sched_stats.h 9.25 KB
Newer Older
1 2 3 4 5 6

#ifdef CONFIG_SCHEDSTATS
/*
 * bump this up when changing the output format or the meaning of an existing
 * format, so that tools can adapt (or abort)
 */
7
#define SCHEDSTAT_VERSION 15
8 9 10 11

static int show_schedstat(struct seq_file *seq, void *v)
{
	int cpu;
12
	int mask_len = DIV_ROUND_UP(NR_CPUS, 32) * 9;
13 14 15 16
	char *mask_str = kmalloc(mask_len, GFP_KERNEL);

	if (mask_str == NULL)
		return -ENOMEM;
17 18 19 20 21 22 23

	seq_printf(seq, "version %d\n", SCHEDSTAT_VERSION);
	seq_printf(seq, "timestamp %lu\n", jiffies);
	for_each_online_cpu(cpu) {
		struct rq *rq = cpu_rq(cpu);
#ifdef CONFIG_SMP
		struct sched_domain *sd;
24
		int dcount = 0;
25 26 27 28
#endif

		/* runqueue-specific stats */
		seq_printf(seq,
29 30
		    "cpu%d %u %u %u %u %u %u %llu %llu %lu",
		    cpu, rq->yld_count,
31 32
		    rq->sched_switch, rq->sched_count, rq->sched_goidle,
		    rq->ttwu_count, rq->ttwu_local,
33
		    rq->rq_cpu_time,
34
		    rq->rq_sched_info.run_delay, rq->rq_sched_info.pcount);
35 36 37 38 39 40 41 42 43

		seq_printf(seq, "\n");

#ifdef CONFIG_SMP
		/* domain-specific stats */
		preempt_disable();
		for_each_domain(cpu, sd) {
			enum cpu_idle_type itype;

44
			cpumask_scnprintf(mask_str, mask_len,
Rusty Russell's avatar
Rusty Russell committed
45
					  sched_domain_span(sd));
46
			seq_printf(seq, "domain%d %s", dcount++, mask_str);
47 48
			for (itype = CPU_IDLE; itype < CPU_MAX_IDLE_TYPES;
					itype++) {
49
				seq_printf(seq, " %u %u %u %u %u %u %u %u",
50
				    sd->lb_count[itype],
51 52 53 54 55 56 57 58
				    sd->lb_balanced[itype],
				    sd->lb_failed[itype],
				    sd->lb_imbalance[itype],
				    sd->lb_gained[itype],
				    sd->lb_hot_gained[itype],
				    sd->lb_nobusyq[itype],
				    sd->lb_nobusyg[itype]);
			}
59 60
			seq_printf(seq,
				   " %u %u %u %u %u %u %u %u %u %u %u %u\n",
61 62 63
			    sd->alb_count, sd->alb_failed, sd->alb_pushed,
			    sd->sbe_count, sd->sbe_balanced, sd->sbe_pushed,
			    sd->sbf_count, sd->sbf_balanced, sd->sbf_pushed,
64 65 66 67 68 69
			    sd->ttwu_wake_remote, sd->ttwu_move_affine,
			    sd->ttwu_move_balance);
		}
		preempt_enable();
#endif
	}
Adrian Bunk's avatar
Adrian Bunk committed
70
	kfree(mask_str);
71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92
	return 0;
}

static int schedstat_open(struct inode *inode, struct file *file)
{
	unsigned int size = PAGE_SIZE * (1 + num_online_cpus() / 32);
	char *buf = kmalloc(size, GFP_KERNEL);
	struct seq_file *m;
	int res;

	if (!buf)
		return -ENOMEM;
	res = single_open(file, show_schedstat, NULL);
	if (!res) {
		m = file->private_data;
		m->buf = buf;
		m->size = size;
	} else
		kfree(buf);
	return res;
}

93
static const struct file_operations proc_schedstat_operations = {
94 95 96 97 98 99
	.open    = schedstat_open,
	.read    = seq_read,
	.llseek  = seq_lseek,
	.release = single_release,
};

100 101 102 103 104 105 106
static int __init proc_schedstat_init(void)
{
	proc_create("schedstat", 0, NULL, &proc_schedstat_operations);
	return 0;
}
module_init(proc_schedstat_init);

