Line data Source code
1 : // SPDX-License-Identifier: GPL-2.0
2 : /*
3 : * kernel/sched/loadavg.c
4 : *
5 : * This file contains the magic bits required to compute the global loadavg
6 : * figure. Its a silly number but people think its important. We go through
7 : * great pains to make it work on big machines and tickless kernels.
8 : */
9 : #include "sched.h"
10 :
11 : /*
12 : * Global load-average calculations
13 : *
14 : * We take a distributed and async approach to calculating the global load-avg
15 : * in order to minimize overhead.
16 : *
17 : * The global load average is an exponentially decaying average of nr_running +
18 : * nr_uninterruptible.
19 : *
20 : * Once every LOAD_FREQ:
21 : *
22 : * nr_active = 0;
23 : * for_each_possible_cpu(cpu)
24 : * nr_active += cpu_of(cpu)->nr_running + cpu_of(cpu)->nr_uninterruptible;
25 : *
26 : * avenrun[n] = avenrun[0] * exp_n + nr_active * (1 - exp_n)
27 : *
28 : * Due to a number of reasons the above turns in the mess below:
29 : *
30 : * - for_each_possible_cpu() is prohibitively expensive on machines with
31 : * serious number of CPUs, therefore we need to take a distributed approach
32 : * to calculating nr_active.
33 : *
34 : * \Sum_i x_i(t) = \Sum_i x_i(t) - x_i(t_0) | x_i(t_0) := 0
35 : * = \Sum_i { \Sum_j=1 x_i(t_j) - x_i(t_j-1) }
36 : *
37 : * So assuming nr_active := 0 when we start out -- true per definition, we
38 : * can simply take per-CPU deltas and fold those into a global accumulate
39 : * to obtain the same result. See calc_load_fold_active().
40 : *
41 : * Furthermore, in order to avoid synchronizing all per-CPU delta folding
42 : * across the machine, we assume 10 ticks is sufficient time for every
43 : * CPU to have completed this task.
44 : *
45 : * This places an upper-bound on the IRQ-off latency of the machine. Then
46 : * again, being late doesn't loose the delta, just wrecks the sample.
47 : *
48 : * - cpu_rq()->nr_uninterruptible isn't accurately tracked per-CPU because
49 : * this would add another cross-CPU cacheline miss and atomic operation
50 : * to the wakeup path. Instead we increment on whatever CPU the task ran
51 : * when it went into uninterruptible state and decrement on whatever CPU
52 : * did the wakeup. This means that only the sum of nr_uninterruptible over
53 : * all CPUs yields the correct result.
54 : *
55 : * This covers the NO_HZ=n code, for extra head-aches, see the comment below.
56 : */
57 :
58 : /* Variables and functions for calc_load */
59 : atomic_long_t calc_load_tasks;
60 : unsigned long calc_load_update;
61 : unsigned long avenrun[3];
62 : EXPORT_SYMBOL(avenrun); /* should be removed */
63 :
64 : /**
65 : * get_avenrun - get the load average array
66 : * @loads: pointer to dest load array
67 : * @offset: offset to add
68 : * @shift: shift count to shift the result left
69 : *
70 : * These values are estimates at best, so no need for locking.
71 : */
72 11 : void get_avenrun(unsigned long *loads, unsigned long offset, int shift)
73 : {
74 11 : loads[0] = (avenrun[0] + offset) << shift;
75 11 : loads[1] = (avenrun[1] + offset) << shift;
76 11 : loads[2] = (avenrun[2] + offset) << shift;
77 11 : }
78 :
79 772 : long calc_load_fold_active(struct rq *this_rq, long adjust)
80 : {
81 772 : long nr_active, delta = 0;
82 :
83 772 : nr_active = this_rq->nr_running - adjust;
84 772 : nr_active += (long)this_rq->nr_uninterruptible;
85 :
86 0 : if (nr_active != this_rq->calc_load_active) {
87 39 : delta = nr_active - this_rq->calc_load_active;
88 0 : this_rq->calc_load_active = nr_active;
89 : }
90 :
91 39 : return delta;
92 : }
93 :
94 : /**
95 : * fixed_power_int - compute: x^n, in O(log n) time
96 : *
97 : * @x: base of the power
98 : * @frac_bits: fractional bits of @x
99 : * @n: power to raise @x to.
