summaryrefslogtreecommitdiffstats
path: root/kernel/time/timer_stats.c
blob: fa3d380ca8c0b1aeb77250743d4cca9e25cdc38c (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
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
/*
 * kernel/time/timer_stats.c
 *
 * Collect timer usage statistics.
 *
 * Copyright(C) 2006, Red Hat, Inc., Ingo Molnar
 * Copyright(C) 2006 Timesys Corp., Thomas Gleixner <tglx@timesys.com>
 *
 * timer_stats is based on timer_top, a similar functionality which was part of
 * Con Kolivas dyntick patch set. It was developed by Daniel Petrini at the
 * Instituto Nokia de Tecnologia - INdT - Manaus. timer_top's design was based
 * on dynamic allocation of the statistics entries and linear search based
 * lookup combined with a global lock, rather than the static array, hash
 * and per-CPU locking which is used by timer_stats. It was written for the
 * pre hrtimer kernel code and therefore did not take hrtimers into account.
 * Nevertheless it provided the base for the timer_stats implementation and
 * was a helpful source of inspiration. Kudos to Daniel and the Nokia folks
 * for this effort.
 *
 * timer_top.c is
 *	Copyright (C) 2005 Instituto Nokia de Tecnologia - INdT - Manaus
 *	Written by Daniel Petrini <d.pensator@gmail.com>
 *	timer_top.c was released under the GNU General Public License version 2
 *
 * We export the addresses and counting of timer functions being called,
 * the pid and cmdline from the owner process if applicable.
 *
 * Start/stop data collection:
 * # echo 1[0] >/proc/timer_stats
 *
 * Display the information collected so far:
 * # cat /proc/timer_stats
 *
 * 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.
 */

#include <linux/proc_fs.h>
#include <linux/module.h>
#include <linux/spinlock.h>
#include <linux/sched.h>
#include <linux/seq_file.h>
#include <linux/kallsyms.h>

#include <asm/uaccess.h>

/*
 * This is our basic unit of interest: a timer expiry event identified
 * by the timer, its start/expire functions and the PID of the task that
 * started the timer. We count the number of times an event happens:
 */
struct entry {
	/*
	 * Hash list:
	 */
	struct entry		*next;

	/*
	 * Hash keys:
	 */
	void			*timer;
	void			*start_func;
	void			*expire_func;
	pid_t			pid;

	/*
	 * Number of timeout events:
	 */
	unsigned long		count;

	/*
	 * We save the command-line string to preserve
	 * this information past task exit:
	 */
	char			comm[TASK_COMM_LEN + 1];

} ____cacheline_aligned_in_smp;

/*
 * Spinlock protecting the tables - not taken during lookup:
 */
static DEFINE_SPINLOCK(table_lock);

/*
 * Per-CPU lookup locks for fast hash lookup:
 */
static DEFINE_PER_CPU(spinlock_t, lookup_lock);

/*
 * Mutex to serialize state changes with show-stats activities:
 */
static DEFINE_MUTEX(show_mutex);

/*
 * Collection status, active/inactive:
 */
static int __read_mostly active;

/*
 * Beginning/end timestamps of measurement:
 */
static ktime_t time_start, time_stop;

/*
 * tstat entry structs only get allocated while collection is
 * active and never freed during that time - this simplifies
 * things quite a bit.
 *
 * They get freed when a new collection period is started.
 */
#define MAX_ENTRIES_BITS	10
#define MAX_ENTRIES		(1UL << MAX_ENTRIES_BITS)

static unsigned long nr_entries;
static struct entry entries[MAX_ENTRIES];

static atomic_t overflow_count;

/*
 * The entries are in a hash-table, for fast lookup:
 */
#define TSTAT_HASH_BITS		(MAX_ENTRIES_BITS - 1)
#define TSTAT_HASH_SIZE		(1UL << TSTAT_HASH_BITS)
#define TSTAT_HASH_MASK		(TSTAT_HASH_SIZE - 1)

#define __tstat_hashfn(entry)						\
	(((unsigned long)(entry)->timer       ^				\
	  (unsigned long)(entry)->start_func  ^				\
	  (unsigned long)(entry)->expire_func ^				\
	  (unsigned long)(entry)->pid		) & TSTAT_HASH_MASK)

#define tstat_hashentry(entry)	(tstat_hash_table + __tstat_hashfn(entry))

static struct entry *tstat_hash_table[TSTAT_HASH_SIZE] __read_mostly;

static void reset_entries(void)
{
	nr_entries = 0;
	memset(entries, 0, sizeof(entries));
	memset(tstat_hash_table, 0, sizeof(tstat_hash_table));
	atomic_set(&overflow_count, 0);
}

static struct entry *alloc_entry(void)
{
	if (nr_entries >= MAX_ENTRIES)
		return NULL;

	return entries + nr_entries++;
}

static int match_entries(struct entry *entry1, struct entry *entry2)
{
	return entry1->timer       == entry2->timer	  &&
	       entry1->start_func  == entry2->start_func  &&
	       entry1->expire_func == entry2->expire_func &&
	       entry1->pid	   == entry2->pid;
}

/*
 * Look up whether an entry matching this item is present
 * in the hash already. Must be called with irqs off and the
 * lookup lock held:
 */
static struct entry *tstat_lookup(struct entry *entry, char *comm)
{
	struct entry **head, *curr, *prev;

	head = tstat_hashentry(entry);
	curr = *head;

