* Request for DIscussion: Cpufreq logging, and frequency floors
@ 2011-10-21 21:31 Steven Finney (Palm GBU)
2011-10-23 11:48 ` Mark Brown
0 siblings, 1 reply; 4+ messages in thread
From: Steven Finney (Palm GBU) @ 2011-10-21 21:31 UTC (permalink / raw)
To: cpufreq
[-- Attachment #1: Type: text/plain, Size: 1497 bytes --]
In both the Touchpad and earlier products, Palm shipped a modified
version of the ondemand.c cpufreq driver that added two facilities:
1) The ability to set a floor frequency for facilities such as
audio that may need a guarantee to avoid dropouts (this differs from
setting cpu_min_freq because it can handle stacked requests and
also cleans up automatically when a process exits (fd close)).
2) The ability to keep a diagnostic log of all the frequency changes so,
e.g., it's possible to determine if bad behavior (e.g. dropouts) is
correlated with a low frequency.
I think both of these are common needs in the embedded world; I know
of at least one other company that independently implemented both of these
(the product didn't ship, so no GPL violation).
We'd be happy to use a standard facility if there were one, but there
doesn't seem to be. Are either of these of general interest, and
should we try to upstream them? It would need some work, as the two
capabilities are tied together in our current 2.6.35 implementation, but
we'd consider separating them out, updating them to TOT, and submitting
one or both as independent patches if there's interest.
The latest Touchpad kernel source should be at
http://opensource.palm.com/3.0.4/index.html; I also
include the cpufreq_ondemand_tickle.c driver as an attachment.
FWIW, I am not the original author of these mods, but I
have done some bug fixes and enhancements.
Steve Finney, Palm/WebOS/HP
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: cpufreq_ondemand_tickle.c --]
[-- Type: text/x-csrc; name="cpufreq_ondemand_tickle.c", Size: 59925 bytes --]
/*
* drivers/cpufreq/cpufreq_ondemand_tickle.c
*
* A version of cpufreq_ondemand supporing hinting, or tickling, into
* high performance levels based on platform defined events. This governor
* also supports setting a temporary frequency floor for maintaining
* minimum required performance levels while still conserving power, such
* as may be required in codecs, com stacks, etc. Ondemand_tickle should
* behave identically to ondemand when neither a tickel nor a floor is active.
*
* Copyright (C) 2001 Russell King
* (C) 2003 Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>.
* Jun Nakajima <jun.nakajima@intel.com>
* (C) 2009 Palm Inc, Corey Tabaka <corey.tabaka@palm.com>
*
* 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.
*/
/* Palm modifications to the ondemand driver fall into 2 categories:
* 1) Floor and tickle capabilities (raise the frequency for controlled
* periods of time)
* 2) Logging of cpufreq data (controlled by "sampling_enabled"; the name
* "sample" for this is a little unfortunate).
* Log data is retrieved from /proc/ondemandtcl_samples, and stored
* with calls to "record_sample". The intent is to store any events that
* might be interesting for post-analysis. The "load" field is overloaded
* for this. For load-based frequency changes, it contains the calculated
* load (0-100). For frequency changes caused for other reasons (tickles,
* floor, limit changes, input subsystem), it will contain a value between
* -1 and -99. Values that do not record an actual frequency change but
* record a relevant event will have a load value < -100; this includes
* CPU1 on/offline transitions. For further detail, look at the code.
*/
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/cpufreq.h>
#include <linux/cpufreq_tickle.h>
#include <linux/cpu.h>
#include <linux/jiffies.h>
#include <linux/kernel_stat.h>
#include <linux/mutex.h>
#include <linux/hrtimer.h>
#include <linux/tick.h>
#include <linux/ktime.h>
#include <linux/sched.h>
#include <linux/input.h>
#include <linux/workqueue.h>
#include <linux/slab.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include <linux/vmalloc.h>
#include <linux/ioctl.h>
#include <linux/fs.h>
#include <linux/cdev.h>
#include <linux/list.h>
/*
* dbs is used in this file as a shortform for demandbased switching
* It helps to keep variable names smaller, simpler
*/
#define DEF_FREQUENCY_DOWN_DIFFERENTIAL (10)
#define DEF_FREQUENCY_UP_THRESHOLD (80)
#define DEF_SAMPLING_DOWN_FACTOR (1)
#define MAX_SAMPLING_DOWN_FACTOR (100000)
#define MICRO_FREQUENCY_DOWN_DIFFERENTIAL (3)
#define MICRO_FREQUENCY_UP_THRESHOLD (95)
#define MICRO_FREQUENCY_MIN_SAMPLE_RATE (10000)
#define MIN_FREQUENCY_UP_THRESHOLD (11)
#define MAX_FREQUENCY_UP_THRESHOLD (100)
#define MIN_FREQUENCY_DOWN_DIFFERENTIAL (1)
/* empirically determined reasonable sampling value; setting it
* here as a default is intended as a temporary fix. Set it to 0 if you
* want this setting to be ignored.
*/
#define DEF_SAMPLING_RATE (200000)
/*
* The max and default time in mS to keep the processor at max freq from a tickle.
*/
#define MAX_TICKLE_WINDOW (10000)
#define DEF_TICKLE_WINDOW (3000)
#if defined(CONFIG_CPU_FREQ_MIN_TICKS)
#define CPU_FREQ_MIN_TICKS CONFIG_CPU_FREQ_MIN_TICKS
#else
#define CPU_FREQ_MIN_TICKS 10
#endif
#if defined(CPU_FREQ_SAMPLING_LATENCY_MULTIPLIER)
#define CPU_FREQ_SAMPLING_LATENCY_MULTIPLIER CONFIG_CPU_FREQ_SAMPLING_LATENCY_MULTIPLIER
#else
#define CPU_FREQ_SAMPLING_LATENCY_MULTIPLIER 1000
#endif
/*
* 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 void do_tickle_timer(unsigned long arg);
/* Sampling types */
enum {DBS_NORMAL_SAMPLE, DBS_SUB_SAMPLE};
/* stashed values from dbs_check_cpu, in us: for log sampling */
struct raw_load_values {
unsigned int wall;
unsigned int idle;
unsigned int iowait;
};
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 tickle_active;
int floor_active;
unsigned int freq_floor;
unsigned int rel_save;
/* values stashed by dbs_check_cpu and retrieved by adjust_for_load */
int cur_load;
unsigned int max_load_freq;
struct raw_load_values rlv;
int cpu;
unsigned int enable:1,
sample_type:1;
/*
* percpu mutex that serializes governor limit change with
* do_dbs_timer invocation. We do not want do_dbs_timer to run
* when user is changing the governor or limits.
*/
struct mutex timer_mutex;
};
static DEFINE_PER_CPU(struct cpu_dbs_info_s, od_cpu_dbs_info);
static unsigned int dbs_enable; /* number of CPUs using this policy */
static void adjust_for_load(struct cpu_dbs_info_s *this_dbs_info);
/*
* dbs_mutex protects data in dbs_tuners_ins from concurrent changes on
* different CPUs. It protects dbs_enable in governor start/stop.
*/
static DEFINE_MUTEX(dbs_mutex);
static struct workqueue_struct *kondemand_wq;
static struct dbs_tuners {
unsigned int sampling_rate;
unsigned int up_threshold;
unsigned int down_differential;
unsigned int ignore_nice;
unsigned int sampling_down_factor;
unsigned int powersave_bias;
unsigned int io_is_busy;
unsigned int max_tickle_window;
} dbs_tuners_ins = {
.up_threshold = DEF_FREQUENCY_UP_THRESHOLD,
.sampling_down_factor = DEF_SAMPLING_DOWN_FACTOR,
.down_differential = DEF_FREQUENCY_DOWN_DIFFERENTIAL,
.ignore_nice = 0,
.powersave_bias = 0,
.max_tickle_window = DEF_TICKLE_WINDOW,
};
/*
* tickle/floor state (shared across CPUs)
* active:floor_active: is a tickle or floor currently held
* active_tickle: is there an active _timed_ tickle.
* active_count, floor_count: number of current tickle/floor
* "clients" (timer = 1, each hold = 1)
*/
static struct {
spinlock_t lock; /* protects this structure, tickle_file_data
* structures, and tickle_clients list
*/
int active;
int active_tickle;
int active_count;
unsigned long tickle_jiffies;
int floor_active;
int floor_count;
unsigned int cur_freq_floor;
struct work_struct tickle_work;
struct work_struct floor_work;
struct timer_list tickle_timer;
} tickle_state = {
.lock = SPIN_LOCK_UNLOCKED,
.active = 0,
.active_tickle = 0,
.active_count = 0,
.floor_active = 0,
.floor_count = 0,
.cur_freq_floor = 0,
};
/*
* Stats for profiling response characteristics.
