1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Timer events oriented CPU idle governor 4 * 5 * Copyright (C) 2018 Intel Corporation 6 * Author: Rafael J. Wysocki <rafael.j.wysocki@intel.com> 7 * 8 * The idea of this governor is based on the observation that on many systems 9 * timer events are two or more orders of magnitude more frequent than any 10 * other interrupts, so they are likely to be the most significant source of CPU 11 * wakeups from idle states. Moreover, information about what happened in the 12 * (relatively recent) past can be used to estimate whether or not the deepest 13 * idle state with target residency within the time to the closest timer is 14 * likely to be suitable for the upcoming idle time of the CPU and, if not, then 15 * which of the shallower idle states to choose. 16 * 17 * Of course, non-timer wakeup sources are more important in some use cases and 18 * they can be covered by taking a few most recent idle time intervals of the 19 * CPU into account. However, even in that case it is not necessary to consider 20 * idle duration values greater than the time till the closest timer, as the 21 * patterns that they may belong to produce average values close enough to 22 * the time till the closest timer (sleep length) anyway. 23 * 24 * Thus this governor estimates whether or not the upcoming idle time of the CPU 25 * is likely to be significantly shorter than the sleep length and selects an 26 * idle state for it in accordance with that, as follows: 27 * 28 * - Find an idle state on the basis of the sleep length and state statistics 29 * collected over time: 30 * 31 * o Find the deepest idle state whose target residency is less than or equal 32 * to the sleep length. 33 * 34 * o Select it if it matched both the sleep length and the observed idle 35 * duration in the past more often than it matched the sleep length alone 36 * (i.e. the observed idle duration was significantly shorter than the sleep 37 * length matched by it). 38 * 39 * o Otherwise, select the shallower state with the greatest matched "early" 40 * wakeups metric. 41 * 42 * - If the majority of the most recent idle duration values are below the 43 * target residency of the idle state selected so far, use those values to 44 * compute the new expected idle duration and find an idle state matching it 45 * (which has to be shallower than the one selected so far). 46 */ 47 48 #include <linux/cpuidle.h> 49 #include <linux/jiffies.h> 50 #include <linux/kernel.h> 51 #include <linux/sched/clock.h> 52 #include <linux/tick.h> 53 54 /* 55 * The PULSE value is added to metrics when they grow and the DECAY_SHIFT value 56 * is used for decreasing metrics on a regular basis. 57 */ 58 #define PULSE 1024 59 #define DECAY_SHIFT 3 60 61 /* 62 * Number of the most recent idle duration values to take into consideration for 63 * the detection of wakeup patterns. 64 */ 65 #define INTERVALS 8 66 67 /** 68 * struct teo_idle_state - Idle state data used by the TEO cpuidle governor. 69 * @early_hits: "Early" CPU wakeups "matching" this state. 70 * @hits: "On time" CPU wakeups "matching" this state. 71 * @misses: CPU wakeups "missing" this state. 72 * 73 * A CPU wakeup is "matched" by a given idle state if the idle duration measured 74 * after the wakeup is between the target residency of that state and the target 75 * residency of the next one (or if this is the deepest available idle state, it 76 * "matches" a CPU wakeup when the measured idle duration is at least equal to 77 * its target residency). 78 * 79 * Also, from the TEO governor perspective, a CPU wakeup from idle is "early" if 80 * it occurs significantly earlier than the closest expected timer event (that 81 * is, early enough to match an idle state shallower than the one matching the 82 * time till the closest timer event). Otherwise, the wakeup is "on time", or 83 * it is a "hit". 84 * 85 * A "miss" occurs when the given state doesn't match the wakeup, but it matches 86 * the time till the closest timer event used for idle state selection. 