107 108 109 110 111 112 113 114
/*
 * Expects runqueue lock to be held for atomicity of update
 */
static inline void
rq_sched_info_arrive(struct rq *rq, unsigned long long delta)
{
	if (rq) {
		rq->rq_sched_info.run_delay += delta;
115
		rq->rq_sched_info.pcount++;
116 117 118 119 120 121 122 123 124 125
	}
}

/*
 * Expects runqueue lock to be held for atomicity of update
 */
static inline void
rq_sched_info_depart(struct rq *rq, unsigned long long delta)
{
	if (rq)
126
		rq->rq_cpu_time += delta;
127
}
128 129 130 131 132 133 134

static inline void
rq_sched_info_dequeued(struct rq *rq, unsigned long long delta)
{
	if (rq)
		rq->rq_sched_info.run_delay += delta;
}
135 136
# define schedstat_inc(rq, field)	do { (rq)->field++; } while (0)
# define schedstat_add(rq, field, amt)	do { (rq)->field += (amt); } while (0)
137
# define schedstat_set(var, val)	do { var = (val); } while (0)
138 139 140 141 142
#else /* !CONFIG_SCHEDSTATS */
static inline void
rq_sched_info_arrive(struct rq *rq, unsigned long long delta)
{}
static inline void
143 144 145
rq_sched_info_dequeued(struct rq *rq, unsigned long long delta)
{}
static inline void
146 147 148 149
rq_sched_info_depart(struct rq *rq, unsigned long long delta)
{}
# define schedstat_inc(rq, field)	do { } while (0)
# define schedstat_add(rq, field, amt)	do { } while (0)
150
# define schedstat_set(var, val)	do { } while (0)
151 152
#endif

153
#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
154 155 156 157 158
static inline void sched_info_reset_dequeued(struct task_struct *t)
{
	t->sched_info.last_queued = 0;
}

159
/*
160
 * We are interested in knowing how long it was from the *first* time a
161 162 163
 * task was queued to the time that it finally hit a cpu, we call this routine
 * from dequeue_task() to account for possible rq->clock skew across cpus. The
 * delta taken on each cpu would annul the skew.
164 165 166
 */
static inline void sched_info_dequeued(struct task_struct *t)
{
167 168 169 170 171 172 173 174 175
	unsigned long long now = task_rq(t)->clock, delta = 0;

	if (unlikely(sched_info_on()))
		if (t->sched_info.last_queued)
			delta = now - t->sched_info.last_queued;
	sched_info_reset_dequeued(t);
	t->sched_info.run_delay += delta;

	rq_sched_info_dequeued(task_rq(t), delta);
176 177 178 179 180 181 182 183 184
}

/*
 * Called when a task finally hits the cpu.  We can now calculate how
 * long it was waiting to run.  We also note when it began so that we
 * can keep stats on how long its timeslice is.
 */
static void sched_info_arrive(struct task_struct *t)
{
185
	unsigned long long now = task_rq(t)->clock, delta = 0;
186 187 188

	if (t->sched_info.last_queued)
		delta = now - t->sched_info.last_queued;
189
	sched_info_reset_dequeued(t);
190 191
	t->sched_info.run_delay += delta;
	t->sched_info.last_arrival = now;
192
	t->sched_info.pcount++;
193 194 195 196 197 198 199 200 201 202 203 204 205

	rq_sched_info_arrive(task_rq(t), delta);
}

/*
 * This function is only called from enqueue_task(), but also only updates
 * the timestamp if it is already not set.  It's assumed that
 * sched_info_dequeued() will clear that stamp when appropriate.
 */
static inline void sched_info_queued(struct task_struct *t)
{
	if (unlikely(sched_info_on()))
		if (!t->sched_info.last_queued)
206
			t->sched_info.last_queued = task_rq(t)->clock;
207 208 209 210 211
}

/*
 * Called when a process ceases being the active-running process, either
 * voluntarily or involuntarily.  Now we can calculate how long we ran.
212 213 214
 * Also, if the process is still in the TASK_RUNNING state, call
 * sched_info_queued() to mark that it has now again started waiting on
 * the runqueue.
215 216 217
 */
static inline void sched_info_depart(struct task_struct *t)
{
218 219
	unsigned long long delta = task_rq(t)->clock -
					t->sched_info.last_arrival;
220 221

	rq_sched_info_depart(task_rq(t), delta);
222 223 224

	if (t->state == TASK_RUNNING)
		sched_info_queued(t);
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
}