*/
#define NUM_SAMPLES 2000
struct sample_data {
unsigned long long timestamp;
cputime64_t user;
cputime64_t system;
unsigned int wall;
unsigned int idle;
unsigned int iowait;
unsigned int cur_freq;
unsigned int target_freq;
int load;
};
struct sample_stats {
struct sample_data *samples;
unsigned int current_sample;
};
static DEFINE_PER_CPU(struct sample_stats, sample_stats);
static int sampling_enabled = 0;
module_param(sampling_enabled, bool, S_IRUGO | S_IWUSR);
static int tickling_enabled = 1;
module_param(tickling_enabled, bool, S_IRUGO | S_IWUSR);
static int clear_samples = 0;
module_param(clear_samples, bool, S_IRUGO | S_IWUSR);
static int print_tickles = 0;
module_param(print_tickles, bool, S_IRUGO | S_IWUSR);
/********************* procfs ********************/
struct stats_state {
int cpu;
int sample_index;
int num_samples;
};
static struct seq_operations stats_op;
static void *stats_start(struct seq_file *, loff_t *);
static void *stats_next(struct seq_file *, void *v, loff_t *);
static void stats_stop(struct seq_file *, void *);
static int stats_show(struct seq_file *, void *);
static int stats_open(struct inode *inode, struct file *file)
{
return seq_open(file, &stats_op);
}
static struct file_operations proc_stats_operations = {
.open = stats_open,
.read = seq_read,
.llseek = seq_lseek,
.release = seq_release,
};
static struct seq_operations stats_op = {
.start = stats_start,
.next = stats_next,
.stop = stats_stop,
.show = stats_show,
};
static int get_num_sample_records(void) {
int i, count = 0;
for (i = 0; i < NR_CPUS; i++) {
struct sample_stats *stats = &per_cpu(sample_stats, i);
if (stats->samples)
count += NUM_SAMPLES;
}
return count;
}
static void *stats_start(struct seq_file *m, loff_t *pos)
{
int i;
struct stats_state *state = kmalloc(sizeof(struct stats_state), GFP_KERNEL);
pr_debug("%s: %u\n", __func__, (unsigned int) *pos);
if (!state)
return ERR_PTR(-ENOMEM);
state->cpu = 0;
state->sample_index = *pos;
state->num_samples = get_num_sample_records();
if (*pos >= state->num_samples)
return NULL;
// Find the first CPU running this governor with sampling enabled
for (i = 0; i < NR_CPUS; i++) {
struct sample_stats *stats = &per_cpu(sample_stats, i);
if (stats->samples) {
state->cpu = i;
break;
}
}
while (state->sample_index >= NUM_SAMPLES) {
struct sample_stats *stats = &per_cpu(sample_stats, state->cpu);
state->cpu++;
if (stats->samples)
state->sample_index -= NUM_SAMPLES;
}
return (void *) state;
}
static void *stats_next(struct seq_file *m, void *v, loff_t *pos)
{
int i;
struct stats_state *state = (struct stats_state *) v;
pr_debug("%s: CPU:%u, Pos:%u\n" , __func__, state->cpu,
(unsigned int) *pos);
++*pos;
if (*pos >= state->num_samples)
return NULL;
state->cpu = 0;
state->sample_index = *pos;
// Find the first CPU running this governor with sampling enabled
for (i = 0; i < NR_CPUS; i++) {
struct sample_stats *stats = &per_cpu(sample_stats, i);
if (stats->samples) {
state->cpu = i;
break;
}
}
while (state->sample_index >= NUM_SAMPLES) {
struct sample_stats *stats = &per_cpu(sample_stats, state->cpu);
state->cpu++;
if (stats->samples)
state->sample_index -= NUM_SAMPLES;
}
return (void *) state;
}
static void stats_stop(struct seq_file *m, void *v)
{
kfree(v);
}
static int stats_show(struct seq_file *m, void *v)
{
struct stats_state *state = v;
struct sample_stats *stats = &per_cpu(sample_stats, state->cpu);
int rc = 0;
unsigned long long t;
unsigned long ms;
if (stats->samples) {
/* conversion borrowed from printk_time */
t = stats->samples[state->sample_index].timestamp;
ms = do_div(t, NSEC_PER_SEC);
ms = ms/NSEC_PER_MSEC;
rc = seq_printf(m, "%u %4u %5lu.%03lu %6llu"
" %6llu %8u %8u %8u %4d %7u %7u\n",
state->cpu,
state->sample_index,
(unsigned long)t,
ms,
stats->samples[state->sample_index].user,
stats->samples[state->sample_index].system,
stats->samples[state->sample_index].wall,
stats->samples[state->sample_index].idle,
stats->samples[state->sample_index].iowait,
stats->samples[state->sample_index].load,
stats->samples[state->sample_index].cur_freq,
stats->samples[state->sample_index].target_freq
);
if (rc)
goto done;
}
done:
return rc;
}
/******************** end procfs ********************/
/********************** ioctl ***********************/
static struct class *tickle_class = NULL;
static struct cdev tickle_cdev;
static dev_t tickle_dev;
static LIST_HEAD(tickle_clients);
struct tickle_file_data {
struct list_head list;
int tickle_hold_flag;
int floor_hold_flag;
unsigned int floor_freq;
};
/* kernel_floor_data can be used safely for a single
* active kernel floor (hold/unhold).
*/
struct tickle_file_data kernel_floor_data;
static int get_max_client_floor(void);
static int tickle_open(struct inode *inode, struct file *filp);
static int tickle_release(struct inode *inode, struct file *filp);
static int tickle_ioctl(struct inode *inode, struct file *filp,
unsigned int cmd, unsigned long arg);
static struct file_operations tickle_fops = {
.owner = THIS_MODULE,
.open = tickle_open,
.release = tickle_release,
.ioctl = tickle_ioctl,
};
static int setup_tickle_device(void)
{
int res = 0;
struct device *dev = NULL;
res = alloc_chrdev_region(&tickle_dev, 0, 1, "ondemandtcl");
if (res < 0) {
printk(KERN_ERR "%s: can't alloc major number (%d)\n", __FILE__, res);
goto error;
}
tickle_class = class_create(THIS_MODULE, "ondemandtcl");
if (IS_ERR(tickle_class)) {
res = PTR_ERR(tickle_class);
printk(KERN_ERR "%s: can't create class (%d)\n", __FILE__, res);
goto error_unregister_region;
}
cdev_init(&tickle_cdev, &tickle_fops);
tickle_cdev.owner = THIS_MODULE;
res = cdev_add(&tickle_cdev, tickle_dev, 1);
if (res < 0) {
printk(KERN_ERR "%s: failed to add device (%d)\n", __FILE__, res);
goto error_remove_class;
}
/* always use ondemandtcl0, since userspace is already depending on that name */
dev = device_create(tickle_class, NULL, tickle_dev, NULL, "ondemandtcl%d", 0);
if (IS_ERR(dev)) {
res = PTR_ERR(dev);
printk(KERN_ERR "%s: failed to create device (%d)\n", __FILE__, res);
goto error_remove_cdev;
}
/* add the default data for a single kernel floor client */
kernel_floor_data.floor_freq = 0;
list_add(&kernel_floor_data.list, &tickle_clients);
return res;
error_remove_cdev:
cdev_del(&tickle_cdev);
error_remove_class:
class_destroy(tickle_class);
tickle_class = NULL;
error_unregister_region:
unregister_chrdev_region(tickle_dev, 1);
error:
return res;
}
static void remove_tickle_device(void)
{
if (tickle_class) {