	/*
	 * The fastpath is when the entry is already hashed,
	 * we do this with the lookup lock held, but with the
	 * table lock not held:
	 */
	while (curr) {
		if (match_entries(curr, entry))
			return curr;

		curr = curr->next;
	}
	/*
	 * Slowpath: allocate, set up and link a new hash entry:
	 */
	prev = NULL;
	curr = *head;

	spin_lock(&table_lock);
	/*
	 * Make sure we have not raced with another CPU:
	 */
	while (curr) {
		if (match_entries(curr, entry))
			goto out_unlock;

		prev = curr;
		curr = curr->next;
	}

	curr = alloc_entry();
	if (curr) {
		*curr = *entry;
		curr->count = 0;
		curr->next = NULL;
		memcpy(curr->comm, comm, TASK_COMM_LEN);

		smp_mb(); /* Ensure that curr is initialized before insert */

		if (prev)
			prev->next = curr;
		else
			*head = curr;
	}
 out_unlock:
	spin_unlock(&table_lock);

	return curr;
}

/**
 * timer_stats_update_stats - Update the statistics for a timer.
 * @timer:	pointer to either a timer_list or a hrtimer
 * @pid:	the pid of the task which set up the timer
 * @startf:	pointer to the function which did the timer setup
 * @timerf:	pointer to the timer callback function of the timer
 * @comm:	name of the process which set up the timer
 *
 * When the timer is already registered, then the event counter is
 * incremented. Otherwise the timer is registered in a free slot.
 */
void timer_stats_update_stats(void *timer, pid_t pid, void *startf,
			      void *timerf, char * comm)
{
	/*
	 * It doesnt matter which lock we take:
	 */
	spinlock_t *lock = &per_cpu(lookup_lock, raw_smp_processor_id());
	struct entry *entry, input;
	unsigned long flags;

	input.timer = timer;
	input.start_func = startf;
	input.expire_func = timerf;
	input.pid = pid;

	spin_lock_irqsave(lock, flags);
	if (!active)
		goto out_unlock;

	entry = tstat_lookup(&input, comm);
	if (likely(entry))
		entry->count++;
	else
		atomic_inc(&overflow_count);

 out_unlock:
	spin_unlock_irqrestore(lock, flags);
}

static void print_name_offset(struct seq_file *m, unsigned long addr)
{
	char symname[KSYM_NAME_LEN+1];

	if (lookup_symbol_name(addr, symname) < 0)
		seq_printf(m, "<%p>", (void *)addr);
	else
		seq_printf(m, "%s", symname);
}

static int tstats_show(struct seq_file *m, void *v)
{
	struct timespec period;
	struct entry *entry;
	unsigned long ms;
	long events = 0;
	ktime_t time;
	int i;

	mutex_lock(&show_mutex);
	/*
	 * If still active then calculate up to now:
	 */
	if (active)
		time_stop = ktime_get();

	time = ktime_sub(time_stop, time_start);

	period = ktime_to_timespec(time);
	ms = period.tv_nsec / 1000000;

	seq_puts(m, "Timer Stats Version: v0.1\n");
	seq_printf(m, "Sample period: %ld.%03ld s\n", period.tv_sec, ms);
	if (atomic_read(&overflow_count))
		seq_printf(m, "Overflow: %d entries\n",
			atomic_read(&overflow_count));

	for (i = 0; i < nr_entries; i++) {
		entry = entries + i;
		seq_printf(m, "%4lu, %5d %-16s ",
				entry->count, entry->pid, entry->comm);

		print_name_offset(m, (unsigned long)entry->start_func);
		seq_puts(m, " (");
		print_name_offset(m, (unsigned long)entry->expire_func);
		seq_puts(m, ")\n");

		events += entry->count;
	}

	ms += period.tv_sec * 1000;
	if (!ms)
		ms = 1;

	if (events && period.tv_sec)
		seq_printf(m, "%ld total events, %ld.%ld events/sec\n", events,
			   events / period.tv_sec, events * 1000 / ms);
	else
		seq_printf(m, "%ld total events\n", events);

	mutex_unlock(&show_mutex);

	return 0;
}

/*
 * After a state change, make sure all concurrent lookup/update
 * activities have stopped:
 */
static void sync_access(void)
{
	unsigned long flags;
	int cpu;

	for_each_online_cpu(cpu) {
		spin_lock_irqsave(&per_cpu(lookup_lock, cpu), flags);
		/* nothing */
		spin_unlock_irqrestore(&per_cpu(lookup_lock, cpu), flags);
	}
}

static ssize_t tstats_write(struct file *file, const char __user *buf,
			    size_t count, loff_t *offs)
{
	char ctl[2];

	if (count != 2 || *offs)
		return -EINVAL;

	if (copy_from_user(ctl, buf, count))
		return -EFAULT;

	mutex_lock(&show_mutex);
	switch (ctl[0]) {
	case '0':
		if (active) {
			active = 0;
			time_stop = ktime_get();
			sync_access();
		}
		break;
	case '1':
		if (!active) {
			reset_entries();
			time_start = ktime_get();
			smp_mb();
			active = 1;
		}
		break;
	default:
		count = -EINVAL;
	}
	mutex_unlock(&show_mutex);

	return count;
}

static int tstats_open(struct inode *inode, struct file *filp)
{
	return single_open(filp, tstats_show, NULL);
}

static struct file_operations tstats_fops = {
	.open		= tstats_open,
	.read		= seq_read,
	.write		= tstats_write,
	.llseek		= seq_lseek,
	.release	= seq_release,
};

void __init init_timer_stats(void)
{
	int cpu;

	for_each_possible_cpu(cpu)
		spin_lock_init(&per_cpu(lookup_lock, cpu));
}

static int __init init_tstats_procfs(void)
{
	struct proc_dir_entry *pe;

	pe = create_proc_entry("timer_stats", 0644, NULL);
	if (!pe)
		return -ENOMEM;

	pe->proc_fops = &tstats_fops;

	return 0;
}
__initcall(init_tstats_procfs);