clock.c 10.2 KB
Newer Older
1 2 3
/*
 * sched_clock for unstable cpu clocks
 *
4
 *  Copyright (C) 2008 Red Hat, Inc., Peter Zijlstra
5
 *
6 7 8
 *  Updates and enhancements:
 *    Copyright (C) 2008 Red Hat, Inc. Steven Rostedt <srostedt@redhat.com>
 *
9 10 11 12
 * Based on code by:
 *   Ingo Molnar <mingo@redhat.com>
 *   Guillaume Chazarain <guichaz@gmail.com>
 *
13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
 *
 * What:
 *
 * cpu_clock(i) provides a fast (execution time) high resolution
 * clock with bounded drift between CPUs. The value of cpu_clock(i)
 * is monotonic for constant i. The timestamp returned is in nanoseconds.
 *
 * ######################### BIG FAT WARNING ##########################
 * # when comparing cpu_clock(i) to cpu_clock(j) for i != j, time can #
 * # go backwards !!                                                  #
 * ####################################################################
 *
 * There is no strict promise about the base, although it tends to start
 * at 0 on boot (but people really shouldn't rely on that).
 *
 * cpu_clock(i)       -- can be used from any context, including NMI.
 * local_clock()      -- is cpu_clock() on the current cpu.
 *
31 32
 * sched_clock_cpu(i)
 *
33 34 35 36 37 38 39 40 41 42 43
 * How:
 *
 * The implementation either uses sched_clock() when
 * !CONFIG_HAVE_UNSTABLE_SCHED_CLOCK, which means in that case the
 * sched_clock() is assumed to provide these properties (mostly it means
 * the architecture provides a globally synchronized highres time source).
 *
 * Otherwise it tries to create a semi stable clock from a mixture of other
 * clocks, including:
 *
 *  - GTOD (clock monotomic)
44 45 46
 *  - sched_clock()
 *  - explicit idle events
 *
47 48 49
 * We use GTOD as base and use sched_clock() deltas to improve resolution. The
 * deltas are filtered to provide monotonicity and keeping it within an
 * expected window.
50 51 52 53 54 55
 *
 * Furthermore, explicit sleep and wakeup hooks allow us to account for time
 * that is otherwise invisible (TSC gets stopped).
 *
 */
#include <linux/spinlock.h>
56
#include <linux/hardirq.h>
57
#include <linux/export.h>
58 59 60
#include <linux/percpu.h>
#include <linux/ktime.h>
#include <linux/sched.h>
61
#include <linux/nmi.h>
62
#include <linux/sched/clock.h>
63
#include <linux/static_key.h>
64
#include <linux/workqueue.h>
65
#include <linux/compiler.h>
66
#include <linux/tick.h>
67

68 69 70 71 72
/*
 * Scheduler clock - returns current time in nanosec units.
 * This is default implementation.
 * Architectures and sub-architectures can override this.
 */
73
unsigned long long __weak sched_clock(void)
74
{
75 76
	return (unsigned long long)(jiffies - INITIAL_JIFFIES)
					* (NSEC_PER_SEC / HZ);
77
}
78
EXPORT_SYMBOL_GPL(sched_clock);
79

80
__read_mostly int sched_clock_running;
81

82 83 84 85 86
void sched_clock_init(void)
{
	sched_clock_running = 1;
}

87
#ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
88 89 90 91 92 93 94
/*
 * We must start with !__sched_clock_stable because the unstable -> stable
 * transition is accurate, while the stable -> unstable transition is not.
 *
 * Similarly we start with __sched_clock_stable_early, thereby assuming we
 * will become stable, such that there's only a single 1 -> 0 transition.
 */
95
static DEFINE_STATIC_KEY_FALSE(__sched_clock_stable);
96
static int __sched_clock_stable_early = 1;
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
/*
 * We want: ktime_get_ns() + gtod_offset == sched_clock() + raw_offset
 */
static __read_mostly u64 raw_offset;
static __read_mostly u64 gtod_offset;

struct sched_clock_data {
	u64			tick_raw;
	u64			tick_gtod;
	u64			clock;
};

static DEFINE_PER_CPU_SHARED_ALIGNED(struct sched_clock_data, sched_clock_data);

static inline struct sched_clock_data *this_scd(void)
{
	return this_cpu_ptr(&sched_clock_data);
}

static inline struct sched_clock_data *cpu_sdc(int cpu)
{
	return &per_cpu(sched_clock_data, cpu);
}

122 123
int sched_clock_stable(void)
{
124
	return static_branch_likely(&__sched_clock_stable);
125 126
}

127
static void __set_sched_clock_stable(void)
128
{
129 130 131 132 133 134 135 136 137 138 139
	struct sched_clock_data *scd = this_scd();