device_destroy(tickle_class, tickle_dev);
class_destroy(tickle_class);
cdev_del(&tickle_cdev);
unregister_chrdev_region(tickle_dev, 1);
/* for posterity */
list_del(&kernel_floor_data.list);
}
}
static int tickle_open(struct inode *inode, struct file *filp)
{
unsigned long flags;
struct tickle_file_data *data = kzalloc(sizeof(struct tickle_file_data), GFP_KERNEL);
if (!data)
return -ENOMEM;
spin_lock_irqsave(&tickle_state.lock, flags);
list_add(&data->list, &tickle_clients);
spin_unlock_irqrestore(&tickle_state.lock, flags);
filp->private_data = data;
return 0;
}
static int tickle_release(struct inode *inode, struct file *filp)
{
unsigned long flags;
struct tickle_file_data *data = filp->private_data;
if (data) {
spin_lock_irqsave(&tickle_state.lock, flags);
list_del(&data->list);
spin_unlock_irqrestore(&tickle_state.lock, flags);
CPUFREQ_UNHOLD_CHECK(&data->tickle_hold_flag);
cpufreq_ondemand_floor_unhold_check(data,
&data->floor_hold_flag);
kfree(data);
}
return 0;
}
/* maximum of one floor hold and one tickle hold per fd */
static int tickle_ioctl(struct inode *inode, struct file *filp, unsigned int cmd, unsigned long arg)
{
unsigned long flags;
struct tickle_file_data *data = filp->private_data;
if (_IOC_TYPE(cmd) != TICKLE_IOC_MAGIC)
return -ENOTTY;
if (_IOC_NR(cmd) > TICKLE_IOC_MAXNR)
return -ENOTTY;
switch (cmd) {
case TICKLE_IOCT_TICKLE:
CPUFREQ_TICKLE_MILLIS((unsigned int) arg);
break;
case TICKLE_IOC_TICKLE_HOLD:
if (data->tickle_hold_flag)
return -EBUSY;
CPUFREQ_HOLD_CHECK(&data->tickle_hold_flag);
break;
case TICKLE_IOC_TICKLE_UNHOLD:
if (data->tickle_hold_flag == 0)
return -EINVAL;
CPUFREQ_UNHOLD_CHECK(&data->tickle_hold_flag);
break;
case TICKLE_IOCT_FLOOR_HOLD:
if (data->floor_hold_flag)
return -EBUSY;
spin_lock_irqsave(&tickle_state.lock, flags);
data->floor_freq = (unsigned int) arg;
spin_unlock_irqrestore(&tickle_state.lock, flags);
cpufreq_ondemand_floor_hold_check(data,
&data->floor_hold_flag);
break;
case TICKLE_IOC_FLOOR_UNHOLD:
if (data->floor_hold_flag == 0)
return -EINVAL;
cpufreq_ondemand_floor_unhold_check(data,
&data->floor_hold_flag);
break;
case TICKLE_IOC_TICKLE_HOLD_SYNC:
CPUFREQ_HOLD_SYNC();
break;
default:
return -ENOTTY;
}
return 0;
}
/* This needs to be called with spinlock tickle_state.lock held */
static int get_max_client_floor(void)
{
int floor_freq = 0;
struct tickle_file_data *file_data;
list_for_each_entry(file_data, &tickle_clients, list) {
if (file_data->floor_freq > floor_freq)
floor_freq = file_data->floor_freq;
}
return floor_freq;
}
/******************** end ioctl *********************/
static inline cputime64_t get_cpu_idle_time_jiffy(unsigned int cpu,
cputime64_t *wall)
{
cputime64_t idle_time;
cputime64_t cur_wall_time;
cputime64_t busy_time;
cur_wall_time = jiffies64_to_cputime64(get_jiffies_64());
busy_time = cputime64_add(kstat_cpu(cpu).cpustat.user,
kstat_cpu(cpu).cpustat.system);
busy_time = cputime64_add(busy_time, kstat_cpu(cpu).cpustat.irq);
busy_time = cputime64_add(busy_time, kstat_cpu(cpu).cpustat.softirq);
busy_time = cputime64_add(busy_time, kstat_cpu(cpu).cpustat.steal);
busy_time = cputime64_add(busy_time, kstat_cpu(cpu).cpustat.nice);
idle_time = cputime64_sub(cur_wall_time, busy_time);
if (wall)
*wall = (cputime64_t)jiffies_to_usecs(cur_wall_time);
return (cputime64_t)jiffies_to_usecs(idle_time);
}
static inline cputime64_t get_cpu_idle_time(unsigned int cpu, cputime64_t *wall)
{
u64 idle_time = get_cpu_idle_time_us(cpu, wall);
if (idle_time == -1ULL)
return get_cpu_idle_time_jiffy(cpu, wall);
return idle_time;
}
static inline cputime64_t get_cpu_iowait_time(unsigned int cpu, cputime64_t *wall)
{
u64 iowait_time = get_cpu_iowait_time_us(cpu, wall);
if (iowait_time == -1ULL)
return 0;
return iowait_time;
}
/*
* Find right freq to be set now with powersave_bias on.
* Returns the freq_hi to be used right now and will set freq_hi_jiffies,
* freq_lo, and freq_lo_jiffies in percpu area for averaging freqs.
*/
static unsigned int powersave_bias_target(struct cpufreq_policy *policy,
unsigned int freq_next,
unsigned int relation)
{
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,
policy->cpu);
if (!dbs_info->freq_table) {
dbs_info->freq_lo = 0;
dbs_info->freq_lo_jiffies = 0;
return freq_next;
}
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_avg = freq_req - freq_reduc;
/* Find freq bounds for freq_avg in freq_table */
index = 0;
cpufreq_frequency_table_target(policy, dbs_info->freq_table, freq_avg,
CPUFREQ_RELATION_H, &index);
freq_lo = dbs_info->freq_table[index].frequency;
index = 0;
cpufreq_frequency_table_target(policy, dbs_info->freq_table, freq_avg,
CPUFREQ_RELATION_L, &index);
freq_hi = dbs_info->freq_table[index].frequency;
/* Find out how long we have to be in hi and lo freqs */
if (freq_hi == freq_lo) {
dbs_info->freq_lo = 0;
dbs_info->freq_lo_jiffies = 0;
return freq_lo;
}
jiffies_total = usecs_to_jiffies(dbs_tuners_ins.sampling_rate);
jiffies_hi = (freq_avg - freq_lo) * jiffies_total;
jiffies_hi += ((freq_hi - freq_lo) / 2);
jiffies_hi /= (freq_hi - freq_lo);
jiffies_lo = jiffies_total - jiffies_hi;
dbs_info->freq_lo = freq_lo;
dbs_info->freq_lo_jiffies = jiffies_lo;
dbs_info->freq_hi_jiffies = jiffies_hi;
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;
for_each_online_cpu(i) {
ondemand_powersave_bias_init_cpu(i);
}
}
/************************** sysfs interface ************************/
static ssize_t show_sampling_rate_max(struct kobject *kobj,
struct attribute *attr, char *buf)
{
printk_once(KERN_INFO "CPUFREQ: ondemand sampling_rate_max "
"sysfs file is deprecated - used by: %s\n", current->comm);
return sprintf(buf, "%u\n", -1U);
}
static ssize_t show_sampling_rate_min(struct kobject *kobj,
struct attribute *attr, char *buf)
{
return sprintf(buf, "%u\n", min_sampling_rate);
}
define_one_global_ro(sampling_rate_max);
define_one_global_ro(sampling_rate_min);
/* 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); \
}
show_one(sampling_rate, sampling_rate);
show_one(io_is_busy, io_is_busy);
show_one(up_threshold, up_threshold);
show_one(down_differential, down_differential);
show_one(sampling_down_factor, sampling_down_factor);
show_one(ignore_nice_load, ignore_nice);
show_one(powersave_bias, powersave_bias);
show_one(max_tickle_window, max_tickle_window);
/*** delete after deprecation time ***/
#define DEPRECATION_MSG(file_name) \
printk_once(KERN_INFO "CPUFREQ: Per core ondemand sysfs " \
"interface is deprecated - " #file_name "\n");
