@ -10,24 +10,23 @@ * published by the Free Software Foundation. */ -#include <linux/kernel.h> -#include <linux/module.h> -#include <linux/init.h> +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + #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/percpu-defs.h> +#include <linux/sysfs.h> #include <linux/tick.h> -#include <linux/ktime.h> -#include <linux/sched.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" +/* On-demand governor macors */ #define DEF_FREQUENCY_DOWN_DIFFERENTIAL (10) #define DEF_FREQUENCY_UP_THRESHOLD (80) #define DEF_SAMPLING_DOWN_FACTOR (1) @@ -38,80 +37,14 @@ #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 void do_dbs_timer(struct work_struct *work); -static int cpufreq_governor_dbs(struct cpufreq_policy *policy, - unsigned int event); +static struct dbs_data od_dbs_data; +static DEFINE_PER_CPU(struct od_cpu_dbs_info_s, od_cpu_dbs_info); #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 */ - -/* - * dbs_mutex protects dbs_enable in governor start/stop. - */ -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 sampling_down_factor; - unsigned int powersave_bias; - unsigned int io_is_busy; -} dbs_tuners_ins = { +static struct od_dbs_tuners od_tuners = { .up_threshold = DEF_FREQUENCY_UP_THRESHOLD, .sampling_down_factor = DEF_SAMPLING_DOWN_FACTOR, .down_differential = DEF_FREQUENCY_DOWN_DIFFERENTIAL, @@ -119,48 +52,35 @@ static struct dbs_tuners { .powersave_bias = 0, }; -static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall) -{ - u64 idle_time; - u64 cur_wall_time; - u64 busy_time; - - cur_wall_time = jiffies64_to_cputime64(get_jiffies_64()); - - busy_time = kcpustat_cpu(cpu).cpustat[CPUTIME_USER]; - busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM]; - busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ]; - busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ]; - busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL]; - busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE]; - - idle_time = cur_wall_time - busy_time; - if (wall) - *wall = jiffies_to_usecs(cur_wall_time); - - return 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, NULL); - - if (idle_time == -1ULL) - return get_cpu_idle_time_jiffy(cpu, wall); - else - idle_time += get_cpu_iowait_time_us(cpu, wall); + struct od_cpu_dbs_info_s *dbs_info = &per_cpu(od_cpu_dbs_info, cpu); - 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 (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) { - 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) 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; +#endif + return 0; } /* @@ -169,14 +89,13 @@ static inline cputime64_t get_cpu_iowait_time(unsigned int cpu, cputime64_t *wal * 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) + 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); if (!dbs_info->freq_table) { @@ -188,7 +107,7 @@ 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 */ @@ -207,7 +126,7 @@ 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); @@ -218,13 +137,6 @@ 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; @@ -233,83 +145,173 @@ static void ondemand_powersave_bias_init(void) } } -/************************** sysfs interface ************************/ +static void dbs_freq_increase(struct cpufreq_policy *p, unsigned int freq) +{ + if (od_tuners.powersave_bias) + freq = powersave_bias_target(p, freq, CPUFREQ_RELATION_H); + else if (p->cur == p->max) + return; -static ssize_t show_sampling_rate_min(struct kobject *kobj, - struct attribute *attr, char *buf) + __cpufreq_driver_target(p, freq, od_tuners.powersave_bias ? + CPUFREQ_RELATION_L : CPUFREQ_RELATION_H); +} + +/* + * 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 + */ +static void od_check_cpu(int cpu, unsigned int load_freq) { - 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; + + dbs_info->freq_lo = 0; + + /* Check for frequency increase */ + if (load_freq > od_tuners.up_threshold * policy->cur) { + /* 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); + 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 (load_freq < (od_tuners.up_threshold - od_tuners.down_differential) * + policy->cur) { + unsigned int freq_next; + freq_next = load_freq / (od_tuners.up_threshold - + od_tuners.down_differential); + + /* No longer fully busy, reset rate_mult */ + dbs_info->rate_mult = 1; + + if (freq_next < policy->min) + freq_next = policy->min; + + if (!od_tuners.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); + } + } } -define_one_global_ro(sampling_rate_min); +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.cpu; + int delay, sample_type = dbs_info->sample_type; + + mutex_lock(&dbs_info->cdbs.timer_mutex); + + /* Common NORMAL_SAMPLE setup */ + dbs_info->sample_type = OD_NORMAL_SAMPLE; + if (sample_type == OD_SUB_SAMPLE) { + delay = dbs_info->freq_lo_jiffies; + __cpufreq_driver_target(dbs_info->cdbs.cur_policy, + dbs_info->freq_lo, CPUFREQ_RELATION_H); + } else { + dbs_check_cpu(&od_dbs_data, cpu); + if (dbs_info->freq_lo) { + /* Setup timer for SUB_SAMPLE */ + dbs_info->sample_type = OD_SUB_SAMPLE; + delay = dbs_info->freq_hi_jiffies; + } else { + delay = delay_for_sampling_rate(od_tuners.sampling_rate + * dbs_info->rate_mult); + } + } + + schedule_delayed_work_on(cpu, &dbs_info->cdbs.work, delay); + mutex_unlock(&dbs_info->cdbs.timer_mutex); +} + +/************************** sysfs interface ************************/ -/* 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); \ +static ssize_t show_sampling_rate_min(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + return sprintf(buf, "%u\n", od_dbs_data.min_sampling_rate); } -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); /** * update_sampling_rate - update sampling rate effective immediately if needed. * @new_rate: new sampling rate * * If new rate is smaller than the old, simply updaing - * 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. + * 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(unsigned int new_rate) { int cpu; - dbs_tuners_ins.sampling_rate = new_rate - = max(new_rate, min_sampling_rate); + od_tuners.sampling_rate = new_rate = max(new_rate, + od_dbs_data.min_sampling_rate); for_each_online_cpu(cpu) { struct cpufreq_policy *policy; - struct cpu_dbs_info_s *dbs_info; + 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, policy->cpu); cpufreq_cpu_put(policy); - mutex_lock(&dbs_info->timer_mutex); + mutex_lock(&dbs_info->cdbs.timer_mutex); - if (!delayed_work_pending(&dbs_info->work)) { - mutex_unlock(&dbs_info->timer_mutex); + 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->work.timer.expires; - + 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->timer_mutex); - cancel_delayed_work_sync(&dbs_info->work); - mutex_lock(&dbs_info->timer_mutex); + mutex_unlock(&dbs_info->cdbs.timer_mutex); + cancel_delayed_work_sync(&dbs_info->cdbs.work); + mutex_lock(&dbs_info->cdbs.timer_mutex); - schedule_delayed_work_on(dbs_info->cpu, &dbs_info->work, - usecs_to_jiffies(new_rate)); + schedule_delayed_work_on(dbs_info->cdbs.cpu, + &dbs_info->cdbs.work, + usecs_to_jiffies(new_rate)); } - mutex_unlock(&dbs_info->timer_mutex); + mutex_unlock(&dbs_info->cdbs.timer_mutex); } } @@ -334,7 +336,7 @@ static ssize_t store_io_is_busy(struct kobject *a, struct attribute *b, ret = sscanf(buf, "%u", &input); if (ret != 1) return -EINVAL; - dbs_tuners_ins.io_is_busy = !!input; + od_tuners.io_is_busy = !!input; return count; } @@ -349,7 +351,7 @@ static ssize_t store_up_threshold(struct kobject *a, struct attribute *b, input < MIN_FREQUENCY_UP_THRESHOLD) { return -EINVAL; } - dbs_tuners_ins.up_threshold = input; + od_tuners.up_threshold = input; return count; } @@ -362,12 +364,12 @@ 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; + 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; } return count; @@ -388,19 +390,20 @@ 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 == od_tuners.ignore_nice) { /* nothing to do */ return count; } - dbs_tuners_ins.ignore_nice = input; + od_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 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 = kcpustat_cpu(j).cpustat[CPUTIME_NICE]; + dbs_info->cdbs.prev_cpu_idle = get_cpu_idle_time(j, + &dbs_info->cdbs.prev_cpu_wall); + if (od_tuners.ignore_nice) + dbs_info->cdbs.prev_cpu_nice = + kcpustat_cpu(j).cpustat[CPUTIME_NICE]; } return count; @@ -419,17 +422,25 @@ static ssize_t store_powersave_bias(struct kobject *a, struct attribute *b, if (input > 1000) input = 1000; - dbs_tuners_ins.powersave_bias = input; + od_tuners.powersave_bias = input; ondemand_powersave_bias_init(); return count; } +show_one(od, sampling_rate, sampling_rate); +show_one(od, io_is_busy, io_is_busy); +show_one(od, up_threshold, up_threshold); +show_one(od, sampling_down_factor, sampling_down_factor); +show_one(od, ignore_nice_load, ignore_nice); +show_one(od, powersave_bias, powersave_bias); + 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); +define_one_global_ro(sampling_rate_min); static struct attribute *dbs_attributes[] = { &sampling_rate_min.attr, @@ -442,354 +453,71 @@ static struct attribute *dbs_attributes[] = { NULL }; -static struct attribute_group dbs_attr_group = { +static struct attribute_group od_attr_group = { .attrs = dbs_attributes, .name = "ondemand", }; /************************** 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) -{ - 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; - - 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; - - j_dbs_info = &per_cpu(od_cpu_dbs_info, j); - - cur_idle_time = get_cpu_idle_time(j, &cur_wall_time); - cur_iowait_time = get_cpu_iowait_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; - - iowait_time = (unsigned int) - (cur_iowait_time - j_dbs_info->prev_cpu_iowait); - j_dbs_info->prev_cpu_iowait = cur_iowait_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); - } - - /* - * 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. - */ - - if (dbs_tuners_ins.io_is_busy && idle_time >= iowait_time) - idle_time -= iowait_time; - - if (unlikely(!wall_time || wall_time < idle_time)) - 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; - } +define_get_cpu_dbs_routines(od_cpu_dbs_info); - /* 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); - - /* No longer fully busy, reset rate_mult */ - this_dbs_info->rate_mult = 1; - - if (freq_next < policy->min) - freq_next = policy->min; - - 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); - } - } -} - -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; - int sample_type = dbs_info->sample_type; - - int 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 { - /* We want all CPUs to do sampling nearly on - * same jiffy - */ - delay = usecs_to_jiffies(dbs_tuners_ins.sampling_rate - * dbs_info->rate_mult); - - if (num_online_cpus() > 1) - delay -= jiffies % delay; - } - } else { - __cpufreq_driver_target(dbs_info->cur_policy, - dbs_info->freq_lo, CPUFREQ_RELATION_H); - delay = dbs_info->freq_lo_jiffies; - } - 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); - - if (num_online_cpus() > 1) - delay -= jiffies % delay; +static struct od_ops od_ops = { + .io_busy = should_io_be_busy, + .powersave_bias_init_cpu = ondemand_powersave_bias_init_cpu, + .powersave_bias_target = powersave_bias_target, + .freq_increase = dbs_freq_increase, +}; - dbs_info->sample_type = DBS_NORMAL_SAMPLE; - INIT_DEFERRABLE_WORK(&dbs_info->work, do_dbs_timer); - schedule_delayed_work_on(dbs_info->cpu, &dbs_info->work, delay); -} +static struct dbs_data od_dbs_data = { + .governor = GOV_ONDEMAND, + .attr_group = &od_attr_group, + .tuners = &od_tuners, + .