	/*
	 * Attempt to make the (initial) unstable->stable transition continuous.
	 */
	raw_offset = (scd->tick_gtod + gtod_offset) - (scd->tick_raw);

	printk(KERN_INFO "sched_clock: Marking stable (%lld, %lld)->(%lld, %lld)\n",
			scd->tick_gtod, gtod_offset,
			scd->tick_raw,  raw_offset);

140
	static_branch_enable(&__sched_clock_stable);
141
	tick_dep_clear(TICK_DEP_BIT_CLOCK_UNSTABLE);
142 143
}

144
static void __clear_sched_clock_stable(struct work_struct *work)
145
{
146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162
	struct sched_clock_data *scd = this_scd();

	/*
	 * Attempt to make the stable->unstable transition continuous.
	 *
	 * Trouble is, this is typically called from the TSC watchdog
	 * timer, which is late per definition. This means the tick
	 * values can already be screwy.
	 *
	 * Still do what we can.
	 */
	gtod_offset = (scd->tick_raw + raw_offset) - (scd->tick_gtod);

	printk(KERN_INFO "sched_clock: Marking unstable (%lld, %lld)<-(%lld, %lld)\n",
			scd->tick_gtod, gtod_offset,
			scd->tick_raw,  raw_offset);

163
	static_branch_disable(&__sched_clock_stable);
164
	tick_dep_set(TICK_DEP_BIT_CLOCK_UNSTABLE);
165
}
166

167 168 169 170
static DECLARE_WORK(sched_clock_work, __clear_sched_clock_stable);

void clear_sched_clock_stable(void)
{
171 172
	__sched_clock_stable_early = 0;

173
	smp_mb(); /* matches sched_clock_init_late() */
174

175 176
	if (sched_clock_running == 2)
		schedule_work(&sched_clock_work);
177 178
}

179
void sched_clock_init_late(void)
180
{
181
	sched_clock_running = 2;
182 183 184 185 186 187 188 189 190 191 192
	/*
	 * Ensure that it is impossible to not do a static_key update.
	 *
	 * Either {set,clear}_sched_clock_stable() must see sched_clock_running
	 * and do the update, or we must see their __sched_clock_stable_early
	 * and do the update, or both.
	 */
	smp_mb(); /* matches {set,clear}_sched_clock_stable() */

	if (__sched_clock_stable_early)
		__set_sched_clock_stable();
193 194
}

195
/*
196
 * min, max except they take wrapping into account
197 198 199 200 201 202 203 204 205 206 207 208
 */

static inline u64 wrap_min(u64 x, u64 y)
{
	return (s64)(x - y) < 0 ? x : y;
}

static inline u64 wrap_max(u64 x, u64 y)
{
	return (s64)(x - y) > 0 ? x : y;
}

209 210 211 212
/*
 * update the percpu scd from the raw @now value
 *
 *  - filter out backward motion
213
 *  - use the GTOD tick value to create a window to filter crazy TSC values
214
 */
215
static u64 sched_clock_local(struct sched_clock_data *scd)
216
{
217 218
	u64 now, clock, old_clock, min_clock, max_clock;
	s64 delta;
219

220 221 222
again:
	now = sched_clock();
	delta = now - scd->tick_raw;
223 224
	if (unlikely(delta < 0))
		delta = 0;
225

226 227
	old_clock = scd->clock;

228 229
	/*
	 * scd->clock = clamp(scd->tick_gtod + delta,
230 231
	 *		      max(scd->tick_gtod, scd->clock),
	 *		      scd->tick_gtod + TICK_NSEC);
232
	 */
233

234
	clock = scd->tick_gtod + gtod_offset + delta;
235 236
	min_clock = wrap_max(scd->tick_gtod, old_clock);
	max_clock = wrap_max(old_clock, scd->tick_gtod + TICK_NSEC);
237

238 239
	clock = wrap_max(clock, min_clock);
	clock = wrap_min(clock, max_clock);
240

241
	if (cmpxchg64(&scd->clock, old_clock, clock) != old_clock)
242
		goto again;
243

244
	return clock;
245 246
}

247
static u64 sched_clock_remote(struct sched_clock_data *scd)
248
{
249 250 251 252
	struct sched_clock_data *my_scd = this_scd();
	u64 this_clock, remote_clock;
	u64 *ptr, old_val, val;