#define show_one_old(file_name) \
static ssize_t show_##file_name##_old \
(struct cpufreq_policy *unused, char *buf) \
{ \
printk_once(KERN_INFO "CPUFREQ: Per core ondemand sysfs " \
"interface is deprecated - " #file_name "\n"); \
return show_##file_name(NULL, NULL, buf); \
}
show_one_old(sampling_rate);
show_one_old(up_threshold);
show_one_old(ignore_nice_load);
show_one_old(powersave_bias);
show_one_old(sampling_rate_min);
show_one_old(sampling_rate_max);
show_one_old(down_differential);
show_one_old(max_tickle_window);
cpufreq_freq_attr_ro_old(sampling_rate_min);
cpufreq_freq_attr_ro_old(sampling_rate_max);
/*** delete after deprecation time ***/
static ssize_t store_sampling_rate(struct kobject *a, struct attribute *b,
const char *buf, size_t count)
{
unsigned int input;
int ret;
ret = sscanf(buf, "%u", &input);
if (ret != 1)
return -EINVAL;
mutex_lock(&dbs_mutex);
dbs_tuners_ins.sampling_rate = max(input, min_sampling_rate);
mutex_unlock(&dbs_mutex);
return count;
}
static ssize_t store_io_is_busy(struct kobject *a, struct attribute *b,
const char *buf, size_t count)
{
unsigned int input;
int ret;
ret = sscanf(buf, "%u", &input);
if (ret != 1)
return -EINVAL;
mutex_lock(&dbs_mutex);
dbs_tuners_ins.io_is_busy = !!input;
mutex_unlock(&dbs_mutex);
return count;
}
static ssize_t store_up_threshold(struct kobject *a, struct attribute *b,
const char *buf, size_t count)
{
unsigned int input;
int ret;
ret = sscanf(buf, "%u", &input);
if (ret != 1 || input > MAX_FREQUENCY_UP_THRESHOLD ||
input < MIN_FREQUENCY_UP_THRESHOLD) {
return -EINVAL;
}
mutex_lock(&dbs_mutex);
dbs_tuners_ins.up_threshold = input;
mutex_unlock(&dbs_mutex);
return count;
}
static ssize_t store_down_differential(struct kobject *a, struct attribute *b,
const char *buf, size_t count)
{
unsigned int input;
int ret;
ret = sscanf(buf, "%u", &input);
mutex_lock(&dbs_mutex);
if (ret != 1 || input >= dbs_tuners_ins.up_threshold ||
input < MIN_FREQUENCY_DOWN_DIFFERENTIAL) {
mutex_unlock(&dbs_mutex);
return -EINVAL;
}
dbs_tuners_ins.down_differential = input;
mutex_unlock(&dbs_mutex);
return count;
}
static ssize_t store_sampling_down_factor(struct kobject *a,
struct attribute *b, const char *buf, size_t count)
{
unsigned int input, j;
int ret;
ret = sscanf(buf, "%u", &input);
if (ret != 1 || input > MAX_SAMPLING_DOWN_FACTOR || input < 1)
return -EINVAL;
mutex_lock(&dbs_mutex);
dbs_tuners_ins.sampling_down_factor = input;
/* 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);
dbs_info->rate_mult = 1;
}
mutex_unlock(&dbs_mutex);
return count;
}
static ssize_t store_ignore_nice_load(struct kobject *a, struct attribute *b,
const char *buf, size_t count)
{
unsigned int input;
int ret;
unsigned int j;
ret = sscanf(buf, "%u", &input);
if (ret != 1)
return -EINVAL;
if (input > 1)
input = 1;
mutex_lock(&dbs_mutex);
if (input == dbs_tuners_ins.ignore_nice) { /* nothing to do */
mutex_unlock(&dbs_mutex);
return count;
}
dbs_tuners_ins.ignore_nice = input;
/* we need to re-evaluate prev_cpu_idle */
for_each_online_cpu(j) {
struct cpu_dbs_info_s *dbs_info;
dbs_info = &per_cpu(od_cpu_dbs_info, j);
dbs_info->prev_cpu_idle = get_cpu_idle_time(j,
&dbs_info->prev_cpu_wall);
if (dbs_tuners_ins.ignore_nice)
dbs_info->prev_cpu_nice = kstat_cpu(j).cpustat.nice;
}
mutex_unlock(&dbs_mutex);
return count;
}
static ssize_t store_powersave_bias(struct kobject *a, struct attribute *b,
const char *buf, size_t count)
{
unsigned int input;
int ret;
ret = sscanf(buf, "%u", &input);
if (ret != 1)
return -EINVAL;
if (input > 1000)
input = 1000;
mutex_lock(&dbs_mutex);
dbs_tuners_ins.powersave_bias = input;
ondemand_powersave_bias_init();
mutex_unlock(&dbs_mutex);
return count;
}
static ssize_t store_max_tickle_window(struct kobject *a, struct attribute *b,
const char *buf, size_t count)
{
unsigned int input;
int ret;
ret = sscanf(buf, "%u", &input);
if (ret != 1)
return -EINVAL;
if (input > MAX_TICKLE_WINDOW)
input = MAX_TICKLE_WINDOW;
mutex_lock(&dbs_mutex);
dbs_tuners_ins.max_tickle_window = input;
mutex_unlock(&dbs_mutex);
return count;
}
define_one_global_rw(sampling_rate);
define_one_global_rw(io_is_busy);
define_one_global_rw(up_threshold);
define_one_global_rw(down_differential);
define_one_global_rw(sampling_down_factor);
define_one_global_rw(ignore_nice_load);
define_one_global_rw(powersave_bias);
define_one_global_rw(max_tickle_window);
static struct attribute *dbs_attributes[] = {
&sampling_rate_max.attr,
&sampling_rate_min.attr,
&sampling_rate.attr,
&up_threshold.attr,
&down_differential.attr,
&sampling_down_factor.attr,
&ignore_nice_load.attr,
&powersave_bias.attr,
&io_is_busy.attr,
&max_tickle_window.attr,
NULL
};
static struct attribute_group dbs_attr_group = {
.attrs = dbs_attributes,
.name = "ondemandtcl",
};
/*** delete after deprecation time ***/
#define write_one_old(file_name) \
static ssize_t store_##file_name##_old \
(struct cpufreq_policy *unused, const char *buf, size_t count) \
{ \
printk_once(KERN_INFO "CPUFREQ: Per core ondemand sysfs " \
"interface is deprecated - " #file_name "\n"); \
return store_##file_name(NULL, NULL, buf, count); \
}
write_one_old(sampling_rate);
write_one_old(up_threshold);
write_one_old(ignore_nice_load);
write_one_old(powersave_bias);
write_one_old(down_differential);
write_one_old(max_tickle_window);
cpufreq_freq_attr_rw_old(sampling_rate);
cpufreq_freq_attr_rw_old(up_threshold);
cpufreq_freq_attr_rw_old(ignore_nice_load);
cpufreq_freq_attr_rw_old(powersave_bias);
cpufreq_freq_attr_rw_old(down_differential);
cpufreq_freq_attr_rw_old(max_tickle_window);
static struct attribute *dbs_attributes_old[] = {
&sampling_rate_max_old.attr,
&sampling_rate_min_old.attr,
&sampling_rate_old.attr,
&up_threshold_old.attr,
&ignore_nice_load_old.attr,
&powersave_bias_old.attr,
&max_tickle_window_old.attr,
&down_differential_old.attr,
NULL
};
static struct attribute_group dbs_attr_group_old = {
.attrs = dbs_attributes_old,
.name = "ondemandtcl",
};
/*** delete after deprecation time ***/
/************************** sysfs end ************************/
static void __attribute__((unused))
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 record_sample(unsigned int cur_freq, unsigned int target_freq,
int load, int cpu, struct raw_load_values *rlvp)
{
struct sample_stats *stats;
int i;
/* abuse a module parameter to trigger clearing the sample buffer. Note
* that after setting the parameter the buffer won't get cleared until
* a sample is made!