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, +}; -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(&od_dbs_data, policy, 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); - - 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 = - kcpustat_cpu(j).cpustat[CPUTIME_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); - 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); - dbs_check_cpu(this_dbs_info); - 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) { u64 idle_time; int cpu = get_cpu(); + mutex_init(&od_dbs_data.mutex); idle_time = get_cpu_idle_time_us(cpu, NULL); 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; + od_tuners.up_threshold = MICRO_FREQUENCY_UP_THRESHOLD; + od_tuners.down_differential = MICRO_FREQUENCY_DOWN_DIFFERENTIAL; /* * 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. */ - min_sampling_rate = MICRO_FREQUENCY_MIN_SAMPLE_RATE; + od_dbs_data.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); + od_dbs_data.min_sampling_rate = MIN_SAMPLING_RATE_RATIO * + jiffies_to_usecs(10); } return cpufreq_register_governor(&cpufreq_gov_ondemand); @@ -800,7 +528,6 @@ static void __exit cpufreq_gov_dbs_exit(void) cpufreq_unregister_governor(&cpufreq_gov_ondemand); } - 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 " diff --git a/drivers/cpufreq/cpufreq_performance.c b/drivers/cpufreq/cpufreq_performance.c index f13a8a9..ceee068 100644 --- a/drivers/cpufreq/cpufreq_performance.c +++ b/drivers/cpufreq/cpufreq_performance.c @@ -10,6 +10,8 @@ * */ +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + #include <linux/kernel.h> #include <linux/module.h> #include <linux/cpufreq.h> diff --git a/drivers/cpufreq/cpufreq_powersave.c b/drivers/cpufreq/cpufreq_powersave.c index 4c2eb51..2d948a1 100644 --- a/drivers/cpufreq/cpufreq_powersave.c +++ b/drivers/cpufreq/cpufreq_powersave.c @@ -10,6 +10,8 @@ * */ +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + #include <linux/kernel.h> #include <linux/module.h> #include <linux/cpufreq.h> diff --git a/drivers/cpufreq/cpufreq_stats.c b/drivers/cpufreq/cpufreq_stats.c index 3998316..e40e508 100644 --- a/drivers/cpufreq/cpufreq_stats.c +++ b/drivers/cpufreq/cpufreq_stats.c @@ -37,7 +37,7 @@ struct cpufreq_stats { unsigned int max_state; unsigned int state_num; unsigned int last_index; - cputime64_t *time_in_state; + u64 *time_in_state; unsigned int *freq_table; #ifdef CONFIG_CPU_FREQ_STAT_DETAILS unsigned int *trans_table; @@ -223,7 +223,7 @@ static int cpufreq_stats_create_table(struct cpufreq_policy *policy, count++; } - alloc_size = count * sizeof(int) + count * sizeof(cputime64_t); + alloc_size = count * sizeof(int) + count * sizeof(u64); #ifdef CONFIG_CPU_FREQ_STAT_DETAILS alloc_size += count * count * sizeof(int); diff --git a/drivers/cpufreq/cpufreq_userspace.c b/drivers/cpufreq/cpufreq_userspace.c index bedac1a..c8c3d29 100644 --- a/drivers/cpufreq/cpufreq_userspace.c +++ b/drivers/cpufreq/cpufreq_userspace.c @@ -11,6 +11,8 @@ * */ +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + #include <linux/kernel.h> #include <linux/module.h> #include <linux/smp.h> diff --git a/drivers/cpufreq/exynos-cpufreq.c