87 */ 88 struct teo_idle_state { 89 unsigned int early_hits; 90 unsigned int hits; 91 unsigned int misses; 92 }; 93 94 /** 95 * struct teo_cpu - CPU data used by the TEO cpuidle governor. 96 * @time_span_ns: Time between idle state selection and post-wakeup update. 97 * @sleep_length_ns: Time till the closest timer event (at the selection time). 98 * @states: Idle states data corresponding to this CPU. 99 * @interval_idx: Index of the most recent saved idle interval. 100 * @intervals: Saved idle duration values. 101 */ 102 struct teo_cpu { 103 u64 time_span_ns; 104 u64 sleep_length_ns; 105 struct teo_idle_state states[CPUIDLE_STATE_MAX]; 106 int interval_idx; 107 u64 intervals[INTERVALS]; 108 }; 109 110 static DEFINE_PER_CPU(struct teo_cpu, teo_cpus); 111 112 /** 113 * teo_update - Update CPU data after wakeup. 114 * @drv: cpuidle driver containing state data. 115 * @dev: Target CPU. 116 */ 117 static void teo_update(struct cpuidle_driver *drv, struct cpuidle_device *dev) 118 { 119 struct teo_cpu *cpu_data = per_cpu_ptr(&teo_cpus, dev->cpu); 120 int i, idx_hit = -1, idx_timer = -1; 121 u64 measured_ns; 122 123 if (cpu_data->time_span_ns >= cpu_data->sleep_length_ns) { 124 /* 125 * One of the safety nets has triggered or the wakeup was close 126 * enough to the closest timer event expected at the idle state 127 * selection time to be discarded. 128 */ 129 measured_ns = U64_MAX; 130 } else { 131 u64 lat_ns = drv->states[dev->last_state_idx].exit_latency_ns; 132 133 /* 134 * The computations below are to determine whether or not the 135 * (saved) time till the next timer event and the measured idle 136 * duration fall into the same "bin", so use last_residency_ns 137 * for that instead of time_span_ns which includes the cpuidle 138 * overhead. 139 */ 140 measured_ns = dev->last_residency_ns; 141 /* 142 * The delay between the wakeup and the first instruction 143 * executed by the CPU is not likely to be worst-case every 144 * time, so take 1/2 of the exit latency as a very rough 145 * approximation of the average of it. 146 */ 147 if (measured_ns >= lat_ns) 148 measured_ns -= lat_ns / 2; 149 else 150 measured_ns /= 2; 151 } 152 153 /* 154 * Decay the "early hits" metric for all of the states and find the 155 * states matching the sleep length and the measured idle duration. 156 */ 157 for (i = 0; i < drv->state_count; i++) { 158 unsigned int early_hits = cpu_data->states[i].early_hits; 159 160 cpu_data->states[i].early_hits -= early_hits >> DECAY_SHIFT; 161 162 if (drv->states[i].target_residency_ns <= cpu_data->sleep_length_ns) { 163 idx_timer = i; 164 if (drv->states[i].target_residency_ns <= measured_ns) 165 idx_hit = i; 166 } 167 } 168 169 /* 170 * Update the "hits" and "misses" data for the state matching the sleep 171 * length. If it matches the measured idle duration too, this is a hit, 172 * so increase the "hits" metric for it then. Otherwise, this is a 173 * miss, so increase the "misses" metric for it. In the latter case 174 * also increase the "early hits" metric for the state that actually 175 * matches the measured idle duration. 176 */ 177 if (idx_timer >= 0) { 178 unsigned int hits = cpu_data->states[idx_timer].hits; 179 unsigned int misses = cpu_data->states[idx_timer].misses; 180 181 hits -= hits >> DECAY_SHIFT; 182 misses -= misses >> DECAY_SHIFT; 183 184 if (idx_timer > idx_hit) { 185 misses += PULSE; 186 if (idx_hit >= 0) 187 cpu_data->states[idx_hit].early_hits += PULSE; 188 } else { 189 hits += PULSE; 190 } 191 192 cpu_data->states[idx_timer].misses = misses; 193 cpu_data->states[idx_timer].hits = hits; 194 } 195 196 /* 197 * Save idle duration values corresponding to non-timer wakeups for 198 * pattern detection. 199 */ 200 cpu_data->intervals[cpu_data->interval_idx++] = measured_ns; 201 if (cpu_data->interval_idx >= INTERVALS) 202 cpu_data->interval_idx = 0; 203 } 204 205 static bool teo_time_ok(u64 interval_ns) 206 { 207 return !tick_nohz_tick_stopped() || interval_ns >= TICK_NSEC; 208 } 209 210 /** 211 * teo_find_shallower_state - Find shallower idle state matching given duration. 