*/
if (clear_samples) {
for_each_present_cpu(i) {
stats = &per_cpu(sample_stats, i);
if (stats->samples)
memset(stats->samples, 0, sizeof(struct sample_data) * NUM_SAMPLES);
stats->current_sample = 0;
}
clear_samples = 0;
}
if (!sampling_enabled)
return;
stats = &per_cpu(sample_stats, cpu);
if (!stats->samples)
return;
i = stats->current_sample;
/* for timestamp, use same timer as printk does. CPU0 and
* CPU1 appear to match to fractions of a millisecond
*/
stats->samples[i].timestamp = cpu_clock(0);
stats->samples[i].user = kstat_cpu(cpu).cpustat.user;
stats->samples[i].system = kstat_cpu(cpu).cpustat.system;
stats->samples[i].wall = rlvp ? rlvp->wall : 0;
stats->samples[i].idle = rlvp ? rlvp->idle : 0;
stats->samples[i].iowait = rlvp ? rlvp->iowait : 0;
stats->samples[i].cur_freq = cur_freq;
stats->samples[i].target_freq = target_freq;
stats->samples[i].load = load;
stats->current_sample = (i + 1) % NUM_SAMPLES;
}
static void do_tickle_state_change(struct work_struct *work)
{
int cpu;
unsigned long flags;
int active, active_tickle, active_count;
unsigned long tickle_jiffies;
spin_lock_irqsave(&tickle_state.lock, flags);
active = tickle_state.active;
active_count = tickle_state.active_count;
active_tickle = tickle_state.active_tickle;
tickle_jiffies = tickle_state.tickle_jiffies;
spin_unlock_irqrestore(&tickle_state.lock, flags);
if (print_tickles)
printk("%s: active=%u, active_count=%u, tickle_jiffies=%lu\n",
__func__, active, active_count, tickle_jiffies);
if (active_count) {
if (!active) {
for_each_online_cpu(cpu) {
struct cpu_dbs_info_s *dbs_info;
struct cpufreq_policy *policy;
mutex_lock(&dbs_mutex);
dbs_info = &per_cpu(od_cpu_dbs_info, cpu);
if (!dbs_info->enable) {
mutex_unlock(&dbs_mutex);
continue;
}
mutex_lock(&dbs_info->timer_mutex);
policy = dbs_info->cur_policy;
/* if we don't have a policy then we probably
* got tickled before setup completed
*/
if (!policy) {
mutex_unlock(&dbs_info->timer_mutex);
mutex_unlock(&dbs_mutex);
continue;
}
/* ramp up to the policy max */
if (policy->cur < policy->max) {
record_sample(policy->cur, policy->max,
-21, policy->cpu, NULL);
__cpufreq_driver_target(policy, policy->max, CPUFREQ_RELATION_H);
}
dbs_info->tickle_active = 1;
mutex_unlock(&dbs_info->timer_mutex);
mutex_unlock(&dbs_mutex);
}
}
if (active_tickle)
mod_timer(&tickle_state.tickle_timer, tickle_jiffies);
active = 1;
} else if (active) {
for_each_online_cpu(cpu) {
struct cpu_dbs_info_s *dbs_info;
struct cpufreq_policy *policy;
mutex_lock(&dbs_mutex);
dbs_info = &per_cpu(od_cpu_dbs_info, cpu);
if (!dbs_info->enable) {
mutex_unlock(&dbs_mutex);
continue;
}
mutex_lock(&dbs_info->timer_mutex);
policy = dbs_info->cur_policy;
dbs_info->tickle_active = 0;
adjust_for_load(dbs_info);
mutex_unlock(&dbs_info->timer_mutex);
mutex_unlock(&dbs_mutex);
}
active = 0;
}
/* update tickle_state */
spin_lock_irqsave(&tickle_state.lock, flags);
tickle_state.active = active;
spin_unlock_irqrestore(&tickle_state.lock, flags);
}
void cpufreq_ondemand_tickle_millis(unsigned int millis)
{
int queue = 0;
unsigned long flags, expire;
if (!tickling_enabled)
return;
if (millis > dbs_tuners_ins.max_tickle_window)
millis = dbs_tuners_ins.max_tickle_window;
expire = jiffies + msecs_to_jiffies(millis);
/* record it on CPU 0 */
record_sample(0, millis, -225, 0, NULL);
spin_lock_irqsave(&tickle_state.lock, flags);
if (time_after(expire, tickle_state.tickle_jiffies)) {
tickle_state.tickle_jiffies = expire;
if (!tickle_state.active_tickle)
tickle_state.active_count += 1;
tickle_state.active_tickle = 1;
queue = 1;
}
spin_unlock_irqrestore(&tickle_state.lock, flags);
if (queue)
queue_work(kondemand_wq, &tickle_state.tickle_work);
}
EXPORT_SYMBOL(cpufreq_ondemand_tickle_millis);
void cpufreq_ondemand_tickle(void)
{
cpufreq_ondemand_tickle_millis(dbs_tuners_ins.max_tickle_window);
}
EXPORT_SYMBOL(cpufreq_ondemand_tickle);
void cpufreq_ondemand_hold(void)
{
unsigned long flags;
/* record it on CPU 0 */
record_sample(0, 0, -221, 0, NULL);
spin_lock_irqsave(&tickle_state.lock, flags);
tickle_state.active_count += 1;
spin_unlock_irqrestore(&tickle_state.lock, flags);
queue_work(kondemand_wq, &tickle_state.tickle_work);
}
EXPORT_SYMBOL(cpufreq_ondemand_hold);
void cpufreq_ondemand_unhold(void)
{
int queue = 0;
unsigned long flags;
/* record it on CPU 0 */
record_sample(0, 0, -220, 0, NULL);
spin_lock_irqsave(&tickle_state.lock, flags);
if (tickle_state.active_count) {
tickle_state.active_count -= 1;
queue = 1;
} else {
printk(KERN_WARNING "%s: attempt to decrement when active_count == 0!\n",
__func__);
}
spin_unlock_irqrestore(&tickle_state.lock, flags);
if (queue)
queue_work(kondemand_wq, &tickle_state.tickle_work);
}
EXPORT_SYMBOL(cpufreq_ondemand_unhold);
void cpufreq_ondemand_hold_check(int *flag)
{
if (!*flag) {
cpufreq_ondemand_hold();
*flag = 1;
}
}
EXPORT_SYMBOL(cpufreq_ondemand_hold_check);
/* this tickle option waits until the CPU is up to the max freq before it returns */
void cpufreq_ondemand_hold_sync(void)
{
cpufreq_ondemand_hold();
flush_work(&tickle_state.tickle_work);
}
EXPORT_SYMBOL(cpufreq_ondemand_hold_sync);
void cpufreq_ondemand_unhold_check(int *flag)
{
if (*flag) {
cpufreq_ondemand_unhold();
*flag = 0;
}
}
EXPORT_SYMBOL(cpufreq_ondemand_unhold_check);
static void do_floor_state_change(struct work_struct *work)
{
int cpu, floor_active, floor_count, cur_freq_floor, pending_freq_floor;
unsigned long flags;
spin_lock_irqsave(&tickle_state.lock, flags);
floor_active = tickle_state.floor_active;
floor_count = tickle_state.floor_count;
cur_freq_floor = tickle_state.cur_freq_floor;
pending_freq_floor = get_max_client_floor();
spin_unlock_irqrestore(&tickle_state.lock, flags);
if (print_tickles)
printk("%s: floor_active=%u, floor_count=%u, "
"cur_freq_floor=%u, pending_freq_floor=%u\n",
__func__, floor_active, floor_count,
cur_freq_floor, pending_freq_floor);
if (floor_count) {
if (!floor_active || pending_freq_floor != cur_freq_floor) {
for_each_online_cpu(cpu) {
unsigned int f = pending_freq_floor;
struct cpu_dbs_info_s *dbs_info;
struct cpufreq_policy *policy;
mutex_lock(&dbs_mutex);
dbs_info = &per_cpu(od_cpu_dbs_info, cpu);
if (!dbs_info->enable) {
mutex_unlock(&dbs_mutex);
continue;
}
mutex_lock(&dbs_info->timer_mutex);
policy = dbs_info->cur_policy;
if (!policy) {
mutex_unlock(&dbs_info->timer_mutex);
mutex_unlock(&dbs_mutex);
continue;
}
if (f < policy->min)
f = policy->min;
if (f > policy->max)
f = policy->max;
dbs_info->freq_floor = f;
dbs_info->floor_active = 1;
if (policy->cur < f) {
record_sample(policy->cur, f, -31,
policy->cpu, NULL);
__cpufreq_driver_target(policy, f,
CPUFREQ_RELATION_L);
}
mutex_unlock(&dbs_info->timer_mutex);
mutex_unlock(&dbs_mutex);
}
}
floor_active = 1;
cur_freq_floor = pending_freq_floor;
} else if (floor_active) {
for_each_online_cpu(cpu) {
struct cpu_dbs_info_s *dbs_info;
struct cpufreq_policy *policy;
mutex_lock(&dbs_mutex);
dbs_info = &per_cpu(od_cpu_dbs_info, cpu);
if (!dbs_info->enable) {
mutex_unlock(&dbs_mutex);
continue;
}
mutex_lock(&dbs_info->timer_mutex);
policy = dbs_info->cur_policy;
dbs_info->floor_active = 0;
adjust_for_load(dbs_info);
mutex_unlock(&dbs_info->timer_mutex);
mutex_unlock(&dbs_mutex);
}
floor_active = 0;
cur_freq_floor = 0;
}
spin_lock_irqsave(&tickle_state.lock, flags);
tickle_state.floor_active = floor_active;
tickle_state.cur_freq_floor = cur_freq_floor;
spin_unlock_irqrestore(&tickle_state.lock, flags);
}
void cpufreq_ondemand_floor_hold(struct tickle_file_data *tfdp)
{
unsigned long flags;
unsigned int freq;
int cpu, index = 0;
struct cpu_dbs_info_s *dbs_info;
/* record it on CPU 0 */
record_sample(0, tfdp->floor_freq, -231, 0, NULL);
/* it is reasonable for a client to set a floor that is not an actual
* CPU frequency, but that will cause a bit of flailing on each sample.