100 : *
101 : * By exploiting the relation between the definition of the natural power
102 : * function: x^n := x*x*...*x (x multiplied by itself for n times), and
103 : * the binary encoding of numbers used by computers: n := \Sum n_i * 2^i,
104 : * (where: n_i \elem {0, 1}, the binary vector representing n),
105 : * we find: x^n := x^(\Sum n_i * 2^i) := \Prod x^(n_i * 2^i), which is
106 : * of course trivially computable in O(log_2 n), the length of our binary
107 : * vector.
108 : */
109 : static unsigned long
110 0 : fixed_power_int(unsigned long x, unsigned int frac_bits, unsigned int n)
111 : {
112 0 : unsigned long result = 1UL << frac_bits;
113 :
114 0 : if (n) {
115 0 : for (;;) {
116 0 : if (n & 1) {
117 0 : result *= x;
118 0 : result += 1UL << (frac_bits - 1);
119 0 : result >>= frac_bits;
120 : }
121 0 : n >>= 1;
122 0 : if (!n)
123 : break;
124 0 : x *= x;
125 0 : x += 1UL << (frac_bits - 1);
126 0 : x >>= frac_bits;
127 : }
128 : }
129 :
130 0 : return result;
131 : }
132 :
133 : /*
134 : * a1 = a0 * e + a * (1 - e)
135 : *
136 : * a2 = a1 * e + a * (1 - e)
137 : * = (a0 * e + a * (1 - e)) * e + a * (1 - e)
138 : * = a0 * e^2 + a * (1 - e) * (1 + e)
139 : *
140 : * a3 = a2 * e + a * (1 - e)
141 : * = (a0 * e^2 + a * (1 - e) * (1 + e)) * e + a * (1 - e)
142 : * = a0 * e^3 + a * (1 - e) * (1 + e + e^2)
143 : *
144 : * ...
145 : *
146 : * an = a0 * e^n + a * (1 - e) * (1 + e + ... + e^n-1) [1]
147 : * = a0 * e^n + a * (1 - e) * (1 - e^n)/(1 - e)
148 : * = a0 * e^n + a * (1 - e^n)
149 : *
150 : * [1] application of the geometric series:
151 : *
152 : * n 1 - x^(n+1)
153 : * S_n := \Sum x^i = -------------
154 : * i=0 1 - x
155 : */
156 : unsigned long
157 0 : calc_load_n(unsigned long load, unsigned long exp,
158 : unsigned long active, unsigned int n)
159 : {
160 0 : return calc_load(load, fixed_power_int(exp, FSHIFT, n), active);
161 : }
162 :
163 : #ifdef CONFIG_NO_HZ_COMMON
164 : /*
165 : * Handle NO_HZ for the global load-average.
166 : *
167 : * Since the above described distributed algorithm to compute the global
168 : * load-average relies on per-CPU sampling from the tick, it is affected by
169 : * NO_HZ.
170 : *
171 : * The basic idea is to fold the nr_active delta into a global NO_HZ-delta upon
172 : * entering NO_HZ state such that we can include this as an 'extra' CPU delta
173 : * when we read the global state.
174 : *
175 : * Obviously reality has to ruin such a delightfully simple scheme:
176 : *
177 : * - When we go NO_HZ idle during the window, we can negate our sample
178 : * contribution, causing under-accounting.
179 : *
180 : * We avoid this by keeping two NO_HZ-delta counters and flipping them
181 : * when the window starts, thus separating old and new NO_HZ load.
182 : *
183 : * The only trick is the slight shift in index flip for read vs write.
