Re: [PATCH] sched/loadavg: Avoid loadavg spikes caused by delayed NO_HZ accounting

From: kbuild test robot
Date: Wed Feb 08 2017 - 11:09:09 EST

Next message:
Previous
In reply
Next in
Messages

Hi Matt,

[auto build test ERROR [also build test ERROR [if your patch is applied
url: # save the make ARCH=i386

All error/warnings
kernel/sched/loadavg.c: >> kernel/sched/loadavg.c:404:0: }

>> kernel/sched/loadavg.c:220:2: if (time_in_range_open(jiffies, ^~
>> kernel/sched/loadavg.c:220:2: kernel/sched/loadavg.c:404:0: }

>> kernel/sched/loadavg.c:220:2: if (time_in_range_open(jiffies, ^~
At top level:
kernel/sched/loadavg.c:100:1: calc_load(unsigned ^~~~~~~~~

vim +/if +404 kernel/sched/l
5ee9e5df kernel/sched/loadavg.c 5ee9e5df kernel/sched/loadavg.c 5ee9e5df kernel/sched/loadavg.c 45ceebf7 kernel/sched/proc.c 5ee9e5df kernel/sched/loadavg.c 5ee9e5df kernel/sched/loadavg.c 5ee9e5df kernel/sched/loadavg.c 5ee9e5df kernel/sched/loadavg.c 5ee9e5df kernel/sched/loadavg.c 5ee9e5df kernel/sched/loadavg.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 3289bdb4 kernel/sched/loadavg.c 3289bdb4 kernel/sched/loadavg.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 3289bdb4 kernel/sched/loadavg.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 3289bdb4 kernel/sched/loadavg.c 3289bdb4 kernel/sched/loadavg.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 3289bdb4 kernel/sched/loadavg.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 3289bdb4 kernel/sched/loadavg.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c d60585c5 kernel/sched/loadavg.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c 45ceebf7 kernel/sched/proc.c
:::::: The code at :::::: 45ceebf77653975815d82
:::::: TO: Paul Gortmaker :::::: CC: Ingo Molnar
---
0-DAY kernel test infrastructure Attachment: Description: application/gzip Vineet Gupta: "Re: [PATCH] ARCv2: intc: Disable all core interrupts by default" message: Mark Rutland: "Re: [PATCH] module: Optimize search_module_extables()" to: Matt Fleming: "[PATCH] sched/loadavg: Avoid loadavg spikes caused by delayed NO_HZ accounting" thread: kbuild test robot: "Re: [PATCH] sched/loadavg: Avoid loadavg spikes caused by delayed NO_HZ accounting" sorted by: [ date ] [ thread ] [ subject ] [ author ] on tip/sched/core]
on v4.10-rc7 next-20170208]
to the wrong git tree, please drop us a note to help improve the system]
href="https://github.com/0day-ci/linux/commits/Matt-Fleming/sched-loadavg-Avoid-loadavg-spikes-caused-by-delayed-NO_HZ-accounting/20170208-224135">https://github.com/0day-ci/linux/commits/Matt-Fleming/sched-loadavg-Avoid-loadavg-spikes-caused-by-delayed-NO_HZ-accounting/20170208-224135
(attached as .config)
6.2.0-3) 6.2.0 20160901
attached .config to linux build tree
(new ones prefixed by >>):
In function 'calc_load_exit_idle':
error: unterminated argument list invoking macro "if"
error: expected '(' at end of input
