summaryrefslogblamecommitdiff
path: root/drivers/thermal/intel_powerclamp.c
blob: b40b37cd25e08b7b2353b6ef6d85c321ad2e08eb (plain) (tree)
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507



















































                                                                                
                           





































































































































































































































































































































































































































































                                                                                
                                                                             






























































































































































































































































































                                                                                
/*
 * intel_powerclamp.c - package c-state idle injection
 *
 * Copyright (c) 2012, Intel Corporation.
 *
 * Authors:
 *     Arjan van de Ven <arjan@linux.intel.com>
 *     Jacob Pan <jacob.jun.pan@linux.intel.com>
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License along with
 * this program; if not, write to the Free Software Foundation, Inc.,
 * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
 *
 *
 *	TODO:
 *           1. better handle wakeup from external interrupts, currently a fixed
 *              compensation is added to clamping duration when excessive amount
 *              of wakeups are observed during idle time. the reason is that in
 *              case of external interrupts without need for ack, clamping down
 *              cpu in non-irq context does not reduce irq. for majority of the
 *              cases, clamping down cpu does help reduce irq as well, we should
 *              be able to differenciate the two cases and give a quantitative
 *              solution for the irqs that we can control. perhaps based on
 *              get_cpu_iowait_time_us()
 *
 *	     2. synchronization with other hw blocks
 *
 *
 */

#define pr_fmt(fmt)	KBUILD_MODNAME ": " fmt

#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/delay.h>
#include <linux/kthread.h>
#include <linux/freezer.h>
#include <linux/cpu.h>
#include <linux/thermal.h>
#include <linux/slab.h>
#include <linux/tick.h>
#include <linux/debugfs.h>
#include <linux/seq_file.h>
#include <linux/sched/rt.h>

#include <asm/nmi.h>
#include <asm/msr.h>
#include <asm/mwait.h>
#include <asm/cpu_device_id.h>
#include <asm/idle.h>
#include <asm/hardirq.h>

#define MAX_TARGET_RATIO (50U)
/* For each undisturbed clamping period (no extra wake ups during idle time),
 * we increment the confidence counter for the given target ratio.
 * CONFIDENCE_OK defines the level where runtime calibration results are
 * valid.
 */
#define CONFIDENCE_OK (3)
/* Default idle injection duration, driver adjust sleep time to meet target
 * idle ratio. Similar to frequency modulation.
 */
#define DEFAULT_DURATION_JIFFIES (6)

static unsigned int target_mwait;
static struct dentry *debug_dir;

/* user selected target */
static unsigned int set_target_ratio;
static unsigned int current_ratio;
static bool should_skip;
static bool reduce_irq;
static atomic_t idle_wakeup_counter;
static unsigned int control_cpu; /* The cpu assigned to collect stat and update
				  * control parameters. default to BSP but BSP
				  * can be offlined.
				  */
static bool clamping;


static struct task_struct * __percpu *powerclamp_thread;
static struct thermal_cooling_device *cooling_dev;
static unsigned long *cpu_clamping_mask;  /* bit map for tracking per cpu
					   * clamping thread
					   */

static unsigned int duration;
static unsigned int pkg_cstate_ratio_cur;
static unsigned int window_size;

static int duration_set(const char *arg, const struct kernel_param *kp)
{
	int ret = 0;
	unsigned long new_duration;

	ret = kstrtoul(arg, 10, &new_duration);
	if (ret)
		goto exit;
	if (new_duration > 25 || new_duration < 6) {
		pr_err("Out of recommended range %lu, between 6-25ms\n",
			new_duration);
		ret = -EINVAL;
	}

	duration = clamp(new_duration, 6ul, 25ul);
	smp_mb();

exit:

	return ret;
}

static struct kernel_param_ops duration_ops = {
	.set = duration_set,
	.get = param_get_int,
};


module_param_cb(duration, &duration_ops, &duration, 0644);
MODULE_PARM_DESC(duration, "forced idle time for each attempt in msec.");

struct powerclamp_calibration_data {
	unsigned long confidence;  /* used for calibration, basically a counter
				    * gets incremented each time a clamping
				    * period is completed without extra wakeups
				    * once that counter is reached given level,
				    * compensation is deemed usable.
				    */
	unsigned long steady_comp; /* steady state compensation used when
				    * no extra wakeups occurred.
				    */
	unsigned long dynamic_comp; /* compensate excessive wakeup from idle
				     * mostly from external interrupts.
				     */
};

