Blame view
kernel/sched_clock.c
7.72 KB
3e51f33fc sched: add option... |
1 2 3 4 5 |
/* * sched_clock for unstable cpu clocks * * Copyright (C) 2008 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com> * |
c300ba252 sched_clock: and ... |
6 7 8 |
* Updates and enhancements: * Copyright (C) 2008 Red Hat, Inc. Steven Rostedt <srostedt@redhat.com> * |
3e51f33fc sched: add option... |
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 |
* Based on code by: * Ingo Molnar <mingo@redhat.com> * Guillaume Chazarain <guichaz@gmail.com> * * Create a semi stable clock from a mixture of other events, including: * - gtod * - jiffies * - sched_clock() * - explicit idle events * * We use gtod as base and the unstable clock deltas. The deltas are filtered, * making it monotonic and keeping it within an expected window. This window * is set up using jiffies. * * Furthermore, explicit sleep and wakeup hooks allow us to account for time * that is otherwise invisible (TSC gets stopped). * * The clock: sched_clock_cpu() is monotonic per cpu, and should be somewhat * consistent between cpus (never more than 1 jiffies difference). */ #include <linux/sched.h> #include <linux/percpu.h> #include <linux/spinlock.h> #include <linux/ktime.h> #include <linux/module.h> #ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK |
c300ba252 sched_clock: and ... |
37 38 39 40 |
#define MULTI_SHIFT 15 /* Max is double, Min is 1/2 */ #define MAX_MULTI (2LL << MULTI_SHIFT) #define MIN_MULTI (1LL << (MULTI_SHIFT-1)) |
3e51f33fc sched: add option... |
41 42 43 44 45 46 47 |
struct sched_clock_data { /* * Raw spinlock - this is a special case: this might be called * from within instrumentation code so we dont want to do any * instrumentation ourselves. */ raw_spinlock_t lock; |
62c43dd98 sched_clock: reco... |
48 |
unsigned long tick_jiffies; |
3e51f33fc sched: add option... |
49 50 51 52 |
u64 prev_raw; u64 tick_raw; u64 tick_gtod; u64 clock; |
c300ba252 sched_clock: and ... |
53 |
s64 multi; |
af52a90a1 sched_clock: stop... |
54 55 56 |
#ifdef CONFIG_NO_HZ int check_max; #endif |
3e51f33fc sched: add option... |
57 58 59 60 61 62 63 64 65 66 67 68 69 |
}; static DEFINE_PER_CPU_SHARED_ALIGNED(struct sched_clock_data, sched_clock_data); static inline struct sched_clock_data *this_scd(void) { return &__get_cpu_var(sched_clock_data); } static inline struct sched_clock_data *cpu_sdc(int cpu) { return &per_cpu(sched_clock_data, cpu); } |
a381759d6 sched: fix sched_... |
70 |
static __read_mostly int sched_clock_running; |
3e51f33fc sched: add option... |
71 72 73 |
void sched_clock_init(void) { u64 ktime_now = ktime_to_ns(ktime_get()); |
a381759d6 sched: fix sched_... |
74 |
unsigned long now_jiffies = jiffies; |
3e51f33fc sched: add option... |
75 76 77 78 79 80 |
int cpu; for_each_possible_cpu(cpu) { struct sched_clock_data *scd = cpu_sdc(cpu); scd->lock = (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED; |
62c43dd98 sched_clock: reco... |
81 |
scd->tick_jiffies = now_jiffies; |
a381759d6 sched: fix sched_... |
82 83 |
scd->prev_raw = 0; scd->tick_raw = 0; |
3e51f33fc sched: add option... |
84 85 |
scd->tick_gtod = ktime_now; scd->clock = ktime_now; |
c300ba252 sched_clock: and ... |
86 |
scd->multi = 1 << MULTI_SHIFT; |
af52a90a1 sched_clock: stop... |
87 88 89 |
#ifdef CONFIG_NO_HZ scd->check_max = 1; #endif |
3e51f33fc sched: add option... |
90 |
} |
a381759d6 sched: fix sched_... |
91 92 |
sched_clock_running = 1; |
3e51f33fc sched: add option... |
93 |
} |
af52a90a1 sched_clock: stop... |
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 |