* Round it up (once) as necessary here.
*/
cpu = smp_processor_id();
dbs_info = &per_cpu(od_cpu_dbs_info, cpu);
cpufreq_frequency_table_target(dbs_info->cur_policy,
dbs_info->freq_table, tfdp->floor_freq, CPUFREQ_RELATION_L,
&index);
freq = dbs_info->freq_table[index].frequency;
spin_lock_irqsave(&tickle_state.lock, flags);
tickle_state.floor_count += 1;
if (tfdp->floor_freq != freq) {
pr_debug("%s: changing floor from %u to %u\n",
__func__, tfdp->floor_freq, freq);
tfdp->floor_freq = freq ;
}
spin_unlock_irqrestore(&tickle_state.lock, flags);
queue_work(kondemand_wq, &tickle_state.floor_work);
}
void cpufreq_ondemand_floor_unhold(struct tickle_file_data *tfdp)
{
int queue = 0;
unsigned long flags;
/* record it on CPU 0 */
record_sample(0, tfdp->floor_freq, -230, 0, NULL);
spin_lock_irqsave(&tickle_state.lock, flags);
tfdp->floor_freq = 0;
if (tickle_state.floor_count) {
tickle_state.floor_count -= 1;
queue = 1;
} else {
printk(KERN_WARNING "%s: attempt to decrement when floor_count == 0!\n",
__func__);
}
spin_unlock_irqrestore(&tickle_state.lock, flags);
if (queue)
queue_work(kondemand_wq, &tickle_state.floor_work);
}
void cpufreq_ondemand_floor_hold_check(struct tickle_file_data *tfdp,
int *flag)
{
if (!*flag) {
cpufreq_ondemand_floor_hold(tfdp);
*flag = 1;
}
}
void cpufreq_ondemand_floor_unhold_check(struct tickle_file_data *tfdp,
int *flag)
{
if (*flag) {
cpufreq_ondemand_floor_unhold(tfdp);
*flag = 0;
}
}
/* allow a single kernel held floor using kernel_floor_data.
* Having more than one kernel floor will require a mechanism
* for creating and unregistering clients.
*/
int cpufreq_ondemand_floor_hold_kernel(unsigned int freq)
{
unsigned long flags;
if (freq == 0)
return -EINVAL;
spin_lock_irqsave(&tickle_state.lock, flags);
if (kernel_floor_data.floor_freq) { /* in use */
spin_unlock_irqrestore(&tickle_state.lock, flags);
return -EBUSY;
}
kernel_floor_data.floor_freq = freq;
spin_unlock_irqrestore(&tickle_state.lock, flags);
cpufreq_ondemand_floor_hold(&kernel_floor_data);
return 0;
}
EXPORT_SYMBOL(cpufreq_ondemand_floor_hold_kernel);
int cpufreq_ondemand_floor_unhold_kernel(void)
{
unsigned long flags;
spin_lock_irqsave(&tickle_state.lock, flags);
if (kernel_floor_data.floor_freq == 0) {
spin_unlock_irqrestore(&tickle_state.lock, flags);
return -EINVAL;
}
spin_unlock_irqrestore(&tickle_state.lock, flags);
cpufreq_ondemand_floor_unhold(&kernel_floor_data);
return 0;
}
EXPORT_SYMBOL(cpufreq_ondemand_floor_unhold_kernel);
/**
* @brief
*
* Should be called with the timer_mutex already held.
* dbs_check_cpu will be called every sample even when a tickle
* or floor is held (so the load data is always current), but if
* a tickle is held it will not change the frequency on that sample.
*
* @param this_dbs_info
*/
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->cur_load = 0;
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) cputime64_sub(cur_wall_time,
j_dbs_info->prev_cpu_wall);
j_dbs_info->prev_cpu_wall = cur_wall_time;
idle_time = (unsigned int) cputime64_sub(cur_idle_time,
j_dbs_info->prev_cpu_idle);
j_dbs_info->prev_cpu_idle = cur_idle_time;
iowait_time = (unsigned int) cputime64_sub(cur_iowait_time,
j_dbs_info->prev_cpu_iowait);
j_dbs_info->prev_cpu_iowait = cur_iowait_time;
if (dbs_tuners_ins.ignore_nice) {
cputime64_t cur_nice;
unsigned long cur_nice_jiffies;
cur_nice = cputime64_sub(kstat_cpu(j).cpustat.nice,
j_dbs_info->prev_cpu_nice);
/*
* Assumption: nice time between sampling periods will
* be less than 2^32 jiffies for 32 bit sys
*/
cur_nice_jiffies = (unsigned long)
cputime64_to_jiffies64(cur_nice);
j_dbs_info->prev_cpu_nice = kstat_cpu(j).cpustat.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;
this_dbs_info->cur_load = load;
/* stash values for later sample storage */
if (sampling_enabled) {
j_dbs_info->rlv.wall = wall_time;
j_dbs_info->rlv.idle = idle_time;
j_dbs_info->rlv.iowait = iowait_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;
}
this_dbs_info->max_load_freq = max_load_freq;
}
/**
* @brief
*
* Should be called with the timer_mutex already held.
*
* @param this_dbs_info
*/
static void adjust_for_load(struct cpu_dbs_info_s *this_dbs_info)
{
unsigned int max_load_freq;
unsigned int load;
struct cpufreq_policy* policy;
if (this_dbs_info->tickle_active) {
return;
}
max_load_freq = this_dbs_info->max_load_freq;
load = this_dbs_info->cur_load;
policy = this_dbs_info->cur_policy;
/* 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;
/* if we are already at full speed then break out early */
if (!dbs_tuners_ins.powersave_bias) {
if (policy->cur == policy->max)
return;
record_sample(policy->cur, policy->max, load,
policy->cpu, &this_dbs_info->rlv);
__cpufreq_driver_target(policy, policy->max,
CPUFREQ_RELATION_H);
} else {
int freq = powersave_bias_target(policy, policy->max,
CPUFREQ_RELATION_H);
if (this_dbs_info->floor_active) {
if (freq <= this_dbs_info->freq_floor) {
freq = this_dbs_info->freq_floor;
}
}
if (policy->cur != freq)
record_sample(policy->cur, freq, load,
policy->cpu, &this_dbs_info->rlv);
__cpufreq_driver_target(policy, freq,
CPUFREQ_RELATION_L);
}
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) {
if (this_dbs_info->floor_active) {
if (freq_next <= this_dbs_info->freq_floor) {
freq_next = this_dbs_info->freq_floor;
}
}
if (policy->cur != freq_next)
record_sample(policy->cur, freq_next, load,
policy->cpu, &this_dbs_info->rlv);
__cpufreq_driver_target(policy, freq_next,
CPUFREQ_RELATION_L);
} else {
int freq = powersave_bias_target(policy, freq_next,
CPUFREQ_RELATION_L);
if (this_dbs_info->floor_active) {
if (freq <= this_dbs_info->freq_floor) {
freq = this_dbs_info->freq_floor;
}
}
if (policy->cur != freq)
record_sample(policy->cur, freq, load,
policy->cpu, &this_dbs_info->rlv);
__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;
/* We want all CPUs to do sampling nearly on same jiffy */
int delay = usecs_to_jiffies(dbs_tuners_ins.sampling_rate
* dbs_info->rate_mult);
if (num_online_cpus() > 1)
delay -= jiffies % delay;
mutex_lock(&dbs_info->timer_mutex);
/* Common NORMAL_SAMPLE setup */
dbs_info->sample_type = DBS_NORMAL_SAMPLE;
if (!dbs_tuners_ins.powersave_bias ||
sample_type == DBS_NORMAL_SAMPLE) {
dbs_check_cpu(dbs_info);
adjust_for_load(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 {
record_sample(dbs_info->cur_policy->cur, dbs_info->freq_lo,
-1, cpu, NULL);
__cpufreq_driver_target(dbs_info->cur_policy,
dbs_info->freq_lo, CPUFREQ_RELATION_H);
}
queue_delayed_work_on(cpu, kondemand_wq, &dbs_info->work, delay);
mutex_unlock(&dbs_info->timer_mutex);
}
static void do_tickle_timer(unsigned long arg)
{
unsigned long flags;