static struct powerclamp_calibration_data cal_data[MAX_TARGET_RATIO];

static int window_size_set(const char *arg, const struct kernel_param *kp)
{
	int ret = 0;
	unsigned long new_window_size;

	ret = kstrtoul(arg, 10, &new_window_size);
	if (ret)
		goto exit_win;
	if (new_window_size > 10 || new_window_size < 2) {
		pr_err("Out of recommended window size %lu, between 2-10\n",
			new_window_size);
		ret = -EINVAL;
	}

	window_size = clamp(new_window_size, 2ul, 10ul);
	smp_mb();

exit_win:

	return ret;
}

static struct kernel_param_ops window_size_ops = {
	.set = window_size_set,
	.get = param_get_int,
};

module_param_cb(window_size, &window_size_ops, &window_size, 0644);
MODULE_PARM_DESC(window_size, "sliding window in number of clamping cycles\n"
	"\tpowerclamp controls idle ratio within this window. larger\n"
	"\twindow size results in slower response time but more smooth\n"
	"\tclamping results. default to 2.");

static void find_target_mwait(void)
{
	unsigned int eax, ebx, ecx, edx;
	unsigned int highest_cstate = 0;
	unsigned int highest_subcstate = 0;
	int i;

	if (boot_cpu_data.cpuid_level < CPUID_MWAIT_LEAF)
		return;

	cpuid(CPUID_MWAIT_LEAF, &eax, &ebx, &ecx, &edx);

	if (!(ecx & CPUID5_ECX_EXTENSIONS_SUPPORTED) ||
	    !(ecx & CPUID5_ECX_INTERRUPT_BREAK))
		return;

	edx >>= MWAIT_SUBSTATE_SIZE;
	for (i = 0; i < 7 && edx; i++, edx >>= MWAIT_SUBSTATE_SIZE) {
		if (edx & MWAIT_SUBSTATE_MASK) {
			highest_cstate = i;
			highest_subcstate = edx & MWAIT_SUBSTATE_MASK;
		}
	}
	target_mwait = (highest_cstate << MWAIT_SUBSTATE_SIZE) |
		(highest_subcstate - 1);

}

static u64 pkg_state_counter(void)
{
	u64 val;
	u64 count = 0;

	static bool skip_c2;
	static bool skip_c3;
	static bool skip_c6;
	static bool skip_c7;

	if (!skip_c2) {
		if (!rdmsrl_safe(MSR_PKG_C2_RESIDENCY, &val))
			count += val;
		else
			skip_c2 = true;
	}

	if (!skip_c3) {
		if (!rdmsrl_safe(MSR_PKG_C3_RESIDENCY, &val))
			count += val;
		else
			skip_c3 = true;
	}

	if (!skip_c6) {
		if (!rdmsrl_safe(MSR_PKG_C6_RESIDENCY, &val))
			count += val;
		else
			skip_c6 = true;
	}

	if (!skip_c7) {
		if (!rdmsrl_safe(MSR_PKG_C7_RESIDENCY, &val))
			count += val;
		else
			skip_c7 = true;
	}

	return count;
}

static void noop_timer(unsigned long foo)
{
	/* empty... just the fact that we get the interrupt wakes us up */
}

static unsigned int get_compensation(int ratio)
{
	unsigned int comp = 0;

	/* we only use compensation if all adjacent ones are good */
	if (ratio == 1 &&
		cal_data[ratio].confidence >= CONFIDENCE_OK &&
		cal_data[ratio + 1].confidence >= CONFIDENCE_OK &&
		cal_data[ratio + 2].confidence >= CONFIDENCE_OK) {
		comp = (cal_data[ratio].steady_comp +
			cal_data[ratio + 1].steady_comp +
			cal_data[ratio + 2].steady_comp) / 3;
	} else if (ratio == MAX_TARGET_RATIO - 1 &&
		cal_data[ratio].confidence >= CONFIDENCE_OK &&
		cal_data[ratio - 1].confidence >= CONFIDENCE_OK &&
		cal_data[ratio - 2].confidence >= CONFIDENCE_OK) {
		comp = (cal_data[ratio].steady_comp +
			cal_data[ratio - 1].steady_comp +
			cal_data[ratio - 2].steady_comp) / 3;
	} else if (cal_data[ratio].confidence >= CONFIDENCE_OK &&
		cal_data[ratio - 1].confidence >= CONFIDENCE_OK &&
		cal_data[ratio + 1].confidence >= CONFIDENCE_OK) {
		comp = (cal_data[ratio].steady_comp +
			cal_data[ratio - 1].steady_comp +
			cal_data[ratio + 1].steady_comp) / 3;
	}

