linux 2.6.16.38 w/ vs2.0.3-rc1
[linux-2.6.git] / drivers / cpufreq / cpufreq_conservative.c
index c4c578d..ac38766 100644 (file)
@@ -22,7 +22,6 @@
 #include <linux/types.h>
 #include <linux/fs.h>
 #include <linux/sysfs.h>
-#include <linux/cpu.h>
 #include <linux/sched.h>
 #include <linux/kmod.h>
 #include <linux/workqueue.h>
  */
 
 #define DEF_FREQUENCY_UP_THRESHOLD             (80)
+#define MIN_FREQUENCY_UP_THRESHOLD             (0)
+#define MAX_FREQUENCY_UP_THRESHOLD             (100)
+
 #define DEF_FREQUENCY_DOWN_THRESHOLD           (20)
+#define MIN_FREQUENCY_DOWN_THRESHOLD           (0)
+#define MAX_FREQUENCY_DOWN_THRESHOLD           (100)
 
 /* 
  * The polling frequency of this governor depends on the capability of 
  * All times here are in uS.
  */
 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_SAMPLING_RATE                      (def_sampling_rate / 2)
 #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 DEF_SAMPLING_RATE_LATENCY_MULTIPLIER   (100000)
+#define DEF_SAMPLING_DOWN_FACTOR               (5)
 #define TRANSITION_LATENCY_LIMIT               (10 * 1000)
 
 static void do_dbs_timer(void *data);
@@ -66,21 +66,11 @@ struct cpu_dbs_info_s {
        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 */
 
-/*
- * 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_WORK    (dbs_work, do_dbs_timer, NULL);
 
@@ -97,8 +87,6 @@ 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,
-       .ignore_nice            = 0,
-       .freq_step              = 5,
 };
 
 static inline unsigned int get_cpu_idle_time(unsigned int cpu)
@@ -148,7 +136,7 @@ static ssize_t store_sampling_down_factor(struct cpufreq_policy *unused,
        unsigned int input;
        int ret;
        ret = sscanf (buf, "%u", &input);
-       if (ret != 1 || input > MAX_SAMPLING_DOWN_FACTOR || input < 1)
+       if (ret != 1 )
                return -EINVAL;
 
        mutex_lock(&dbs_mutex);
@@ -185,7 +173,9 @@ 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 <= dbs_tuners_ins.down_threshold) {
+       if (ret != 1 || input > MAX_FREQUENCY_UP_THRESHOLD || 
+                       input < MIN_FREQUENCY_UP_THRESHOLD ||
+                       input <= dbs_tuners_ins.down_threshold) {
                mutex_unlock(&dbs_mutex);
                return -EINVAL;
        }
@@ -204,7 +194,9 @@ 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 >= dbs_tuners_ins.up_threshold) {
+       if (ret != 1 || input > MAX_FREQUENCY_DOWN_THRESHOLD || 
+                       input < MIN_FREQUENCY_DOWN_THRESHOLD ||
+                       input >= dbs_tuners_ins.up_threshold) {
                mutex_unlock(&dbs_mutex);
                return -EINVAL;
        }
@@ -305,17 +297,31 @@ static struct attribute_group dbs_attr_group = {
 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_down_sampling_rate;
-       struct cpu_dbs_info_s *this_dbs_info = &per_cpu(cpu_dbs_info, cpu);
+       static int down_skip[NR_CPUS];
+       static int requested_freq[NR_CPUS];
+       static unsigned short init_flag = 0;
+       struct cpu_dbs_info_s *this_dbs_info;
+       struct cpu_dbs_info_s *dbs_info;
+
        struct cpufreq_policy *policy;
+       unsigned int j;
 
+       this_dbs_info = &per_cpu(cpu_dbs_info, cpu);
        if (!this_dbs_info->enable)
                return;
 
        policy = this_dbs_info->cur_policy;
 
+       if ( init_flag == 0 ) {
+               for_each_online_cpu(j) {
+                       dbs_info = &per_cpu(cpu_dbs_info, j);
+                       requested_freq[j] = dbs_info->cur_policy->cur;
+               }
+               init_flag = 1;
+       }
+       
        /* 
         * The default safe range is 20% to 80% 
         * Every sampling_rate, we check
@@ -331,29 +337,39 @@ static void dbs_check_cpu(int cpu)
         */
 