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
#if BITS_PER_LONG != 64
again:
	/*
	 * Careful here: The local and the remote clock values need to
	 * be read out atomic as we need to compare the values and
	 * then update either the local or the remote side. So the
	 * cmpxchg64 below only protects one readout.
	 *
	 * We must reread via sched_clock_local() in the retry case on
	 * 32bit as an NMI could use sched_clock_local() via the
	 * tracer and hit between the readout of
	 * the low32bit and the high 32bit portion.
	 */
	this_clock = sched_clock_local(my_scd);
	/*
	 * We must enforce atomic readout on 32bit, otherwise the
	 * update on the remote cpu can hit inbetween the readout of
	 * the low32bit and the high 32bit portion.
	 */
	remote_clock = cmpxchg64(&scd->clock, 0, 0);
#else
	/*
	 * On 64bit the read of [my]scd->clock is atomic versus the
	 * update, so we can avoid the above 32bit dance.
	 */
278 279 280 281
	sched_clock_local(my_scd);
again:
	this_clock = my_scd->clock;
	remote_clock = scd->clock;
282
#endif
283 284 285 286 287 288 289 290 291 292 293

	/*
	 * Use the opportunity that we have both locks
	 * taken to couple the two clocks: we take the
	 * larger time as the latest time for both
	 * runqueues. (this creates monotonic movement)
	 */
	if (likely((s64)(remote_clock - this_clock) < 0)) {
		ptr = &scd->clock;
		old_val = remote_clock;
		val = this_clock;
294
	} else {
295 296 297 298 299 300
		/*
		 * Should be rare, but possible:
		 */
		ptr = &my_scd->clock;
		old_val = this_clock;
		val = remote_clock;
301
	}
302

303
	if (cmpxchg64(ptr, old_val, val) != old_val)
304 305 306
		goto again;

	return val;
307 308
}

309 310 311 312 313
/*
 * Similar to cpu_clock(), but requires local IRQs to be disabled.
 *
 * See cpu_clock().
 */
314 315
u64 sched_clock_cpu(int cpu)
{
316
	struct sched_clock_data *scd;
317 318
	u64 clock;

319
	if (sched_clock_stable())
320
		return sched_clock() + raw_offset;
321 322 323 324

	if (unlikely(!sched_clock_running))
		return 0ull;

325
	preempt_disable_notrace();
326
	scd = cpu_sdc(cpu);
327

328 329 330 331
	if (cpu != smp_processor_id())
		clock = sched_clock_remote(scd);
	else
		clock = sched_clock_local(scd);
332
	preempt_enable_notrace();
333

334 335
	return clock;
}
336
EXPORT_SYMBOL_GPL(sched_clock_cpu);
337 338 339

void sched_clock_tick(void)
{
Peter Zijlstra's avatar
Peter Zijlstra committed
340
	struct sched_clock_data *scd;
341

342 343
	WARN_ON_ONCE(!irqs_disabled());

344 345 346 347 348 349 350
	/*
	 * Update these values even if sched_clock_stable(), because it can
	 * become unstable at any point in time at which point we need some
	 * values to fall back on.
	 *
	 * XXX arguably we can skip this if we expose tsc_clocksource_reliable
	 */
Peter Zijlstra's avatar
Peter Zijlstra committed
351
	scd = this_scd();
352 353
	scd->tick_raw  = sched_clock();
	scd->tick_gtod = ktime_get_ns();
354

355 356
	if (!sched_clock_stable() && likely(sched_clock_running))
		sched_clock_local(scd);
357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372
}

/*
 * We are going deep-idle (irqs are disabled):
 */
void sched_clock_idle_sleep_event(void)
{
	sched_clock_cpu(smp_processor_id());
}
EXPORT_SYMBOL_GPL(sched_clock_idle_sleep_event);

/*
 * We just idled delta nanoseconds (called with irqs disabled):
 */
void sched_clock_idle_wakeup_event(u64 delta_ns)
{
373 374 375
	if (timekeeping_suspended)
		return;

376
	sched_clock_tick();
377
	touch_softlockup_watchdog_sched();
378 379 380
}
EXPORT_SYMBOL_GPL(sched_clock_idle_wakeup_event);

Peter Zijlstra's avatar
Peter Zijlstra committed
381 382 383 384 385 386 387 388 389
#else /* CONFIG_HAVE_UNSTABLE_SCHED_CLOCK */

u64 sched_clock_cpu(int cpu)
{
	if (unlikely(!sched_clock_running))
		return 0;

	return sched_clock();
}
390

391
#endif /* CONFIG_HAVE_UNSTABLE_SCHED_CLOCK */
392

393 394 395 396 397 398 399 400 401 402 403 404
/*
 * Running clock - returns the time that has elapsed while a guest has been
 * running.
 * On a guest this value should be local_clock minus the time the guest was
 * suspended by the hypervisor (for any reason).
 * On bare metal this function should return the same as local_clock.
 * Architectures and sub-architectures can override this.
 */
u64 __weak running_clock(void)
{
	return local_clock();
}