212 * @drv: cpuidle driver containing state data. 213 * @dev: Target CPU. 214 * @state_idx: Index of the capping idle state. 215 * @duration_ns: Idle duration value to match. 216 */ 217 static int teo_find_shallower_state(struct cpuidle_driver *drv, 218 struct cpuidle_device *dev, int state_idx, 219 u64 duration_ns) 220 { 221 int i; 222 223 for (i = state_idx - 1; i >= 0; i--) { 224 if (dev->states_usage[i].disable) 225 continue; 226 227 state_idx = i; 228 if (drv->states[i].target_residency_ns <= duration_ns) 229 break; 230 } 231 return state_idx; 232 } 233 234 /** 235 * teo_select - Selects the next idle state to enter. 236 * @drv: cpuidle driver containing state data. 237 * @dev: Target CPU. 238 * @stop_tick: Indication on whether or not to stop the scheduler tick. 239 */ 240 static int teo_select(struct cpuidle_driver *drv, struct cpuidle_device *dev, 241 bool *stop_tick) 242 { 243 struct teo_cpu *cpu_data = per_cpu_ptr(&teo_cpus, dev->cpu); 244 s64 latency_req = cpuidle_governor_latency_req(dev->cpu); 245 u64 duration_ns; 246 unsigned int hits, misses, early_hits; 247 int max_early_idx, prev_max_early_idx, constraint_idx, idx, i; 248 ktime_t delta_tick; 249 250 if (dev->last_state_idx >= 0) { 251 teo_update(drv, dev); 252 dev->last_state_idx = -1; 253 } 254 255 cpu_data->time_span_ns = local_clock(); 256 257 duration_ns = tick_nohz_get_sleep_length(&delta_tick); 258 cpu_data->sleep_length_ns = duration_ns; 259 260 hits = 0; 261 misses = 0; 262 early_hits = 0; 263 max_early_idx = -1; 264 prev_max_early_idx = -1; 265 constraint_idx = drv->state_count; 266 idx = -1; 267 268 for (i = 0; i < drv->state_count; i++) { 269 struct cpuidle_state *s = &drv->states[i]; 270 271 if (dev->states_usage[i].disable) { 272 /* 273 * Ignore disabled states with target residencies beyond 274 * the anticipated idle duration. 275 */ 276 if (s->target_residency_ns > duration_ns) 277 continue; 278 279 /* 280 * This state is disabled, so the range of idle duration 281 * values corresponding to it is covered by the current 282 * candidate state, but still the "hits" and "misses" 283 * metrics of the disabled state need to be used to 284 * decide whether or not the state covering the range in 285 * question is good enough. 286 */ 287 hits = cpu_data->states[i].hits; 288 misses = cpu_data->states[i].misses; 289 290 if (early_hits >= cpu_data->states[i].early_hits || 291 idx < 0) 292 continue; 293 294 /* 295 * If the current candidate state has been the one with 296 * the maximum "early hits" metric so far, the "early 297 * hits" metric of the disabled state replaces the 298 * current "early hits" count to avoid selecting a 299 * deeper state with lower "early hits" metric. 300 */ 301 if (max_early_idx == idx) { 302 early_hits = cpu_data->states[i].early_hits; 303 continue; 304 } 305 306 /* 307 * The current candidate state is closer to the disabled 308 * one than the current maximum "early hits" state, so 309 * replace the latter with it, but in case the maximum 310 * "early hits" state index has not been set so far, 311 * check if the current candidate state is not too 312 * shallow for that role. 313 */ 314 if (teo_time_ok(drv->states[idx].target_residency_ns)) { 315 prev_max_early_idx = max_early_idx; 316 early_hits = cpu_data->states[i].early_hits; 317 max_early_idx = idx; 318 } 319 320 continue; 321 } 322 323 if (idx < 0) { 324 idx = i; /* first enabled state */ 325 hits = cpu_data->states[i].hits; 326 misses = cpu_data->states[i].misses; 327 } 328 329 if (s->target_residency_ns > duration_ns) 330 break; 331 332 if (s->exit_latency_ns > latency_req && constraint_idx > i) 333 constraint_idx = i; 334 335 idx = i; 336 hits = cpu_data->states[i].hits; 337 misses = cpu_data->states[i].misses; 338 339 if (early_hits < cpu_data->states[i].early_hits && 340 teo_time_ok(drv->states[i].target_residency_ns)) { 341 prev_max_early_idx = max_early_idx; 342 early_hits = cpu_data->states[i].early_hits; 343 max_early_idx = i; 344 } 345 } 346 347 /* 348 * If the "hits" metric of the idle state matching the sleep length is 349 * greater than its "misses" metric, that is the one to use. Otherwise, 350 * it is more likely that one of the shallower states will match the 351 * idle duration observed after wakeup, so take the one with the maximum 352 * "early hits" metric, but if that cannot be determined, just use the 353 * state selected so far. 354 */ 355 if (hits <= misses) { 356 /* 357 * The current candidate state is not suitable, so take the one 358 * whose "early hits" metric is the maximum for the range of 359 * shallower states. 