	/* REVISIT: simple penalty of double idle injection */
	if (reduce_irq)
		comp = ratio;
	/* do not exceed limit */
	if (comp + ratio >= MAX_TARGET_RATIO)
		comp = MAX_TARGET_RATIO - ratio - 1;

	return comp;
}

static void adjust_compensation(int target_ratio, unsigned int win)
{
	int delta;
	struct powerclamp_calibration_data *d = &cal_data[target_ratio];

	/*
	 * adjust compensations if confidence level has not been reached or
	 * there are too many wakeups during the last idle injection period, we
	 * cannot trust the data for compensation.
	 */
	if (d->confidence >= CONFIDENCE_OK ||
		atomic_read(&idle_wakeup_counter) >
		win * num_online_cpus())
		return;

	delta = set_target_ratio - current_ratio;
	/* filter out bad data */
	if (delta >= 0 && delta <= (1+target_ratio/10)) {
		if (d->steady_comp)
			d->steady_comp =
				roundup(delta+d->steady_comp, 2)/2;
		else
			d->steady_comp = delta;
		d->confidence++;
	}
}

static bool powerclamp_adjust_controls(unsigned int target_ratio,
				unsigned int guard, unsigned int win)
{
	static u64 msr_last, tsc_last;
	u64 msr_now, tsc_now;
	u64 val64;

	/* check result for the last window */
	msr_now = pkg_state_counter();
	rdtscll(tsc_now);

	/* calculate pkg cstate vs tsc ratio */
	if (!msr_last || !tsc_last)
		current_ratio = 1;
	else if (tsc_now-tsc_last) {
		val64 = 100*(msr_now-msr_last);
		do_div(val64, (tsc_now-tsc_last));
		current_ratio = val64;
	}

	/* update record */
	msr_last = msr_now;
	tsc_last = tsc_now;

	adjust_compensation(target_ratio, win);
	/*
	 * too many external interrupts, set flag such
	 * that we can take measure later.
	 */
	reduce_irq = atomic_read(&idle_wakeup_counter) >=
		2 * win * num_online_cpus();

	atomic_set(&idle_wakeup_counter, 0);
	/* if we are above target+guard, skip */
	return set_target_ratio + guard <= current_ratio;
}

static int clamp_thread(void *arg)
{
	int cpunr = (unsigned long)arg;
	DEFINE_TIMER(wakeup_timer, noop_timer, 0, 0);
	static const struct sched_param param = {
		.sched_priority = MAX_USER_RT_PRIO/2,
	};
	unsigned int count = 0;
	unsigned int target_ratio;

	set_bit(cpunr, cpu_clamping_mask);
	set_freezable();
	init_timer_on_stack(&wakeup_timer);
	sched_setscheduler(current, SCHED_FIFO, &param);

	while (true == clamping && !kthread_should_stop() &&
		cpu_online(cpunr)) {
		int sleeptime;
		unsigned long target_jiffies;
		unsigned int guard;
		unsigned int compensation = 0;
		int interval; /* jiffies to sleep for each attempt */
		unsigned int duration_jiffies = msecs_to_jiffies(duration);
		unsigned int window_size_now;

		try_to_freeze();
		/*
		 * make sure user selected ratio does not take effect until
		 * the next round. adjust target_ratio if user has changed
		 * target such that we can converge quickly.
		 */
		target_ratio = set_target_ratio;
		guard = 1 + target_ratio/20;
		window_size_now = window_size;
		count++;

		/*
		 * systems may have different ability to enter package level
		 * c-states, thus we need to compensate the injected idle ratio
		 * to achieve the actual target reported by the HW.
		 */
		compensation = get_compensation(target_ratio);
		interval = duration_jiffies*100/(target_ratio+compensation);