184 : *
185 : * 0s 5s 10s 15s
186 : * +10 +10 +10 +10
187 : * |-|-----------|-|-----------|-|-----------|-|
188 : * r:0 0 1 1 0 0 1 1 0
189 : * w:0 1 1 0 0 1 1 0 0
190 : *
191 : * This ensures we'll fold the old NO_HZ contribution in this window while
192 : * accumlating the new one.
193 : *
194 : * - When we wake up from NO_HZ during the window, we push up our
195 : * contribution, since we effectively move our sample point to a known
196 : * busy state.
197 : *
198 : * This is solved by pushing the window forward, and thus skipping the
199 : * sample, for this CPU (effectively using the NO_HZ-delta for this CPU which
200 : * was in effect at the time the window opened). This also solves the issue
201 : * of having to deal with a CPU having been in NO_HZ for multiple LOAD_FREQ
202 : * intervals.
203 : *
204 : * When making the ILB scale, we should try to pull this in as well.
205 : */
206 : static atomic_long_t calc_load_nohz[2];
207 : static int calc_load_idx;
208 :
209 30 : static inline int calc_load_write_idx(void)
210 : {
211 30 : int idx = calc_load_idx;
212 :
213 : /*
214 : * See calc_global_nohz(), if we observe the new index, we also
215 : * need to observe the new update time.
216 : */
217 30 : smp_rmb();
218 :
219 : /*
220 : * If the folding window started, make sure we start writing in the
221 : * next NO_HZ-delta.
222 : */
223 30 : if (!time_before(jiffies, READ_ONCE(calc_load_update)))
224 0 : idx++;
225 :
226 30 : return idx & 1;
227 : }
228 :
229 6 : static inline int calc_load_read_idx(void)
230 : {
231 6 : return calc_load_idx & 1;
232 : }
233 :
234 751 : static void calc_load_nohz_fold(struct rq *rq)
235 : {
236 751 : long delta;
237 :
238 751 : delta = calc_load_fold_active(rq, 0);
239 30 : if (delta) {
240 30 : int idx = calc_load_write_idx();
241 :
242 30 : atomic_long_add(delta, &calc_load_nohz[idx]);
243 : }
244 751 : }
245 :
246 751 : void calc_load_nohz_start(void)
247 : {
248 : /*
249 : * We're going into NO_HZ mode, if there's any pending delta, fold it
250 : * into the pending NO_HZ delta.
251 : */
252 751 : calc_load_nohz_fold(this_rq());
253 751 : }
254 :
255 : /*
256 : * Keep track of the load for NOHZ_FULL, must be called between
257 : * calc_load_nohz_{start,stop}().
258 : */
259 0 : void calc_load_nohz_remote(struct rq *rq)
260 : {
261 0 : calc_load_nohz_fold(rq);
262 0 : }
263 :
264 750 : void calc_load_nohz_stop(void)
265 : {
266 750 : struct rq *this_rq = this_rq();
267 :
268 : /*
269 : * If we're still before the pending sample window, we're done.
270 : */
271 750 : this_rq->calc_load_update = READ_ONCE(calc_load_update);
272 750 : if (time_before(jiffies, this_rq->calc_load_update))
273 : return;
274 :
275 : /*
276 : * We woke inside or after the sample window, this means we're already
277 : * accounted through the nohz accounting, so skip the entire deal and
278 : * sync up for the next window.
279 : */
280 4 : if (time_before(jiffies, this_rq->calc_load_update + 10))
281 4 : this_rq->calc_load_update += LOAD_FREQ;
282 : }
283 :
284 6 : static long calc_load_nohz_read(void)
285 : {
286 6 : int idx = calc_load_read_idx();
287 6 : long delta = 0;
288 :
289 6 : if (atomic_long_read(&calc_load_nohz[idx]))
290 3 : delta = atomic_long_xchg(&calc_load_nohz[idx], 0);
291 :
292 6 : return delta;
293 : }
294 :
295 : /*
296 : * NO_HZ can leave us missing all per-CPU ticks calling
297 : * calc_load_fold_active(), but since a NO_HZ CPU folds its delta into
298 : * calc_load_nohz per calc_load_nohz_start(), all we need to do is fold
299 : * in the pending NO_HZ delta if our NO_HZ period crossed a load cycle boundary.