next_window, next_window + 10)
warning: this 'if' clause does not guard... [-Wmisleading-indentation]
note: ...this statement, but the latter is misleadingly indented as if it is guarded by the 'if'
error: expected declaration or statement at end of input
next_window, next_window + 10)
warning: 'calc_load' defined but not used [-Wunused-function]
long load, unsigned long exp, unsigned long active)
oadavg.c
Matt Fleming 2017-02-08 214 *
Matt Fleming 2017-02-08 215 * The next window is 'calc_load_update' if we haven't reached it yet,
Matt Fleming 2017-02-08 216 * and 'calc_load_update + 10' if we're inside the current window.
Paul Gortmaker 2013-04-19 217 */
Matt Fleming 2017-02-08 218 next_window = calc_load_update;
Matt Fleming 2017-02-08 219
Matt Fleming 2017-02-08 @220 if (time_in_range_open(jiffies, next_window, next_window + 10)
Matt Fleming 2017-02-08 221 next_window += LOAD_FREQ;
Matt Fleming 2017-02-08 222
Matt Fleming 2017-02-08 223 this_rq->calc_load_update = next_window;
Paul Gortmaker 2013-04-19 224 }
Paul Gortmaker 2013-04-19 225
Paul Gortmaker 2013-04-19 226 static long calc_load_fold_idle(void)
Paul Gortmaker 2013-04-19 227 {
Paul Gortmaker 2013-04-19 228 int idx = calc_load_read_idx();
Paul Gortmaker 2013-04-19 229 long delta = 0;
Paul Gortmaker 2013-04-19 230
Paul Gortmaker 2013-04-19 231 if (atomic_long_read(&calc_load_idle[idx]))
Paul Gortmaker 2013-04-19 232 delta = atomic_long_xchg(&calc_load_idle[idx], 0);
Paul Gortmaker 2013-04-19 233
Paul Gortmaker 2013-04-19 234 return delta;
Paul Gortmaker 2013-04-19 235 }
Paul Gortmaker 2013-04-19 236
Paul Gortmaker 2013-04-19 237 /**
Paul Gortmaker 2013-04-19 238 * fixed_power_int - compute: x^n, in O(log n) time
Paul Gortmaker 2013-04-19 239 *
Paul Gortmaker 2013-04-19 240 * @x: base of the power
Paul Gortmaker 2013-04-19 241 * @frac_bits: fractional bits of @x
Paul Gortmaker 2013-04-19 242 * @n: power to raise @x to.
Paul Gortmaker 2013-04-19 243 *
Paul Gortmaker 2013-04-19 244 * By exploiting the relation between the definition of the natural power
Paul Gortmaker 2013-04-19 245 * function: x^n := x*x*...*x (x multiplied by itself for n times), and
Paul Gortmaker 2013-04-19 246 * the binary encoding of numbers used by computers: n := \Sum n_i * 2^i,
Paul Gortmaker 2013-04-19 247 * (where: n_i \elem {0, 1}, the binary vector representing n),
Paul Gortmaker 2013-04-19 248 * we find: x^n := x^(\Sum n_i * 2^i) := \Prod x^(n_i * 2^i), which is
Paul Gortmaker 2013-04-19 249 * of course trivially computable in O(log_2 n), the length of our binary
Paul Gortmaker 2013-04-19 250 * vector.
Paul Gortmaker 2013-04-19 251 */
Paul Gortmaker 2013-04-19 252 static unsigned long
Paul Gortmaker 2013-04-19 253 fixed_power_int(unsigned long x, unsigned int frac_bits, unsigned int n)
Paul Gortmaker 2013-04-19 254 {
Paul Gortmaker 2013-04-19 255 unsigned long result = 1UL << frac_bits;