360 */ 361 if (idx == max_early_idx) 362 max_early_idx = prev_max_early_idx; 363 364 if (max_early_idx >= 0) { 365 idx = max_early_idx; 366 duration_ns = drv->states[idx].target_residency_ns; 367 } 368 } 369 370 /* 371 * If there is a latency constraint, it may be necessary to use a 372 * shallower idle state than the one selected so far. 373 */ 374 if (constraint_idx < idx) 375 idx = constraint_idx; 376 377 if (idx < 0) { 378 idx = 0; /* No states enabled. Must use 0. */ 379 } else if (idx > 0) { 380 unsigned int count = 0; 381 u64 sum = 0; 382 383 /* 384 * Count and sum the most recent idle duration values less than 385 * the current expected idle duration value. 386 */ 387 for (i = 0; i < INTERVALS; i++) { 388 u64 val = cpu_data->intervals[i]; 389 390 if (val >= duration_ns) 391 continue; 392 393 count++; 394 sum += val; 395 } 396 397 /* 398 * Give up unless the majority of the most recent idle duration 399 * values are in the interesting range. 400 */ 401 if (count > INTERVALS / 2) { 402 u64 avg_ns = div64_u64(sum, count); 403 404 /* 405 * Avoid spending too much time in an idle state that 406 * would be too shallow. 407 */ 408 if (teo_time_ok(avg_ns)) { 409 duration_ns = avg_ns; 410 if (drv->states[idx].target_residency_ns > avg_ns) 411 idx = teo_find_shallower_state(drv, dev, 412 idx, avg_ns); 413 } 414 } 415 } 416 417 /* 418 * Don't stop the tick if the selected state is a polling one or if the 419 * expected idle duration is shorter than the tick period length. 420 */ 421 if (((drv->states[idx].flags & CPUIDLE_FLAG_POLLING) || 422 duration_ns < TICK_NSEC) && !tick_nohz_tick_stopped()) { 423 *stop_tick = false; 424 425 /* 426 * The tick is not going to be stopped, so if the target 427 * residency of the state to be returned is not within the time 428 * till the closest timer including the tick, try to correct 429 * that. 430 */ 431 if (idx > 0 && drv->states[idx].target_residency_ns > delta_tick) 432 idx = teo_find_shallower_state(drv, dev, idx, delta_tick); 433 } 434 435 return idx; 436 } 437 438 /** 439 * teo_reflect - Note that governor data for the CPU need to be updated. 440 * @dev: Target CPU. 441 * @state: Entered state. 442 */ 443 static void teo_reflect(struct cpuidle_device *dev, int state) 444 { 445 struct teo_cpu *cpu_data = per_cpu_ptr(&teo_cpus, dev->cpu); 446 447 dev->last_state_idx = state; 448 /* 449 * If the wakeup was not "natural", but triggered by one of the safety 450 * nets, assume that the CPU might have been idle for the entire sleep 451 * length time. 452 */ 453 if (dev->poll_time_limit || 454 (tick_nohz_idle_got_tick() && cpu_data->sleep_length_ns > TICK_NSEC)) { 455 dev->poll_time_limit = false; 456 cpu_data->time_span_ns = cpu_data->sleep_length_ns; 457 } else { 458 cpu_data->time_span_ns = local_clock() - cpu_data->time_span_ns; 459 } 460 } 461 462 /** 463 * teo_enable_device - Initialize the governor's data for the target CPU. 464 * @drv: cpuidle driver (not used). 465 * @dev: Target CPU. 466 */ 467 static int teo_enable_device(struct cpuidle_driver *drv, 468 struct cpuidle_device *dev) 469 { 470 struct teo_cpu *cpu_data = per_cpu_ptr(&teo_cpus, dev->cpu); 471 int i; 472 473 memset(cpu_data, 0, sizeof(*cpu_data)); 474 475 for (i = 0; i < INTERVALS; i++) 476 cpu_data->intervals[i] = U64_MAX; 477 478 return 0; 479 } 480 481 static struct cpuidle_governor teo_governor = { 482 .name = "teo", 483 .rating = 19, 484 .enable = teo_enable_device, 485 .select = teo_select, 486 .reflect = teo_reflect, 487 }; 488 489 static int __init teo_governor_init(void) 490 { 491 return cpuidle_register_governor(&teo_governor); 492 } 493 494 postcore_initcall(teo_governor_init); 495