delayacct.c 4.62 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
/* delayacct.c - per-task delay accounting
 *
 * Copyright (C) Shailabh Nagar, IBM Corp. 2006
 *
 * This program is free software;  you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it would 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.
 */

#include <linux/sched.h>
17
#include <linux/sched/task.h>
18
#include <linux/sched/cputime.h>
19
#include <linux/slab.h>
20
#include <linux/taskstats.h>
21 22 23
#include <linux/time.h>
#include <linux/sysctl.h>
#include <linux/delayacct.h>
Glauber Costa's avatar
Glauber Costa committed
24
#include <linux/module.h>
25

26
int delayacct_on __read_mostly = 1;	/* Delay accounting turned on/off */
Glauber Costa's avatar
Glauber Costa committed
27
EXPORT_SYMBOL_GPL(delayacct_on);
28
struct kmem_cache *delayacct_cache;
29

30
static int __init delayacct_setup_disable(char *str)
31
{
32
	delayacct_on = 0;
33 34
	return 1;
}
35
__setup("nodelayacct", delayacct_setup_disable);
36 37 38

void delayacct_init(void)
{
39
	delayacct_cache = KMEM_CACHE(task_delay_info, SLAB_PANIC|SLAB_ACCOUNT);
40 41 42 43 44
	delayacct_tsk_init(&init_task);
}

void __delayacct_tsk_init(struct task_struct *tsk)
{
45
	tsk->delays = kmem_cache_zalloc(delayacct_cache, GFP_KERNEL);
46 47 48 49 50
	if (tsk->delays)
		spin_lock_init(&tsk->delays->lock);
}

/*
51 52
 * Finish delay accounting for a statistic using its timestamps (@start),
 * accumalator (@total) and @count
53
 */
54
static void delayacct_end(spinlock_t *lock, u64 *start, u64 *total, u32 *count)
55
{
56
	s64 ns = ktime_get_ns() - *start;
57
	unsigned long flags;
58

59
	if (ns > 0) {
60
		spin_lock_irqsave(lock, flags);
61 62
		*total += ns;
		(*count)++;
63
		spin_unlock_irqrestore(lock, flags);
64
	}
65 66
}

67 68
void __delayacct_blkio_start(void)
{
69
	current->delays->blkio_start = ktime_get_ns();
70 71
}

72 73 74 75 76
/*
 * We cannot rely on the `current` macro, as we haven't yet switched back to
 * the process being woken.
 */
void __delayacct_blkio_end(struct task_struct *p)
77
{
78 79 80 81 82 83 84 85 86 87 88 89 90
	struct task_delay_info *delays = p->delays;
	u64 *total;
	u32 *count;

	if (p->delays->flags & DELAYACCT_PF_SWAPIN) {
		total = &delays->swapin_delay;
		count = &delays->swapin_count;
	} else {
		total = &delays->blkio_delay;
		count = &delays->blkio_count;
	}

	delayacct_end(&delays->lock, &delays->blkio_start, total, count);
91
}
92 93 94

int __delayacct_add_tsk(struct taskstats *d, struct task_struct *tsk)
{
95
	u64 utime, stime, stimescaled, utimescaled;
96 97 98
	unsigned long long t2, t3;
	unsigned long flags, t1;
	s64 tmp;
99

100
	task_cputime(tsk, &utime, &stime);
101
	tmp = (s64)d->cpu_run_real_total;
102
	tmp += utime + stime;
103 104
	d->cpu_run_real_total = (tmp < (s64)d->cpu_run_real_total) ? 0 : tmp;

105
	task_cputime_scaled(tsk, &utimescaled, &stimescaled);
106
	tmp = (s64)d->cpu_scaled_run_real_total;
107
	tmp += utimescaled + stimescaled;
108 109 110
	d->cpu_scaled_run_real_total =
		(tmp < (s64)d->cpu_scaled_run_real_total) ? 0 : tmp;

111 112 113 114
	/*
	 * No locking available for sched_info (and too expensive to add one)
	 * Mitigate by taking snapshot of values
	 */
115
	t1 = tsk->sched_info.pcount;
116
	t2 = tsk->sched_info.run_delay;
117
	t3 = tsk->se.sum_exec_runtime;
118 119 120

	d->cpu_count += t1;

121
	tmp = (s64)d->cpu_delay_total + t2;
122 123
	d->cpu_delay_total = (tmp < (s64)d->cpu_delay_total) ? 0 : tmp;

124
	tmp = (s64)d->cpu_run_virtual_total + t3;
125 126 127 128 129
	d->cpu_run_virtual_total =
		(tmp < (s64)d->cpu_run_virtual_total) ?	0 : tmp;

	/* zero XXX_total, non-zero XXX_count implies XXX stat overflowed */

130
	spin_lock_irqsave(&tsk->delays->lock, flags);
131 132 133 134
	tmp = d->blkio_delay_total + tsk->delays->blkio_delay;
	d->blkio_delay_total = (tmp < d->blkio_delay_total) ? 0 : tmp;
	tmp = d->swapin_delay_total + tsk->delays->swapin_delay;
	d->swapin_delay_total = (tmp < d->swapin_delay_total) ? 0 : tmp;
135 136
	tmp = d->freepages_delay_total + tsk->delays->freepages_delay;
	d->freepages_delay_total = (tmp < d->freepages_delay_total) ? 0 : tmp;
137 138
	d->blkio_count += tsk->delays->blkio_count;
	d->swapin_count += tsk->delays->swapin_count;
139
	d->freepages_count += tsk->delays->freepages_count;
140
	spin_unlock_irqrestore(&tsk->delays->lock, flags);
141 142 143

	return 0;
}
144 145 146 147

__u64 __delayacct_blkio_ticks(struct task_struct *tsk)
{
	__u64 ret;
148
	unsigned long flags;
149

150
	spin_lock_irqsave(&tsk->delays->lock, flags);
151 152
	ret = nsec_to_clock_t(tsk->delays->blkio_delay +
				tsk->delays->swapin_delay);
153
	spin_unlock_irqrestore(&tsk->delays->lock, flags);
154 155 156
	return ret;
}

157 158
void __delayacct_freepages_start(void)
{
159
	current->delays->freepages_start = ktime_get_ns();
160 161 162 163
}

void __delayacct_freepages_end(void)
{
164 165 166 167 168
	delayacct_end(
		&current->delays->lock,
		&current->delays->freepages_start,
		&current->delays->freepages_delay,
		&current->delays->freepages_count);
169 170
}