        /* Check for frequency increase */
-       idle_ticks = UINT_MAX;
 
-       /* Check for frequency increase */
-       total_idle_ticks = get_cpu_idle_time(cpu);
-       tmp_idle_ticks = total_idle_ticks -
-               this_dbs_info->prev_cpu_idle_up;
-       this_dbs_info->prev_cpu_idle_up = total_idle_ticks;
+       idle_ticks = UINT_MAX;
+       for_each_cpu_mask(j, policy->cpus) {
+               unsigned int tmp_idle_ticks, total_idle_ticks;
+               struct cpu_dbs_info_s *j_dbs_info;
 
-       if (tmp_idle_ticks < idle_ticks)
-               idle_ticks = tmp_idle_ticks;
+               j_dbs_info = &per_cpu(cpu_dbs_info, j);
+               /* Check for frequency increase */
+               total_idle_ticks = get_cpu_idle_time(j);
+               tmp_idle_ticks = total_idle_ticks -
+                       j_dbs_info->prev_cpu_idle_up;
+               j_dbs_info->prev_cpu_idle_up = total_idle_ticks;
+
+               if (tmp_idle_ticks < idle_ticks)
+                       idle_ticks = tmp_idle_ticks;
+       }
 
        /* Scale idle ticks by 100 and compare with up and down ticks */
        idle_ticks *= 100;
        up_idle_ticks = (100 - dbs_tuners_ins.up_threshold) *
-                       usecs_to_jiffies(dbs_tuners_ins.sampling_rate);
+               usecs_to_jiffies(dbs_tuners_ins.sampling_rate);
 
        if (idle_ticks < up_idle_ticks) {
-               this_dbs_info->down_skip = 0;
-               this_dbs_info->prev_cpu_idle_down =
-                       this_dbs_info->prev_cpu_idle_up;
+               down_skip[cpu] = 0;
+               for_each_cpu_mask(j, policy->cpus) {
+                       struct cpu_dbs_info_s *j_dbs_info;
 
+                       j_dbs_info = &per_cpu(cpu_dbs_info, j);
+                       j_dbs_info->prev_cpu_idle_down = 
+                                       j_dbs_info->prev_cpu_idle_up;
+               }
                /* if we are already at full speed then break out early */
-               if (this_dbs_info->requested_freq == policy->max)
+               if (requested_freq[cpu] == policy->max)
                        return;
                
                freq_step = (dbs_tuners_ins.freq_step * policy->max) / 100;
@@ -362,45 +378,49 @@ static void dbs_check_cpu(int cpu)
                if (unlikely(freq_step == 0))
                        freq_step = 5;
                
-               this_dbs_info->requested_freq += freq_step;
-               if (this_dbs_info->requested_freq > policy->max)
-                       this_dbs_info->requested_freq = policy->max;
+               requested_freq[cpu] += freq_step;
+               if (requested_freq[cpu] > policy->max)
+                       requested_freq[cpu] = policy->max;
 
-               __cpufreq_driver_target(policy, this_dbs_info->requested_freq,
+               __cpufreq_driver_target(policy, requested_freq[cpu], 
                        CPUFREQ_RELATION_H);
                return;
        }
 
        /* Check for frequency decrease */
-       this_dbs_info->down_skip++;
-       if (this_dbs_info->down_skip < dbs_tuners_ins.sampling_down_factor)
+       down_skip[cpu]++;
+       if (down_skip[cpu] < dbs_tuners_ins.sampling_down_factor)
                return;
 
-       /* Check for frequency decrease */
-       total_idle_ticks = this_dbs_info->prev_cpu_idle_up;
-       tmp_idle_ticks = total_idle_ticks -
-               this_dbs_info->prev_cpu_idle_down;
-       this_dbs_info->prev_cpu_idle_down = total_idle_ticks;
+       idle_ticks = UINT_MAX;
+       for_each_cpu_mask(j, policy->cpus) {
+               unsigned int tmp_idle_ticks, total_idle_ticks;
+               struct cpu_dbs_info_s *j_dbs_info;
+
+               j_dbs_info = &per_cpu(cpu_dbs_info, j);
+               total_idle_ticks = j_dbs_info->prev_cpu_idle_up;
+               tmp_idle_ticks = total_idle_ticks -
+                       j_dbs_info->prev_cpu_idle_down;
+               j_dbs_info->prev_cpu_idle_down = total_idle_ticks;
 