Paul Gortmaker 2013-04-19 256
Peter Zijlstra 2015-04-14 257 if (n) {
Peter Zijlstra 2015-04-14 258 for (;;) {
Paul Gortmaker 2013-04-19 259 if (n & 1) {
Paul Gortmaker 2013-04-19 260 result *= x;
Paul Gortmaker 2013-04-19 261 result += 1UL << (frac_bits - 1);
Paul Gortmaker 2013-04-19 262 result >>= frac_bits;
Paul Gortmaker 2013-04-19 263 }
Paul Gortmaker 2013-04-19 264 n >>= 1;
Paul Gortmaker 2013-04-19 265 if (!n)
Paul Gortmaker 2013-04-19 266 break;
Paul Gortmaker 2013-04-19 267 x *= x;
Paul Gortmaker 2013-04-19 268 x += 1UL << (frac_bits - 1);
Paul Gortmaker 2013-04-19 269 x >>= frac_bits;
Paul Gortmaker 2013-04-19 270 }
Peter Zijlstra 2015-04-14 271 }
Paul Gortmaker 2013-04-19 272
Paul Gortmaker 2013-04-19 273 return result;
Paul Gortmaker 2013-04-19 274 }
Paul Gortmaker 2013-04-19 275
Paul Gortmaker 2013-04-19 276 /*
Paul Gortmaker 2013-04-19 277 * a1 = a0 * e + a * (1 - e)
Paul Gortmaker 2013-04-19 278 *
Paul Gortmaker 2013-04-19 279 * a2 = a1 * e + a * (1 - e)
Paul Gortmaker 2013-04-19 280 * = (a0 * e + a * (1 - e)) * e + a * (1 - e)
Paul Gortmaker 2013-04-19 281 * = a0 * e^2 + a * (1 - e) * (1 + e)
Paul Gortmaker 2013-04-19 282 *
Paul Gortmaker 2013-04-19 283 * a3 = a2 * e + a * (1 - e)
Paul Gortmaker 2013-04-19 284 * = (a0 * e^2 + a * (1 - e) * (1 + e)) * e + a * (1 - e)
Paul Gortmaker 2013-04-19 285 * = a0 * e^3 + a * (1 - e) * (1 + e + e^2)
Paul Gortmaker 2013-04-19 286 *
Paul Gortmaker 2013-04-19 287 * ...
Paul Gortmaker 2013-04-19 288 *
Paul Gortmaker 2013-04-19 289 * an = a0 * e^n + a * (1 - e) * (1 + e + ... + e^n-1) [1]
Paul Gortmaker 2013-04-19 290 * = a0 * e^n + a * (1 - e) * (1 - e^n)/(1 - e)
Paul Gortmaker 2013-04-19 291 * = a0 * e^n + a * (1 - e^n)
Paul Gortmaker 2013-04-19 292 *
Paul Gortmaker 2013-04-19 293 * [1] application of the geometric series:
Paul Gortmaker 2013-04-19 294 *
Paul Gortmaker 2013-04-19 295 * n 1 - x^(n+1)
Paul Gortmaker 2013-04-19 296 * S_n := \Sum x^i = -------------
Paul Gortmaker 2013-04-19 297 * i=0 1 - x
Paul Gortmaker 2013-04-19 298 */
Paul Gortmaker 2013-04-19 299 static unsigned long
Paul Gortmaker 2013-04-19 300 calc_load_n(unsigned long load, unsigned long exp,
Paul Gortmaker 2013-04-19 301 unsigned long active, unsigned int n)
Paul Gortmaker 2013-04-19 302 {
Paul Gortmaker 2013-04-19 303 return calc_load(load, fixed_power_int(exp, FSHIFT, n), active);
Paul Gortmaker 2013-04-19 304 }
Paul Gortmaker 2013-04-19 305
Paul Gortmaker 2013-04-19 306 /*
Paul Gortmaker 2013-04-19 307 * NO_HZ can leave us missing all per-cpu ticks calling
Paul Gortmaker 2013-04-19 308 * calc_load_account_active(), but since an idle CPU folds its delta into
Paul Gortmaker 2013-04-19 309 * calc_load_tasks_idle per calc_load_account_idle(), all we need to do is fold
Paul Gortmaker 2013-04-19 310 * in the pending idle delta if our idle period crossed a load cycle boundary.
Paul Gortmaker 2013-04-19 311 *