		/* align idle time */
		target_jiffies = roundup(jiffies, interval);
		sleeptime = target_jiffies - jiffies;
		if (sleeptime <= 0)
			sleeptime = 1;
		schedule_timeout_interruptible(sleeptime);
		/*
		 * only elected controlling cpu can collect stats and update
		 * control parameters.
		 */
		if (cpunr == control_cpu && !(count%window_size_now)) {
			should_skip =
				powerclamp_adjust_controls(target_ratio,
							guard, window_size_now);
			smp_mb();
		}

		if (should_skip)
			continue;

		target_jiffies = jiffies + duration_jiffies;
		mod_timer(&wakeup_timer, target_jiffies);
		if (unlikely(local_softirq_pending()))
			continue;
		/*
		 * stop tick sched during idle time, interrupts are still
		 * allowed. thus jiffies are updated properly.
		 */
		preempt_disable();
		tick_nohz_idle_enter();
		/* mwait until target jiffies is reached */
		while (time_before(jiffies, target_jiffies)) {
			unsigned long ecx = 1;
			unsigned long eax = target_mwait;

			/*
			 * REVISIT: may call enter_idle() to notify drivers who
			 * can save power during cpu idle. same for exit_idle()
			 */
			local_touch_nmi();
			stop_critical_timings();
			__monitor((void *)&current_thread_info()->flags, 0, 0);
			cpu_relax(); /* allow HT sibling to run */
			__mwait(eax, ecx);
			start_critical_timings();
			atomic_inc(&idle_wakeup_counter);
		}
		tick_nohz_idle_exit();
		preempt_enable_no_resched();
	}
	del_timer_sync(&wakeup_timer);
	clear_bit(cpunr, cpu_clamping_mask);

	return 0;
}

/*
 * 1 HZ polling while clamping is active, useful for userspace
 * to monitor actual idle ratio.
 */
static void poll_pkg_cstate(struct work_struct *dummy);
static DECLARE_DELAYED_WORK(poll_pkg_cstate_work, poll_pkg_cstate);
static void poll_pkg_cstate(struct work_struct *dummy)
{
	static u64 msr_last;
	static u64 tsc_last;
	static unsigned long jiffies_last;

	u64 msr_now;
	unsigned long jiffies_now;
	u64 tsc_now;
	u64 val64;

	msr_now = pkg_state_counter();
	rdtscll(tsc_now);
	jiffies_now = jiffies;

	/* calculate pkg cstate vs tsc ratio */
	if (!msr_last || !tsc_last)
		pkg_cstate_ratio_cur = 1;
	else {
		if (tsc_now - tsc_last) {
			val64 = 100 * (msr_now - msr_last);
			do_div(val64, (tsc_now - tsc_last));
			pkg_cstate_ratio_cur = val64;
		}
	}

	/* update record */
	msr_last = msr_now;
	jiffies_last = jiffies_now;
	tsc_last = tsc_now;

	if (true == clamping)
		schedule_delayed_work(&poll_pkg_cstate_work, HZ);
}

static int start_power_clamp(void)
{
	unsigned long cpu;
	struct task_struct *thread;

	/* check if pkg cstate counter is completely 0, abort in this case */
	if (!pkg_state_counter()) {
		pr_err("pkg cstate counter not functional, abort\n");
		return -EINVAL;
	}

	set_target_ratio = clamp(set_target_ratio, 0U, MAX_TARGET_RATIO - 1);
	/* prevent cpu hotplug */
	get_online_cpus();

	/* prefer BSP */
	control_cpu = 0;
	if (!cpu_online(control_cpu))
		control_cpu = smp_processor_id();

	clamping = true;
	schedule_delayed_work(&poll_pkg_cstate_work, 0);

	/* start one thread per online cpu */
	for_each_online_cpu(cpu) {
		struct task_struct **p =
			per_cpu_ptr(powerclamp_thread, cpu);

		thread = kthread_create_on_node(clamp_thread,
						(void *) cpu,
						cpu_to_node(cpu),
						"kidle_inject/%ld", cpu);
		/* bind to cpu here */
		if (likely(!IS_ERR(thread))) {
			kthread_bind(thread, cpu);
			wake_up_process(thread);
			*p = thread;
		}