spin_lock_irqsave(&tickle_state.lock, flags);
if (tickle_state.active_count) {
if (tickle_state.active_tickle) {
tickle_state.active_count -= 1;
tickle_state.active_tickle = 0;
}
} else {
printk(KERN_WARNING "%s: attempt to decrement when active_count == 0!\n",
__func__);
}
spin_unlock_irqrestore(&tickle_state.lock, flags);
queue_work(kondemand_wq, &tickle_state.tickle_work);
}
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;
dbs_info->enable = 1;
dbs_info->sample_type = DBS_NORMAL_SAMPLE;
INIT_DELAYED_WORK_DEFERRABLE(&dbs_info->work, do_dbs_timer);
queue_delayed_work_on(dbs_info->cpu, kondemand_wq, &dbs_info->work,
delay);
dbs_info->tickle_active = 0;
dbs_info->floor_active = 0;
}
static inline void dbs_timer_exit(struct cpu_dbs_info_s *dbs_info)
{
dbs_info->enable = 0;
cancel_delayed_work_sync(&dbs_info->work);
}
/*
* 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;
#endif
return 0;
}
static void dbs_refresh_callback(struct work_struct *unused)
{
struct cpufreq_policy *policy;
struct cpu_dbs_info_s *this_dbs_info;
mutex_lock(&dbs_mutex);
this_dbs_info = &per_cpu(od_cpu_dbs_info, 0);
if (!this_dbs_info->enable) {
mutex_unlock(&dbs_mutex);
return;
}
mutex_lock(&this_dbs_info->timer_mutex);
policy = this_dbs_info->cur_policy;
record_sample(0, 0, -251, policy->cpu, NULL);
if (policy->cur < policy->max) {
record_sample(policy->cur, policy->max, -51, policy->cpu, NULL);
policy->cur = policy->max;
__cpufreq_driver_target(policy, policy->max,
CPUFREQ_RELATION_L);
this_dbs_info->prev_cpu_idle = get_cpu_idle_time(0,
&this_dbs_info->prev_cpu_wall);
}
mutex_unlock(&this_dbs_info->timer_mutex);
mutex_unlock(&dbs_mutex);
}
static DECLARE_WORK(dbs_refresh_work, dbs_refresh_callback);
static void dbs_input_event(struct input_handle *handle, unsigned int type,
unsigned int code, int value)
{
schedule_work_on(0, &dbs_refresh_work);
}
static int dbs_input_connect(struct input_handler *handler,
struct input_dev *dev, const struct input_device_id *id)
{
struct input_handle *handle;
int error;
handle = kzalloc(sizeof(struct input_handle), GFP_KERNEL);
if (!handle)
return -ENOMEM;
handle->dev = dev;
handle->handler = handler;
handle->name = "cpufreq";
error = input_register_handle(handle);
if (error)
goto err2;
error = input_open_device(handle);
if (error)
goto err1;
return 0;
err1:
input_unregister_handle(handle);
err2:
kfree(handle);
return error;
}
static void dbs_input_disconnect(struct input_handle *handle)
{
input_close_device(handle);
input_unregister_handle(handle);
kfree(handle);
}
static const struct input_device_id dbs_ids[] = {
{ .driver_info = 1 },
{ },
};
static struct input_handler dbs_input_handler = {
.event = dbs_input_event,
.connect = dbs_input_connect,
.disconnect = dbs_input_disconnect,
.name = "cpufreq_ond_tcl",
.id_table = dbs_ids,
};
/* support for logging CPU hotplug in the sample file. Getting the
* frequency right is a little messy, hence the separate notifiers
* with different priorities. Only used if logging is enabled.
*/
static int cpufreq_hotplug_online_callback(struct notifier_block *nfb,
unsigned long action,
void *hcpu)
{ unsigned int cpu = (unsigned long)hcpu;
switch (action) {
case CPU_ONLINE:
record_sample(0, cpufreq_quick_get(cpu), -501, cpu, NULL);
break;
}
return NOTIFY_OK;
}
static int cpufreq_hotplug_offline_callback(struct notifier_block *nfb,
unsigned long action,
void *hcpu)
{ unsigned int cpu = (unsigned long)hcpu;
switch (action) {
case CPU_DOWN_PREPARE:
record_sample(cpufreq_quick_get(cpu), 0, -500, cpu, NULL);
break;
}
return NOTIFY_OK;
}
static struct notifier_block cpufreq_online_notifier = {
.notifier_call = cpufreq_hotplug_online_callback,
.priority = 0,
};
static struct notifier_block cpufreq_offline_notifier = {
.notifier_call = cpufreq_hotplug_offline_callback,
.priority = 1,
};
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;
struct proc_dir_entry *entry;
this_dbs_info = &per_cpu(od_cpu_dbs_info, cpu);
switch (event) {
case CPUFREQ_GOV_START:
if ((!cpu_online(cpu)) || (!policy->cur))
return -EINVAL;
mutex_lock(&dbs_mutex);
rc = sysfs_create_group(&policy->kobj, &dbs_attr_group_old);
if (rc) {
mutex_unlock(&dbs_mutex);
return rc;
}
dbs_enable++;
for_each_cpu(j, policy->cpus) {
struct cpu_dbs_info_s *j_dbs_info;
j_dbs_info = &per_cpu(od_cpu_dbs_info, j);
j_dbs_info->cur_policy = policy;
j_dbs_info->prev_cpu_idle = get_cpu_idle_time(j,
&j_dbs_info->prev_cpu_wall);
if (dbs_tuners_ins.ignore_nice) {
j_dbs_info->prev_cpu_nice =
kstat_cpu(j).cpustat.nice;
}
}
this_dbs_info->cpu = cpu;
this_dbs_info->rate_mult = 1;
ondemand_powersave_bias_init_cpu(cpu);
/* sampling must be enabled prior to becoming the active governor */
if (sampling_enabled) {
for_each_cpu(j, policy->cpus) {
struct sample_stats *stats = &per_cpu(sample_stats, j);
// if allocation fails, sampling is disabled on this cpu
if (!stats->samples) {
stats->samples = vmalloc(sizeof(struct sample_data) * NUM_SAMPLES);
if (!stats->samples)
continue;
/* if it's a new allocation, touch
* each page to force allocation of
* physical pages. This also initializes
* the structure to 0.
*/
memset(stats->samples, 0, sizeof(struct sample_data) *
NUM_SAMPLES);
stats->current_sample = 0;
}
}
}
/*
* Start the timerschedule work and create procfs entry
* for sampling statistics when this governor is used for
* the first time.
*/
if (dbs_enable == 1) {
unsigned int latency;
entry = create_proc_entry("ondemandtcl_samples", 0444, NULL);
if (entry)
entry->proc_fops = &proc_stats_operations;
if (sampling_enabled) {
int rc1 = 0;
rc = register_hotcpu_notifier(&cpufreq_online_notifier);
if (rc == 0) {
rc1 = register_hotcpu_notifier(&cpufreq_offline_notifier);
if (rc1)
unregister_hotcpu_notifier(&cpufreq_online_notifier);
}
if (rc || rc1)
pr_err("ondemandtcl not logging CPU hotplug\n");
}
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);
/* Until user-space startup scripts are set up,
* override it here
*/
if (DEF_SAMPLING_RATE > 0)
dbs_tuners_ins.sampling_rate = DEF_SAMPLING_RATE;
dbs_tuners_ins.io_is_busy = should_io_be_busy();
}
if (!cpu)
rc = input_register_handler(&dbs_input_handler);
mutex_unlock(&dbs_mutex);
mutex_init(&this_dbs_info->timer_mutex);
dbs_timer_init(this_dbs_info);
break;
case CPUFREQ_GOV_STOP:
dbs_timer_exit(this_dbs_info);
mutex_lock(&dbs_mutex);
sysfs_remove_group(&policy->kobj, &dbs_attr_group_old);
mutex_destroy(&this_dbs_info->timer_mutex);
dbs_enable--;
if (!cpu)
input_unregister_handler(&dbs_input_handler);
if (!dbs_enable) {
/* don't free sampling data for _any_ CPU until
* there's no one using this governor, then free it
* all. Implicitly assumes that all CPU's are using
* this governor, but will be harmless if not.