300 : *
301 : * Once we've updated the global active value, we need to apply the exponential
302 : * weights adjusted to the number of cycles missed.
303 : */
304 6 : static void calc_global_nohz(void)
305 : {
306 6 : unsigned long sample_window;
307 6 : long delta, active, n;
308 :
309 6 : sample_window = READ_ONCE(calc_load_update);
310 6 : if (!time_before(jiffies, sample_window + 10)) {
311 : /*
312 : * Catch-up, fold however many we are behind still
313 : */
314 0 : delta = jiffies - sample_window - 10;
315 0 : n = 1 + (delta / LOAD_FREQ);
316 :
317 0 : active = atomic_long_read(&calc_load_tasks);
318 0 : active = active > 0 ? active * FIXED_1 : 0;
319 :
320 0 : avenrun[0] = calc_load_n(avenrun[0], EXP_1, active, n);
321 0 : avenrun[1] = calc_load_n(avenrun[1], EXP_5, active, n);
322 0 : avenrun[2] = calc_load_n(avenrun[2], EXP_15, active, n);
323 :
324 0 : WRITE_ONCE(calc_load_update, sample_window + n * LOAD_FREQ);
325 : }
326 :
327 : /*
328 : * Flip the NO_HZ index...
329 : *
330 : * Make sure we first write the new time then flip the index, so that
331 : * calc_load_write_idx() will see the new time when it reads the new
332 : * index, this avoids a double flip messing things up.
333 : */
334 6 : smp_wmb();
335 6 : calc_load_idx++;
336 6 : }
337 : #else /* !CONFIG_NO_HZ_COMMON */
338 :
339 : static inline long calc_load_nohz_read(void) { return 0; }
340 : static inline void calc_global_nohz(void) { }
341 :
342 : #endif /* CONFIG_NO_HZ_COMMON */
343 :
344 : /*
345 : * calc_load - update the avenrun load estimates 10 ticks after the
346 : * CPUs have updated calc_load_tasks.
347 : *
348 : * Called from the global timer code.
349 : */
350 6997 : void calc_global_load(void)
351 : {
352 6997 : unsigned long sample_window;
353 6997 : long active, delta;
354 :
355 6997 : sample_window = READ_ONCE(calc_load_update);
356 6997 : if (time_before(jiffies, sample_window + 10))
357 : return;
358 :
359 : /*
360 : * Fold the 'old' NO_HZ-delta to include all NO_HZ CPUs.
361 : */
362 6 : delta = calc_load_nohz_read();
363 6 : if (delta)
364 3 : atomic_long_add(delta, &calc_load_tasks);
365 :
366 6 : active = atomic_long_read(&calc_load_tasks);
367 6 : active = active > 0 ? active * FIXED_1 : 0;
368 :
369 6 : avenrun[0] = calc_load(avenrun[0], EXP_1, active);
370 6 : avenrun[1] = calc_load(avenrun[1], EXP_5, active);
371 6 : avenrun[2] = calc_load(avenrun[2], EXP_15, active);
372 :
373 6 : WRITE_ONCE(calc_load_update, sample_window + LOAD_FREQ);
374 :
375 : /*
376 : * In case we went to NO_HZ for multiple LOAD_FREQ intervals
377 : * catch up in bulk.
378 : */
379 6 : calc_global_nohz();
380 : }
381 :
382 : /*
383 : * Called from scheduler_tick() to periodically update this CPU's
384 : * active count.
385 : */
386 24914 : void calc_global_load_tick(struct rq *this_rq)
387 : {
388 24914 : long delta;
389 :
390 24914 : if (time_before(jiffies, this_rq->calc_load_update))
391 : return;
392 :
393 21 : delta = calc_load_fold_active(this_rq, 0);
394 9 : if (delta)
395 9 : atomic_long_add(delta, &calc_load_tasks);
396 :
397 21 : this_rq->calc_load_update += LOAD_FREQ;
398 : }
|