	}
	put_online_cpus();

	return 0;
}

static void end_power_clamp(void)
{
	int i;
	struct task_struct *thread;

	clamping = false;
	/*
	 * make clamping visible to other cpus and give per cpu clamping threads
	 * sometime to exit, or gets killed later.
	 */
	smp_mb();
	msleep(20);
	if (bitmap_weight(cpu_clamping_mask, num_possible_cpus())) {
		for_each_set_bit(i, cpu_clamping_mask, num_possible_cpus()) {
			pr_debug("clamping thread for cpu %d alive, kill\n", i);
			thread = *per_cpu_ptr(powerclamp_thread, i);
			kthread_stop(thread);
		}
	}
}

static int powerclamp_cpu_callback(struct notifier_block *nfb,
				unsigned long action, void *hcpu)
{
	unsigned long cpu = (unsigned long)hcpu;
	struct task_struct *thread;
	struct task_struct **percpu_thread =
		per_cpu_ptr(powerclamp_thread, cpu);

	if (false == clamping)
		goto exit_ok;

	switch (action) {
	case CPU_ONLINE:
		thread = kthread_create_on_node(clamp_thread,
						(void *) cpu,
						cpu_to_node(cpu),
						"kidle_inject/%lu", cpu);
		if (likely(!IS_ERR(thread))) {
			kthread_bind(thread, cpu);
			wake_up_process(thread);
			*percpu_thread = thread;
		}
		/* prefer BSP as controlling CPU */
		if (cpu == 0) {
			control_cpu = 0;
			smp_mb();
		}
		break;
	case CPU_DEAD:
		if (test_bit(cpu, cpu_clamping_mask)) {
			pr_err("cpu %lu dead but powerclamping thread is not\n",
				cpu);
			kthread_stop(*percpu_thread);
		}
		if (cpu == control_cpu) {
			control_cpu = smp_processor_id();
			smp_mb();
		}
	}

exit_ok:
	return NOTIFY_OK;
}

static struct notifier_block powerclamp_cpu_notifier = {
	.notifier_call = powerclamp_cpu_callback,
};

static int powerclamp_get_max_state(struct thermal_cooling_device *cdev,
				 unsigned long *state)
{
	*state = MAX_TARGET_RATIO;

	return 0;
}

static int powerclamp_get_cur_state(struct thermal_cooling_device *cdev,
				 unsigned long *state)
{
	if (true == clamping)
		*state = pkg_cstate_ratio_cur;
	else
		/* to save power, do not poll idle ratio while not clamping */
		*state = -1; /* indicates invalid state */

	return 0;
}

static int powerclamp_set_cur_state(struct thermal_cooling_device *cdev,
				 unsigned long new_target_ratio)
{
	int ret = 0;

	new_target_ratio = clamp(new_target_ratio, 0UL,
				(unsigned long) (MAX_TARGET_RATIO-1));
	if (set_target_ratio == 0 && new_target_ratio > 0) {
		pr_info("Start idle injection to reduce power\n");
		set_target_ratio = new_target_ratio;
		ret = start_power_clamp();
		goto exit_set;
	} else	if (set_target_ratio > 0 && new_target_ratio == 0) {
		pr_info("Stop forced idle injection\n");
		set_target_ratio = 0;
		end_power_clamp();
	} else	/* adjust currently running */ {
		set_target_ratio = new_target_ratio;
		/* make new set_target_ratio visible to other cpus */
		smp_mb();
	}

exit_set:
	return ret;
}

/* bind to generic thermal layer as cooling device*/
static struct thermal_cooling_device_ops powerclamp_cooling_ops = {
	.get_max_state = powerclamp_get_max_state,
	.get_cur_state = powerclamp_get_cur_state,
	.set_cur_state = powerclamp_set_cur_state,
};