Paul Gortmaker 2013-04-19 312 * Once we've updated the global active value, we need to apply the exponential
Paul Gortmaker 2013-04-19 313 * weights adjusted to the number of cycles missed.
Paul Gortmaker 2013-04-19 314 */
Paul Gortmaker 2013-04-19 315 static void calc_global_nohz(void)
Paul Gortmaker 2013-04-19 316 {
Paul Gortmaker 2013-04-19 317 long delta, active, n;
Paul Gortmaker 2013-04-19 318
Paul Gortmaker 2013-04-19 319 if (!time_before(jiffies, calc_load_update + 10)) {
Paul Gortmaker 2013-04-19 320 /*
Paul Gortmaker 2013-04-19 321 * Catch-up, fold however many we are behind still
Paul Gortmaker 2013-04-19 322 */
Paul Gortmaker 2013-04-19 323 delta = jiffies - calc_load_update - 10;
Paul Gortmaker 2013-04-19 324 n = 1 + (delta / LOAD_FREQ);
Paul Gortmaker 2013-04-19 325
Paul Gortmaker 2013-04-19 326 active = atomic_long_read(&calc_load_tasks);
Paul Gortmaker 2013-04-19 327 active = active > 0 ? active * FIXED_1 : 0;
Paul Gortmaker 2013-04-19 328
Paul Gortmaker 2013-04-19 329 avenrun[0] = calc_load_n(avenrun[0], EXP_1, active, n);
Paul Gortmaker 2013-04-19 330 avenrun[1] = calc_load_n(avenrun[1], EXP_5, active, n);
Paul Gortmaker 2013-04-19 331 avenrun[2] = calc_load_n(avenrun[2], EXP_15, active, n);
Paul Gortmaker 2013-04-19 332
Paul Gortmaker 2013-04-19 333 calc_load_update += n * LOAD_FREQ;
Paul Gortmaker 2013-04-19 334 }
Paul Gortmaker 2013-04-19 335
Paul Gortmaker 2013-04-19 336 /*
Paul Gortmaker 2013-04-19 337 * Flip the idle index...
Paul Gortmaker 2013-04-19 338 *
Paul Gortmaker 2013-04-19 339 * Make sure we first write the new time then flip the index, so that
Paul Gortmaker 2013-04-19 340 * calc_load_write_idx() will see the new time when it reads the new
Paul Gortmaker 2013-04-19 341 * index, this avoids a double flip messing things up.
Paul Gortmaker 2013-04-19 342 */
Paul Gortmaker 2013-04-19 343 smp_wmb();
Paul Gortmaker 2013-04-19 344 calc_load_idx++;
Paul Gortmaker 2013-04-19 345 }
Paul Gortmaker 2013-04-19 346 #else /* !CONFIG_NO_HZ_COMMON */
Paul Gortmaker 2013-04-19 347
Paul Gortmaker 2013-04-19 348 static inline long calc_load_fold_idle(void) { return 0; }
Paul Gortmaker 2013-04-19 349 static inline void calc_global_nohz(void) { }
Paul Gortmaker 2013-04-19 350
Paul Gortmaker 2013-04-19 351 #endif /* CONFIG_NO_HZ_COMMON */
Paul Gortmaker 2013-04-19 352
Paul Gortmaker 2013-04-19 353 /*
Paul Gortmaker 2013-04-19 354 * calc_load - update the avenrun load estimates 10 ticks after the
Paul Gortmaker 2013-04-19 355 * CPUs have updated calc_load_tasks.
Peter Zijlstra 2015-04-14 356 *
Peter Zijlstra 2015-04-14 357 * Called from the global timer code.
Paul Gortmaker 2013-04-19 358 */
Paul Gortmaker 2013-04-19 359 void calc_global_load(unsigned long ticks)
Paul Gortmaker 2013-04-19 360 {
Paul Gortmaker 2013-04-19 361 long active, delta;
Paul Gortmaker 2013-04-19 362
Paul Gortmaker 2013-04-19 363 if (time_before(jiffies, calc_load_update + 10))