b/drivers/cpufreq/exynos-cpufreq.c index af2d81e..7012ea8 100644 --- a/drivers/cpufreq/exynos-cpufreq.c +++ b/drivers/cpufreq/exynos-cpufreq.c @@ -31,13 +31,13 @@ static unsigned int locking_frequency; static bool frequency_locked; static DEFINE_MUTEX(cpufreq_lock); -int exynos_verify_speed(struct cpufreq_policy *policy) +static int exynos_verify_speed(struct cpufreq_policy *policy) { return cpufreq_frequency_table_verify(policy, exynos_info->freq_table); } -unsigned int exynos_getspeed(unsigned int cpu) +static unsigned int exynos_getspeed(unsigned int cpu) { return clk_get_rate(exynos_info->cpu_clk) / 1000; } @@ -100,7 +100,8 @@ static int exynos_target(struct cpufreq_policy *policy, } arm_volt = volt_table[index]; - cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE); + for_each_cpu(freqs.cpu, policy->cpus) + cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE); /* When the new frequency is higher than current frequency */ if ((freqs.new > freqs.old) && !safe_arm_volt) { @@ -115,7 +116,8 @@ static int exynos_target(struct cpufreq_policy *policy, if (freqs.new != freqs.old) exynos_info->set_freq(old_index, index); - cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE); + for_each_cpu(freqs.cpu, policy->cpus) + cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE); /* When the new frequency is lower than current frequency */ if ((freqs.new < freqs.old) || @@ -235,6 +237,7 @@ static int exynos_cpufreq_cpu_init(struct cpufreq_policy *policy) cpumask_copy(policy->related_cpus, cpu_possible_mask); cpumask_copy(policy->cpus, cpu_online_mask); } else { + policy->shared_type = CPUFREQ_SHARED_TYPE_ANY; cpumask_setall(policy->cpus); } diff --git a/drivers/cpufreq/freq_table.c b/drivers/cpufreq/freq_table.c index 90431cb..49cda25 100644 --- a/drivers/cpufreq/freq_table.c +++ b/drivers/cpufreq/freq_table.c @@ -9,6 +9,8 @@ * */ +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + #include <linux/kernel.h> #include <linux/module.h> #include <linux/init.h> diff --git a/drivers/cpufreq/longhaul.c b/drivers/cpufreq/longhaul.c index 53ddbc7..f1fa500 100644 --- a/drivers/cpufreq/longhaul.c +++ b/drivers/cpufreq/longhaul.c @@ -930,7 +930,7 @@ static int __cpuinit longhaul_cpu_init(struct cpufreq_policy *policy) return 0; } -static int __devexit longhaul_cpu_exit(struct cpufreq_policy *policy) +static int longhaul_cpu_exit(struct cpufreq_policy *policy) { cpufreq_frequency_table_put_attr(policy->cpu); return 0; @@ -946,7 +946,7 @@ static struct cpufreq_driver longhaul_driver = { .target = longhaul_target, .get = longhaul_get, .init = longhaul_cpu_init, - .exit = __devexit_p(longhaul_cpu_exit), + .exit = longhaul_cpu_exit, .name = "longhaul", .owner = THIS_MODULE, .attr = longhaul_attr, diff --git a/drivers/cpufreq/powernow-k8.c b/drivers/cpufreq/powernow-k8.c index e3ebb4f..056faf6 100644 --- a/drivers/cpufreq/powernow-k8.c +++ b/drivers/cpufreq/powernow-k8.c @@ -1186,7 +1186,7 @@ err_out: return -ENODEV; } -static int __devexit powernowk8_cpu_exit(struct cpufreq_policy *pol) +static int powernowk8_cpu_exit(struct cpufreq_policy *pol) { struct powernow_k8_data *data = per_cpu(powernow_data, pol->cpu); @@ -1242,7 +1242,7 @@ static struct cpufreq_driver cpufreq_amd64_driver = { .target = powernowk8_target, .bios_limit = acpi_processor_get_bios_limit, .init = powernowk8_cpu_init, - .exit = __devexit_p(powernowk8_cpu_exit), + .exit = powernowk8_cpu_exit, .get = powernowk8_get, .name = "powernow-k8", .owner = THIS_MODULE, diff --git a/drivers/cpufreq/spear-cpufreq.c b/drivers/cpufreq/spear-cpufreq.c new file mode 100644 index 0000000..4575cfe --- /dev/null +++ b/drivers/cpufreq/spear-cpufreq.c @@ -0,0 +1,291 @@ +/* + * drivers/cpufreq/spear-cpufreq.c + * + * CPU Frequency Scaling for SPEAr platform + * + * Copyright (C) 2012 ST Microelectronics + * Deepak Sikri <deepak.sikri@st.com> + * + * This file is licensed under the terms of the GNU General Public + * License version 2. This program is licensed "as is" without any + * warranty of any kind, whether express or implied. + */ + +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + +#include <linux/clk.h> +#include <linux/cpufreq.h> +#include <linux/err.h> +#include <linux/init.h> +#include <linux/module.h> +#include <linux/of.h> +#include <linux/slab.h> +#include <linux/types.h> + +/* SPEAr CPUFreq driver data structure */ +static struct { + struct clk *clk; + unsigned int transition_latency; + struct cpufreq_frequency_table *freq_tbl; + u32 cnt; +} spear_cpufreq; + +int spear_cpufreq_verify(struct cpufreq_policy *policy) +{ + return cpufreq_frequency_table_verify(policy, spear_cpufreq.freq_tbl); +} + +static unsigned int spear_cpufreq_get(unsigned int cpu) +{ + return clk_get_rate(spear_cpufreq.clk) / 1000; +} + +static struct clk *spear1340_cpu_get_possible_parent(unsigned long newfreq) +{ + struct clk *sys_pclk; + int pclk; + /* + * In SPEAr1340, cpu clk's parent sys clk can take input from + * following sources + */ + const char *sys_clk_src[] = { + "sys_syn_clk", + "pll1_clk", + "pll2_clk", + "pll3_clk", + }; + + /* + * As sys clk can have multiple source with their own range + * limitation so we choose possible sources accordingly + */ + if (newfreq <= 300000000) + pclk = 0; /* src is sys_syn_clk */ + else if (newfreq > 300000000 && newfreq <= 500000000) + pclk = 3; /* src is pll3_clk */ + else if (newfreq == 600000000) + pclk = 1; /* src is pll1_clk */ + else + return ERR_PTR(-EINVAL); + + /* Get parent to sys clock */ + sys_pclk = clk_get(NULL, sys_clk_src[pclk]); + if (IS_ERR(sys_pclk)) + pr_err("Failed to get %s clock\n", sys_clk_src[pclk]); + + return sys_pclk; +} + +/* + * In SPEAr1340, we cannot use newfreq directly because we need to actually + * access a source clock (clk) which might not be ancestor of cpu at present. + * Hence in SPEAr1340 we would operate on source clock directly before switching + * cpu clock to it. + */ +static int spear1340_set_cpu_rate(struct clk *sys_pclk, unsigned long newfreq) +{ + struct clk *sys_clk; + int ret = 0; + + sys_clk = clk_get_parent(spear_cpufreq.clk); + if (IS_ERR(sys_clk)) { + pr_err("failed to get cpu's parent (sys) clock\n"); + return PTR_ERR(sys_clk); + } + + /* Set the rate of the source clock before changing the parent */ + ret = clk_set_rate(sys_pclk, newfreq); + if (ret) { + pr_err("Failed to set sys clk rate to %lu\n", newfreq); + return ret; + } + + ret = clk_set_parent(sys_clk, sys_pclk); + if (ret) { + pr_err("Failed to set sys clk parent\n"); + return ret; + } + + return 0; +} + +static int spear_cpufreq_target(struct cpufreq_policy *policy, + unsigned int target_freq, unsigned int relation) +{ + struct cpufreq_freqs freqs; + unsigned long newfreq; + struct clk *srcclk; + int index, ret, mult = 1; + + if (cpufreq_frequency_table_target(policy, spear_cpufreq.freq_tbl, + target_freq, relation, &index)) + return -EINVAL; + + freqs.cpu = policy->cpu; + freqs.old = spear_cpufreq_get(0); + + newfreq = spear_cpufreq.freq_tbl[index].frequency * 1000; + if (of_machine_is_compatible("st,spear1340")) { + /* + * SPEAr1340 is special in the sense that due to the possibility + * of multiple clock sources for cpu clk's parent we can have + * different clock source for different frequency of cpu clk. + * Hence we need to choose one from amongst these possible clock + * sources. + */ + srcclk = spear1340_cpu_get_possible_parent(newfreq); + if (IS_ERR(srcclk)) { + pr_err("Failed