#ifdef CONFIG_NO_HZ /* * The dynamic ticks makes the delta jiffies inaccurate. This * prevents us from checking the maximum time update. * Disable the maximum check during stopped ticks. */ void sched_clock_tick_stop(int cpu) { struct sched_clock_data *scd = cpu_sdc(cpu); scd->check_max = 0; } void sched_clock_tick_start(int cpu) { struct sched_clock_data *scd = cpu_sdc(cpu); scd->check_max = 1; } static int check_max(struct sched_clock_data *scd) { return scd->check_max; } #else static int check_max(struct sched_clock_data *scd) { return 1; } #endif /* CONFIG_NO_HZ */ |
3e51f33fc sched: add option... |
124 125 126 127 128 129 |
/* * update the percpu scd from the raw @now value * * - filter out backward motion * - use jiffies to generate a min,max window to clip the raw values */ |
c0c87734f sched_clock: only... |
130 |
static void __update_sched_clock(struct sched_clock_data *scd, u64 now, u64 *time) |
3e51f33fc sched: add option... |
131 132 |
{ unsigned long now_jiffies = jiffies; |
62c43dd98 sched_clock: reco... |
133 |
long delta_jiffies = now_jiffies - scd->tick_jiffies; |
3e51f33fc sched: add option... |
134 135 136 137 138 |
u64 clock = scd->clock; u64 min_clock, max_clock; s64 delta = now - scd->prev_raw; WARN_ON_ONCE(!irqs_disabled()); |
f7cce27f5 sched_clock: wide... |
139 |
|
c300ba252 sched_clock: and ... |
140 141 142 143 144 145 146 |
/* * At schedule tick the clock can be just under the gtod. We don't * want to push it too prematurely. */ min_clock = scd->tick_gtod + (delta_jiffies * TICK_NSEC); if (min_clock > TICK_NSEC) min_clock -= TICK_NSEC / 2; |
3e51f33fc sched: add option... |
147 148 149 150 151 |
if (unlikely(delta < 0)) { clock++; goto out; } |
f7cce27f5 sched_clock: wide... |
152 153 154 155 156 157 |
/* * The clock must stay within a jiffie of the gtod. * But since we may be at the start of a jiffy or the end of one * we add another jiffy buffer. */ max_clock = scd->tick_gtod + (2 + delta_jiffies) * TICK_NSEC; |
3e51f33fc sched: add option... |
158 |
|
c300ba252 sched_clock: and ... |
159 160 |
delta *= scd->multi; delta >>= MULTI_SHIFT; |
3e51f33fc sched: add option... |
161 |
|
af52a90a1 sched_clock: stop... |
162 |
if (unlikely(clock + delta > max_clock) && check_max(scd)) { |
3e51f33fc sched: add option... |
163 164 165 166 167 168 169 170 171 172 173 |
if (clock < max_clock) clock = max_clock; else clock++; } else { clock += delta; } out: if (unlikely(clock < min_clock)) clock = min_clock; |
c0c87734f sched_clock: only... |
174 175 176 177 178 179 |
if (time) *time = clock; else { scd->prev_raw = now; scd->clock = clock; } |
3e51f33fc sched: add option... |
180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 |
} static void lock_double_clock(struct sched_clock_data *data1, struct sched_clock_data *data2) { if (data1 < data2) { __raw_spin_lock(&data1->lock); __raw_spin_lock(&data2->lock); } else { __raw_spin_lock(&data2->lock); __raw_spin_lock(&data1->lock); } } u64 sched_clock_cpu(int cpu) { struct sched_clock_data *scd = cpu_sdc(cpu); u64 now, clock; |
a381759d6 sched: fix sched_... |
198 199 |
if (unlikely(!sched_clock_running)) return 0ull; |
3e51f33fc sched: add option... |
200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 |
WARN_ON_ONCE(!irqs_disabled()); now = sched_clock(); if (cpu != raw_smp_processor_id()) { /* * in order to update a remote cpu's clock based on our * unstable raw time rebase it against: * tick_raw (offset between raw counters) * tick_gotd (tick offset between cpus) */ struct sched_clock_data *my_scd = this_scd(); lock_double_clock(scd, my_scd); now -= my_scd->tick_raw; now += scd->tick_raw; |
2b8a0cf48 sched_clock: fix ... |
216 217 |
now += my_scd->tick_gtod; now -= scd->tick_gtod; |