Paul Gortmaker 2013-04-19 364 return;
Paul Gortmaker 2013-04-19 365
Paul Gortmaker 2013-04-19 366 /*
Paul Gortmaker 2013-04-19 367 * Fold the 'old' idle-delta to include all NO_HZ cpus.
Paul Gortmaker 2013-04-19 368 */
Paul Gortmaker 2013-04-19 369 delta = calc_load_fold_idle();
Paul Gortmaker 2013-04-19 370 if (delta)
Paul Gortmaker 2013-04-19 371 atomic_long_add(delta, &calc_load_tasks);
Paul Gortmaker 2013-04-19 372
Paul Gortmaker 2013-04-19 373 active = atomic_long_read(&calc_load_tasks);
Paul Gortmaker 2013-04-19 374 active = active > 0 ? active * FIXED_1 : 0;
Paul Gortmaker 2013-04-19 375
Paul Gortmaker 2013-04-19 376 avenrun[0] = calc_load(avenrun[0], EXP_1, active);
Paul Gortmaker 2013-04-19 377 avenrun[1] = calc_load(avenrun[1], EXP_5, active);
Paul Gortmaker 2013-04-19 378 avenrun[2] = calc_load(avenrun[2], EXP_15, active);
Paul Gortmaker 2013-04-19 379
Paul Gortmaker 2013-04-19 380 calc_load_update += LOAD_FREQ;
Paul Gortmaker 2013-04-19 381
Paul Gortmaker 2013-04-19 382 /*
Paul Gortmaker 2013-04-19 383 * In case we idled for multiple LOAD_FREQ intervals, catch up in bulk.
Paul Gortmaker 2013-04-19 384 */
Paul Gortmaker 2013-04-19 385 calc_global_nohz();
Paul Gortmaker 2013-04-19 386 }
Paul Gortmaker 2013-04-19 387
Paul Gortmaker 2013-04-19 388 /*
Peter Zijlstra 2015-04-14 389 * Called from scheduler_tick() to periodically update this CPU's
Paul Gortmaker 2013-04-19 390 * active count.
Paul Gortmaker 2013-04-19 391 */
Peter Zijlstra 2015-04-14 392 void calc_global_load_tick(struct rq *this_rq)
Paul Gortmaker 2013-04-19 393 {
Paul Gortmaker 2013-04-19 394 long delta;
Paul Gortmaker 2013-04-19 395
Paul Gortmaker 2013-04-19 396 if (time_before(jiffies, this_rq->calc_load_update))
Paul Gortmaker 2013-04-19 397 return;
Paul Gortmaker 2013-04-19 398
Thomas Gleixner 2016-07-12 399 delta = calc_load_fold_active(this_rq, 0);
Paul Gortmaker 2013-04-19 400 if (delta)
Paul Gortmaker 2013-04-19 401 atomic_long_add(delta, &calc_load_tasks);
Paul Gortmaker 2013-04-19 402
Paul Gortmaker 2013-04-19 403 this_rq->calc_load_update += LOAD_FREQ;
Paul Gortmaker 2013-04-19 @404 }
line 404 was first introduced by commit
fcf7cec0a164215ae11 sched: Factor out load calculation code from sched/core.c --> sched/proc.c
<paul.gortmaker@xxxxxxxxxxxxx>
<mingo@xxxxxxxxxx>
Open Source Technology Center
href="https://lists.01.org/pipermail/kbuild-all">https://lists.01.org/pipermail/kbuild-all Intel Corporation
>.config.gz

Next message: Vineet Gupta: "Re: [PATCH] ARCv2: intc: Disable all core interrupts by default"
Previous message: Mark Rutland: "Re: [PATCH] module: Optimize search_module_extables()"
In reply to: Matt Fleming: "[PATCH] sched/loadavg: Avoid loadavg spikes caused by delayed NO_HZ accounting"
Next in thread: kbuild test robot: "Re: [PATCH] sched/loadavg: Avoid loadavg spikes caused by delayed NO_HZ accounting"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]