to get src clk\n"); + return PTR_ERR(srcclk); + } + + /* SPEAr1340: src clk is always 2 * intended cpu clk */ + mult = 2; + } else { + /* + * src clock to be altered is ancestor of cpu clock. Hence we + * can directly work on cpu clk + */ + srcclk = spear_cpufreq.clk; + } + + newfreq = clk_round_rate(srcclk, newfreq * mult); + if (newfreq < 0) { + pr_err("clk_round_rate failed for cpu src clock\n"); + return newfreq; + } + + freqs.new = newfreq / 1000; + freqs.new /= mult; + cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE); + + if (mult == 2) + ret = spear1340_set_cpu_rate(srcclk, newfreq); + else + ret = clk_set_rate(spear_cpufreq.clk, newfreq); + + /* Get current rate after clk_set_rate, in case of failure */ + if (ret) { + pr_err("CPU Freq: cpu clk_set_rate failed: %d\n", ret); + freqs.new = clk_get_rate(spear_cpufreq.clk) / 1000; + } + + cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE); + return ret; +} + +static int spear_cpufreq_init(struct cpufreq_policy *policy) +{ + int ret; + + ret = cpufreq_frequency_table_cpuinfo(policy, spear_cpufreq.freq_tbl); + if (ret) { + pr_err("cpufreq_frequency_table_cpuinfo() failed"); + return ret; + } + + cpufreq_frequency_table_get_attr(spear_cpufreq.freq_tbl, policy->cpu); + policy->cpuinfo.transition_latency = spear_cpufreq.transition_latency; + policy->cur = spear_cpufreq_get(0); + + cpumask_copy(policy->cpus, topology_core_cpumask(policy->cpu)); + cpumask_copy(policy->related_cpus, policy->cpus); + + return 0; +} + +static int spear_cpufreq_exit(struct cpufreq_policy *policy) +{ + cpufreq_frequency_table_put_attr(policy->cpu); + return 0; +} + +static struct freq_attr *spear_cpufreq_attr[] = { + &cpufreq_freq_attr_scaling_available_freqs, + NULL, +}; + +static struct cpufreq_driver spear_cpufreq_driver = { + .name = "cpufreq-spear", + .flags = CPUFREQ_STICKY, + .verify = spear_cpufreq_verify, + .target = spear_cpufreq_target, + .get = spear_cpufreq_get, + .init = spear_cpufreq_init, + .exit = spear_cpufreq_exit, + .attr = spear_cpufreq_attr, +}; + +static int spear_cpufreq_driver_init(void) +{ + struct device_node *np; + const struct property *prop; + struct cpufreq_frequency_table *freq_tbl; + const __be32 *val; + int cnt, i, ret; + + np = of_find_node_by_path("/cpus/cpu@0"); + if (!np) { + pr_err("No cpu node found"); + return -ENODEV; + } + + if (of_property_read_u32(np, "clock-latency", + &spear_cpufreq.transition_latency)) + spear_cpufreq.transition_latency = CPUFREQ_ETERNAL; + + prop = of_find_property(np, "cpufreq_tbl", NULL); + if (!prop || !prop->value) { + pr_err("Invalid cpufreq_tbl"); + ret = -ENODEV; + goto out_put_node; + } + + cnt = prop->length / sizeof(u32); + val = prop->value; + + freq_tbl = kmalloc(sizeof(*freq_tbl) * (cnt + 1), GFP_KERNEL); + if (!freq_tbl) { + ret = -ENOMEM; + goto out_put_node; + } + + for (i = 0; i < cnt; i++) { + freq_tbl[i].index = i; + freq_tbl[i].frequency = be32_to_cpup(val++); + } + + freq_tbl[i].index = i; + freq_tbl[i].frequency = CPUFREQ_TABLE_END; + + spear_cpufreq.freq_tbl = freq_tbl; + + of_node_put(np); + + spear_cpufreq.clk = clk_get(NULL, "cpu_clk"); + if (IS_ERR(spear_cpufreq.clk)) { + pr_err("Unable to get CPU clock\n"); + ret = PTR_ERR(spear_cpufreq.clk); + goto out_put_mem; + } + + ret = cpufreq_register_driver(&spear_cpufreq_driver); + if (!ret) + return 0; + + pr_err("failed register driver: %d\n", ret); + clk_put(spear_cpufreq.clk); + +out_put_mem: + kfree(freq_tbl); + return ret; + +out_put_node: + of_node_put(np); + return ret; +} +late_initcall(spear_cpufreq_driver_init); + +MODULE_AUTHOR("Deepak Sikri <deepak.sikri@st.com>"); +MODULE_DESCRIPTION("SPEAr CPUFreq driver"); +MODULE_LICENSE("GPL"); |