-       if (tmp_idle_ticks < idle_ticks)
-               idle_ticks = tmp_idle_ticks;
+               if (tmp_idle_ticks < idle_ticks)
+                       idle_ticks = tmp_idle_ticks;
+       }
 
        /* Scale idle ticks by 100 and compare with up and down ticks */
        idle_ticks *= 100;
-       this_dbs_info->down_skip = 0;
+       down_skip[cpu] = 0;
 
        freq_down_sampling_rate = dbs_tuners_ins.sampling_rate *
                dbs_tuners_ins.sampling_down_factor;
        down_idle_ticks = (100 - dbs_tuners_ins.down_threshold) *
-               usecs_to_jiffies(freq_down_sampling_rate);
+                       usecs_to_jiffies(freq_down_sampling_rate);
 
        if (idle_ticks > down_idle_ticks) {
-               /*
-                * if we are already at the lowest speed then break out early
+               /* 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
-                */
-               if (this_dbs_info->requested_freq == policy->min
+                * freq_step to be zero */
+               if (requested_freq[cpu] == policy->min
                                || dbs_tuners_ins.freq_step == 0)
                        return;
 
@@ -410,12 +430,13 @@ static void dbs_check_cpu(int cpu)
                if (unlikely(freq_step == 0))
                        freq_step = 5;
 
-               this_dbs_info->requested_freq -= freq_step;
-               if (this_dbs_info->requested_freq < policy->min)
-                       this_dbs_info->requested_freq = policy->min;
+               requested_freq[cpu] -= freq_step;
+               if (requested_freq[cpu] < policy->min)
+                       requested_freq[cpu] = policy->min;
 
-               __cpufreq_driver_target(policy, this_dbs_info->requested_freq,
-                               CPUFREQ_RELATION_H);
+               __cpufreq_driver_target(policy,
+                       requested_freq[cpu],
+                       CPUFREQ_RELATION_H);
                return;
        }
 }
@@ -423,14 +444,12 @@ static void dbs_check_cpu(int cpu)
 static void do_dbs_timer(void *data)
 { 
        int i;
-       lock_cpu_hotplug();
        mutex_lock(&dbs_mutex);
        for_each_online_cpu(i)
                dbs_check_cpu(i);
        schedule_delayed_work(&dbs_work, 
                        usecs_to_jiffies(dbs_tuners_ins.sampling_rate));
        mutex_unlock(&dbs_mutex);
-       unlock_cpu_hotplug();
 } 
 
 static inline void dbs_timer_init(void)
@@ -474,13 +493,11 @@ static int cpufreq_governor_dbs(struct cpufreq_policy *policy,
                        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_up = get_cpu_idle_time(j);
                        j_dbs_info->prev_cpu_idle_down
                                = j_dbs_info->prev_cpu_idle_up;
                }
                this_dbs_info->enable = 1;
-               this_dbs_info->down_skip = 0;
-               this_dbs_info->requested_freq = policy->cur;
                sysfs_create_group(&policy->kobj, &dbs_attr_group);
                dbs_enable++;
                /*
@@ -490,17 +507,16 @@ static int cpufreq_governor_dbs(struct cpufreq_policy *policy,
                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;
-
-                       def_sampling_rate = 10 * latency *
-                                       DEF_SAMPLING_RATE_LATENCY_MULTIPLIER;
 
-                       if (def_sampling_rate < MIN_STAT_SAMPLING_RATE)
-                               def_sampling_rate = MIN_STAT_SAMPLING_RATE;
+                       latency = policy->cpuinfo.transition_latency;
+                       if (latency < 1000)
+                               latency = 1000;
 
+                       def_sampling_rate = (latency / 1000) *
+                                       DEF_SAMPLING_RATE_LATENCY_MULTIPLIER;
                        dbs_tuners_ins.sampling_rate = def_sampling_rate;
+                       dbs_tuners_ins.ignore_nice = 0;
+                       dbs_tuners_ins.freq_step = 5;
 
                        dbs_timer_init();
                }