/*
 * Called when tasks are switched involuntarily due, typically, to expiring
 * their time slice.  (This may also be called when switching to or from
 * the idle task.)  We are only called when prev != next.
 */
static inline void
__sched_info_switch(struct task_struct *prev, struct task_struct *next)
{
	struct rq *rq = task_rq(prev);

	/*
	 * prev now departs the cpu.  It's not interesting to record
	 * stats about how efficient we were at scheduling the idle
	 * process, however.
	 */
	if (prev != rq->idle)
		sched_info_depart(prev);

	if (next != rq->idle)
		sched_info_arrive(next);
}
static inline void
sched_info_switch(struct task_struct *prev, struct task_struct *next)
{
	if (unlikely(sched_info_on()))
		__sched_info_switch(prev, next);
}
#else
255 256 257 258
#define sched_info_queued(t)			do { } while (0)
#define sched_info_reset_dequeued(t)	do { } while (0)
#define sched_info_dequeued(t)			do { } while (0)
#define sched_info_switch(t, next)		do { } while (0)
259
#endif /* CONFIG_SCHEDSTATS || CONFIG_TASK_DELAY_ACCT */
260

261 262 263 264 265 266 267
/*
 * The following are functions that support scheduler-internal time accounting.
 * These functions are generally called at the timer tick.  None of this depends
 * on CONFIG_SCHEDSTATS.
 */

/**
268
 * account_group_user_time - Maintain utime for a thread group.
269
 *
270 271 272
 * @tsk:	Pointer to task structure.
 * @cputime:	Time value by which to increment the utime field of the
 *		thread_group_cputime structure.
273 274 275 276
 *
 * If thread group time is being maintained, get the structure for the
 * running CPU and update the utime field there.
 */
277 278
static inline void account_group_user_time(struct task_struct *tsk,
					   cputime_t cputime)
279
{
280
	struct thread_group_cputimer *cputimer = &tsk->signal->cputimer;
281

282 283 284 285 286 287 288
	if (!cputimer->running)
		return;

	spin_lock(&cputimer->lock);
	cputimer->cputime.utime =
		cputime_add(cputimer->cputime.utime, cputime);
	spin_unlock(&cputimer->lock);
289 290 291
}

/**
292
 * account_group_system_time - Maintain stime for a thread group.
293
 *
294 295 296
 * @tsk:	Pointer to task structure.
 * @cputime:	Time value by which to increment the stime field of the
 *		thread_group_cputime structure.
297 298 299 300
 *
 * If thread group time is being maintained, get the structure for the
 * running CPU and update the stime field there.
 */
301 302
static inline void account_group_system_time(struct task_struct *tsk,
					     cputime_t cputime)
303
{
304
	struct thread_group_cputimer *cputimer = &tsk->signal->cputimer;
305 306 307

	if (!cputimer->running)
		return;
308

309 310 311 312
	spin_lock(&cputimer->lock);
	cputimer->cputime.stime =
		cputime_add(cputimer->cputime.stime, cputime);
	spin_unlock(&cputimer->lock);
313 314 315
}

/**
316
 * account_group_exec_runtime - Maintain exec runtime for a thread group.
317
 *
318
 * @tsk:	Pointer to task structure.
319
 * @ns:		Time value by which to increment the sum_exec_runtime field
320
 *		of the thread_group_cputime structure.
321 322 323 324
 *
 * If thread group time is being maintained, get the structure for the
 * running CPU and update the sum_exec_runtime field there.
 */
325 326
static inline void account_group_exec_runtime(struct task_struct *tsk,
					      unsigned long long ns)
327
{
328
	struct thread_group_cputimer *cputimer = &tsk->signal->cputimer;
329 330 331

	if (!cputimer->running)
		return;
332

333 334 335
	spin_lock(&cputimer->lock);
	cputimer->cputime.sum_exec_runtime += ns;
	spin_unlock(&cputimer->lock);
336
}