*/
for_each_possible_cpu(j) {
struct sample_stats *stats;
stats = &per_cpu(sample_stats, j);
if (stats->samples) {
vfree(stats->samples);
stats->samples = NULL;
}
}
/* if registration failed earlier, unregistration will
* benignly return an ENOENT error code)
*/
unregister_hotcpu_notifier(&cpufreq_online_notifier);
unregister_hotcpu_notifier(&cpufreq_offline_notifier);
remove_proc_entry("ondemandtcl_samples", NULL);
sysfs_remove_group(cpufreq_global_kobject,
&dbs_attr_group);
}
mutex_unlock(&dbs_mutex);
break;
case CPUFREQ_GOV_LIMITS:
mutex_lock(&this_dbs_info->timer_mutex);
record_sample(policy->min, policy->max, -240,
policy->cpu, NULL);
if (policy->max < this_dbs_info->cur_policy->cur) {
record_sample(policy->cur, policy->max, -40,
policy->cpu, NULL);
__cpufreq_driver_target(this_dbs_info->cur_policy,
policy->max,
CPUFREQ_RELATION_H);
} else if (policy->min > this_dbs_info->cur_policy->cur) {
record_sample(policy->cur, policy->min, -41,
policy->cpu, NULL);
__cpufreq_driver_target(this_dbs_info->cur_policy,
policy->min,
CPUFREQ_RELATION_L);
}
mutex_unlock(&this_dbs_info->timer_mutex);
break;
}
return 0;
}
#ifndef CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND_TICKLE
static
#endif
struct cpufreq_governor cpufreq_gov_ondemand_tickle = {
.name = "ondemandtcl",
.governor = cpufreq_governor_dbs,
.max_transition_latency = TRANSITION_LATENCY_LIMIT,
.owner = THIS_MODULE,
};
static int __init cpufreq_gov_dbs_init(void)
{
int err;
cputime64_t wall;
u64 idle_time;
int cpu = get_cpu();
idle_time = get_cpu_idle_time_us(cpu, &wall);
put_cpu();
if (idle_time != -1ULL) {
/* Idle micro accounting is supported. Use finer thresholds */
dbs_tuners_ins.up_threshold = MICRO_FREQUENCY_UP_THRESHOLD;
dbs_tuners_ins.down_differential =
MICRO_FREQUENCY_DOWN_DIFFERENTIAL;
/*
* In no_hz/micro accounting case we set the minimum frequency
* not depending on HZ, but fixed (very low). The deferred
* timer might skip some samples if idle/sleeping as needed.
*/
min_sampling_rate = MICRO_FREQUENCY_MIN_SAMPLE_RATE;
} else {
/* For correct statistics, we need 10 ticks for each measure */
min_sampling_rate =
MIN_SAMPLING_RATE_RATIO * jiffies_to_usecs(10);
}
kondemand_wq = create_workqueue("kondemandtcl");
if (!kondemand_wq) {
printk(KERN_ERR "Creation of kondemandtcl failed\n");
return -EFAULT;
}
init_timer(&tickle_state.tickle_timer);
tickle_state.tickle_timer.function = do_tickle_timer;
INIT_WORK(&tickle_state.tickle_work, do_tickle_state_change);
INIT_WORK(&tickle_state.floor_work, do_floor_state_change);
/* init this to current jiffies or short circuiting
* doesn't work until jiffies wraps
*/
tickle_state.tickle_jiffies = jiffies;
err = setup_tickle_device();
if (err < 0) {
return err;
}
err = cpufreq_register_governor(&cpufreq_gov_ondemand_tickle);
if (err)
destroy_workqueue(kondemand_wq);
return err;
}
static void __exit cpufreq_gov_dbs_exit(void)
{
remove_tickle_device();
cpufreq_unregister_governor(&cpufreq_gov_ondemand_tickle);
/* cancel any pending timers before destroying the work queue */
del_timer(&tickle_state.tickle_timer);
destroy_workqueue(kondemand_wq);
}
MODULE_AUTHOR("Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>");
MODULE_AUTHOR("Alexey Starikovskiy <alexey.y.starikovskiy@intel.com>");
MODULE_AUTHOR("Corey Tabaka <corey.tabaka@palm.com>");
MODULE_DESCRIPTION("'cpufreq_ondemand_tickle' - A dynamic cpufreq governor for "
"Low Latency Frequency Transition capable processors");
MODULE_LICENSE("GPL");
#ifdef CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND_TICKLE
fs_initcall(cpufreq_gov_dbs_init);
#else
module_init(cpufreq_gov_dbs_init);
#endif
module_exit(cpufreq_gov_dbs_exit);
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: Request for DIscussion: Cpufreq logging, and frequency floors
2011-10-21 21:31 Request for DIscussion: Cpufreq logging, and frequency floors Steven Finney (Palm GBU)
@ 2011-10-23 11:48 ` Mark Brown
2011-10-24 0:01 ` Dave Jones
0 siblings, 1 reply; 4+ messages in thread
From: Mark Brown @ 2011-10-23 11:48 UTC (permalink / raw)
To: Steven Finney (Palm GBU); +Cc: cpufreq
On Fri, Oct 21, 2011 at 02:31:57PM -0700, Steven Finney (Palm GBU) wrote:
> 2) The ability to keep a diagnostic log of all the frequency changes so,
> e.g., it's possible to determine if bad behavior (e.g. dropouts) is
> correlated with a low frequency.
This is a really good and useful idea but it seems to me like it would
be better done with the standard trace subsystem - that provides good
facilities for enabling and disabling the trace as needed and would make
it easy to tie in with the other subsystems that are in play.
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: Request for DIscussion: Cpufreq logging, and frequency floors
2011-10-23 11:48 ` Mark Brown
@ 2011-10-24 0:01 ` Dave Jones
2011-10-28 16:45 ` Steven Finney (Palm GBU)
0 siblings, 1 reply; 4+ messages in thread
From: Dave Jones @ 2011-10-24 0:01 UTC (permalink / raw)
To: Mark Brown; +Cc: Steven Finney (Palm GBU), cpufreq
On Sun, Oct 23, 2011 at 12:48:03PM +0100, Mark Brown wrote:
> On Fri, Oct 21, 2011 at 02:31:57PM -0700, Steven Finney (Palm GBU) wrote:
>
> > 2) The ability to keep a diagnostic log of all the frequency changes so,
> > e.g., it's possible to determine if bad behavior (e.g. dropouts) is
> > correlated with a low frequency.
>
> This is a really good and useful idea but it seems to me like it would
> be better done with the standard trace subsystem - that provides good
> facilities for enabling and disabling the trace as needed and would make
> it easy to tie in with the other subsystems that are in play.
Indeed. This sounds like the direction to go towards. Having played a little
with Steven Rostedt's kernelshark tool, I could see interesting things coming
from being able to correlate transitions with other system events graphically.
Dave
^ permalink raw reply [flat|nested] 4+ messages in thread
* RE: Request for DIscussion: Cpufreq logging, and frequency floors
2011-10-24 0:01 ` Dave Jones
@ 2011-10-28 16:45 ` Steven Finney (Palm GBU)
0 siblings, 0 replies; 4+ messages in thread
From: Steven Finney (Palm GBU) @ 2011-10-28 16:45 UTC (permalink / raw)
To: Dave Jones, Mark Brown, Dmitry Fink (Palm GBU); +Cc: cpufreq
> From: Dave Jones [davej@redhat.com]
> Sent: Sunday, October 23, 2011 5:01 PM
> Subject: Re: Request for DIscussion: Cpufreq logging, and frequency floors
>
> On Sun, Oct 23, 2011 at 12:48:03PM +0100, Mark Brown wrote:
> > On Fri, Oct 21, 2011 at 02:31:57PM -0700, Steven Finney (Palm GBU) wrote:
> >
> > > 2) The ability to keep a diagnostic log of all the frequency changes so,
> > > e.g., it's possible to determine if bad behavior (e.g. dropouts) is
> > > correlated with a low frequency.
> >
> > This is a really good and useful idea but it seems to me like it would
> > be better done with the standard trace subsystem - that provides good
> > facilities for enabling and disabling the trace as needed and would make
> > it easy to tie in with the other subsystems that are in play.
>
> Indeed. This sounds like the direction to go towards. Having played a little
> with Steven Rostedt's kernelshark tool, I could see interesting things coming
> from being able to correlate transitions with other system events graphically.
Actually, it turns out there's at least some trace-system based cpufreq
tracing that was added sometime after our 2.6.35 kernel; 2.6.38 cpufreq.c
contains a trace_cpu_frequency() added by Thomas Renninger (so: generic,
and not restricted to Intel architectures). Perhaps this
is adequate; I haven't had a chance to try it yet. Do you know of problems
with that implementation? (One thing that might be nice to add is the option
of logging the decision made at each sample, with the load value,rather
than just the actual cpufreq changes).
sf
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2011-10-28 16:45 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-10-21 21:31 Request for DIscussion: Cpufreq logging, and frequency floors Steven Finney (Palm GBU)
2011-10-23 11:48 ` Mark Brown
2011-10-24 0:01 ` Dave Jones
2011-10-28 16:45 ` Steven Finney (Palm GBU)
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.