/* runs on Nehalem and later */
static const struct x86_cpu_id intel_powerclamp_ids[] = {
	{ X86_VENDOR_INTEL, 6, 0x1a},
	{ X86_VENDOR_INTEL, 6, 0x1c},
	{ X86_VENDOR_INTEL, 6, 0x1e},
	{ X86_VENDOR_INTEL, 6, 0x1f},
	{ X86_VENDOR_INTEL, 6, 0x25},
	{ X86_VENDOR_INTEL, 6, 0x26},
	{ X86_VENDOR_INTEL, 6, 0x2a},
	{ X86_VENDOR_INTEL, 6, 0x2c},
	{ X86_VENDOR_INTEL, 6, 0x2d},
	{ X86_VENDOR_INTEL, 6, 0x2e},
	{ X86_VENDOR_INTEL, 6, 0x2f},
	{ X86_VENDOR_INTEL, 6, 0x3a},
	{}
};
MODULE_DEVICE_TABLE(x86cpu, intel_powerclamp_ids);

static int powerclamp_probe(void)
{
	if (!x86_match_cpu(intel_powerclamp_ids)) {
		pr_err("Intel powerclamp does not run on family %d model %d\n",
				boot_cpu_data.x86, boot_cpu_data.x86_model);
		return -ENODEV;
	}
	if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC) ||
		!boot_cpu_has(X86_FEATURE_CONSTANT_TSC) ||
		!boot_cpu_has(X86_FEATURE_MWAIT) ||
		!boot_cpu_has(X86_FEATURE_ARAT))
		return -ENODEV;

	/* find the deepest mwait value */
	find_target_mwait();

	return 0;
}

static int powerclamp_debug_show(struct seq_file *m, void *unused)
{
	int i = 0;

	seq_printf(m, "controlling cpu: %d\n", control_cpu);
	seq_printf(m, "pct confidence steady dynamic (compensation)\n");
	for (i = 0; i < MAX_TARGET_RATIO; i++) {
		seq_printf(m, "%d\t%lu\t%lu\t%lu\n",
			i,
			cal_data[i].confidence,
			cal_data[i].steady_comp,
			cal_data[i].dynamic_comp);
	}

	return 0;
}

static int powerclamp_debug_open(struct inode *inode,
			struct file *file)
{
	return single_open(file, powerclamp_debug_show, inode->i_private);
}

static const struct file_operations powerclamp_debug_fops = {
	.open		= powerclamp_debug_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
	.owner		= THIS_MODULE,
};

static inline void powerclamp_create_debug_files(void)
{
	debug_dir = debugfs_create_dir("intel_powerclamp", NULL);
	if (!debug_dir)
		return;

	if (!debugfs_create_file("powerclamp_calib", S_IRUGO, debug_dir,
					cal_data, &powerclamp_debug_fops))
		goto file_error;

	return;

file_error:
	debugfs_remove_recursive(debug_dir);
}

static int powerclamp_init(void)
{
	int retval;
	int bitmap_size;

	bitmap_size = BITS_TO_LONGS(num_possible_cpus()) * sizeof(long);
	cpu_clamping_mask = kzalloc(bitmap_size, GFP_KERNEL);
	if (!cpu_clamping_mask)
		return -ENOMEM;

	/* probe cpu features and ids here */
	retval = powerclamp_probe();
	if (retval)
		return retval;
	/* set default limit, maybe adjusted during runtime based on feedback */
	window_size = 2;
	register_hotcpu_notifier(&powerclamp_cpu_notifier);
	powerclamp_thread = alloc_percpu(struct task_struct *);
	cooling_dev = thermal_cooling_device_register("intel_powerclamp", NULL,
						&powerclamp_cooling_ops);
	if (IS_ERR(cooling_dev))
		return -ENODEV;

	if (!duration)
		duration = jiffies_to_msecs(DEFAULT_DURATION_JIFFIES);
	powerclamp_create_debug_files();

	return 0;
}
module_init(powerclamp_init);

static void powerclamp_exit(void)
{
	unregister_hotcpu_notifier(&powerclamp_cpu_notifier);
	end_power_clamp();
	free_percpu(powerclamp_thread);
	thermal_cooling_device_unregister(cooling_dev);
	kfree(cpu_clamping_mask);

	cancel_delayed_work_sync(&poll_pkg_cstate_work);
	debugfs_remove_recursive(debug_dir);
}
module_exit(powerclamp_exit);

MODULE_LICENSE("GPL");
MODULE_AUTHOR("Arjan van de Ven <arjan@linux.intel.com>");
MODULE_AUTHOR("Jacob Pan <jacob.jun.pan@linux.intel.com>");
MODULE_DESCRIPTION("Package Level C-state Idle Injection for Intel CPUs");