3e51f33fc sched: add option... |
218 219 |
__raw_spin_unlock(&my_scd->lock); |
c0c87734f sched_clock: only... |
220 221 222 223 |
__update_sched_clock(scd, now, &clock); __raw_spin_unlock(&scd->lock); |
3e51f33fc sched: add option... |
224 225 |
} else { __raw_spin_lock(&scd->lock); |
c0c87734f sched_clock: only... |
226 227 228 |
__update_sched_clock(scd, now, NULL); clock = scd->clock; __raw_spin_unlock(&scd->lock); |
3e51f33fc sched: add option... |
229 |
} |
3e51f33fc sched: add option... |
230 231 232 233 234 235 |
return clock; } void sched_clock_tick(void) { struct sched_clock_data *scd = this_scd(); |
62c43dd98 sched_clock: reco... |
236 |
unsigned long now_jiffies = jiffies; |
c300ba252 sched_clock: and ... |
237 |
s64 mult, delta_gtod, delta_raw; |
3e51f33fc sched: add option... |
238 |
u64 now, now_gtod; |
a381759d6 sched: fix sched_... |
239 240 |
if (unlikely(!sched_clock_running)) return; |
3e51f33fc sched: add option... |
241 |
WARN_ON_ONCE(!irqs_disabled()); |
3e51f33fc sched: add option... |
242 |
now_gtod = ktime_to_ns(ktime_get()); |
a83bc47c3 sched_clock: reco... |
243 |
now = sched_clock(); |
3e51f33fc sched: add option... |
244 245 |
__raw_spin_lock(&scd->lock); |
c0c87734f sched_clock: only... |
246 |
__update_sched_clock(scd, now, NULL); |
3e51f33fc sched: add option... |
247 248 249 250 251 |
/* * update tick_gtod after __update_sched_clock() because that will * already observe 1 new jiffy; adding a new tick_gtod to that would * increase the clock 2 jiffies. */ |
c300ba252 sched_clock: and ... |
252 253 254 255 256 257 258 259 260 261 262 263 264 |
delta_gtod = now_gtod - scd->tick_gtod; delta_raw = now - scd->tick_raw; if ((long)delta_raw > 0) { mult = delta_gtod << MULTI_SHIFT; do_div(mult, delta_raw); scd->multi = mult; if (scd->multi > MAX_MULTI) scd->multi = MAX_MULTI; else if (scd->multi < MIN_MULTI) scd->multi = MIN_MULTI; } else scd->multi = 1 << MULTI_SHIFT; |
3e51f33fc sched: add option... |
265 266 |
scd->tick_raw = now; scd->tick_gtod = now_gtod; |
c300ba252 sched_clock: and ... |
267 |
scd->tick_jiffies = now_jiffies; |
3e51f33fc sched: add option... |
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 |
__raw_spin_unlock(&scd->lock); } /* * 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) { struct sched_clock_data *scd = this_scd(); u64 now = sched_clock(); /* * Override the previous timestamp and ignore all * sched_clock() deltas that occured while we idled, * and use the PM-provided delta_ns to advance the * rq clock: */ __raw_spin_lock(&scd->lock); scd->prev_raw = now; scd->clock += delta_ns; |
c300ba252 sched_clock: and ... |
297 |
scd->multi = 1 << MULTI_SHIFT; |
3e51f33fc sched: add option... |
298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 |
__raw_spin_unlock(&scd->lock); touch_softlockup_watchdog(); } EXPORT_SYMBOL_GPL(sched_clock_idle_wakeup_event); #endif /* * Scheduler clock - returns current time in nanosec units. * This is default implementation. * Architectures and sub-architectures can override this. */ unsigned long long __attribute__((weak)) sched_clock(void) { return (unsigned long long)jiffies * (NSEC_PER_SEC / HZ); } |
76a2a6ee8 sched: sched_cloc... |
315 316 317 318 319 |
unsigned long long cpu_clock(int cpu) { unsigned long long clock; unsigned long flags; |
2d452c9b1 sched: sched_cloc... |
320 |
local_irq_save(flags); |
76a2a6ee8 sched: sched_cloc... |
321 |
clock = sched_clock_cpu(cpu); |
2d452c9b1 sched: sched_cloc... |
322 |
local_irq_restore(flags); |
76a2a6ee8 sched: sched_cloc... |
323 324 325 |
return clock; } |
4c9fe8ad8 sched: export cpu... |
326 |
EXPORT_SYMBOL_GPL(cpu_clock); |