1sched(7)               Miscellaneous Information Manual               sched(7)
2
3
4

NAME

6       sched - overview of CPU scheduling
7

DESCRIPTION

9       Since  Linux 2.6.23, the default scheduler is CFS, the "Completely Fair
10       Scheduler".  The CFS scheduler replaced the earlier "O(1)" scheduler.
11
12   API summary
13       Linux provides the following  system  calls  for  controlling  the  CPU
14       scheduling  behavior,  policy, and priority of processes (or, more pre‐
15       cisely, threads).
16
17       nice(2)
18              Set a new nice value for the calling thread, and return the  new
19              nice value.
20
21       getpriority(2)
22              Return  the  nice value of a thread, a process group, or the set
23              of threads owned by a specified user.
24
25       setpriority(2)
26              Set the nice value of a thread, a process group, or the  set  of
27              threads owned by a specified user.
28
29       sched_setscheduler(2)
30              Set the scheduling policy and parameters of a specified thread.
31
32       sched_getscheduler(2)
33              Return the scheduling policy of a specified thread.
34
35       sched_setparam(2)
36              Set the scheduling parameters of a specified thread.
37
38       sched_getparam(2)
39              Fetch the scheduling parameters of a specified thread.
40
41       sched_get_priority_max(2)
42              Return  the maximum priority available in a specified scheduling
43              policy.
44
45       sched_get_priority_min(2)
46              Return the minimum priority available in a specified  scheduling
47              policy.
48
49       sched_rr_get_interval(2)
50              Fetch  the quantum used for threads that are scheduled under the
51              "round-robin" scheduling policy.
52
53       sched_yield(2)
54              Cause the caller to relinquish  the  CPU,  so  that  some  other
55              thread be executed.
56
57       sched_setaffinity(2)
58              (Linux-specific) Set the CPU affinity of a specified thread.
59
60       sched_getaffinity(2)
61              (Linux-specific) Get the CPU affinity of a specified thread.
62
63       sched_setattr(2)
64              Set  the scheduling policy and parameters of a specified thread.
65              This (Linux-specific) system call provides  a  superset  of  the
66              functionality of sched_setscheduler(2) and sched_setparam(2).
67
68       sched_getattr(2)
69              Fetch  the  scheduling  policy  and  parameters  of  a specified
70              thread.  This (Linux-specific) system call provides  a  superset
71              of  the  functionality  of  sched_getscheduler(2) and sched_get‐
72              param(2).
73
74   Scheduling policies
75       The scheduler is the  kernel  component  that  decides  which  runnable
76       thread will be executed by the CPU next.  Each thread has an associated
77       scheduling policy and a  static  scheduling  priority,  sched_priority.
78       The  scheduler makes its decisions based on knowledge of the scheduling
79       policy and static priority of all threads on the system.
80
81       For threads scheduled under  one  of  the  normal  scheduling  policies
82       (SCHED_OTHER,  SCHED_IDLE,  SCHED_BATCH), sched_priority is not used in
83       scheduling decisions (it must be specified as 0).
84
85       Processes scheduled under one of the  real-time  policies  (SCHED_FIFO,
86       SCHED_RR)  have  a  sched_priority  value  in  the  range 1 (low) to 99
87       (high).  (As the numbers imply, real-time threads  always  have  higher
88       priority  than  normal threads.)  Note well: POSIX.1 requires an imple‐
89       mentation to support only a minimum 32 distinct priority levels for the
90       real-time  policies, and some systems supply just this minimum.  Porta‐
91       ble programs should use sched_get_priority_min(2) and  sched_get_prior‐
92       ity_max(2)  to  find the range of priorities supported for a particular
93       policy.
94
95       Conceptually, the scheduler maintains a list of  runnable  threads  for
96       each possible sched_priority value.  In order to determine which thread
97       runs next, the scheduler looks for the nonempty list with  the  highest
98       static priority and selects the thread at the head of this list.
99
100       A  thread's scheduling policy determines where it will be inserted into
101       the list of threads with equal static priority and how it will move in‐
102       side this list.
103
104       All scheduling is preemptive: if a thread with a higher static priority
105       becomes ready to run, the currently running thread  will  be  preempted
106       and  returned  to  the  wait  list  for its static priority level.  The
107       scheduling policy determines the  ordering  only  within  the  list  of
108       runnable threads with equal static priority.
109
110   SCHED_FIFO: First in-first out scheduling
111       SCHED_FIFO can be used only with static priorities higher than 0, which
112       means that when a SCHED_FIFO thread becomes runnable,  it  will  always
113       immediately  preempt any currently running SCHED_OTHER, SCHED_BATCH, or
114       SCHED_IDLE thread.  SCHED_FIFO is a simple scheduling algorithm without
115       time  slicing.   For threads scheduled under the SCHED_FIFO policy, the
116       following rules apply:
117
118       •  A running SCHED_FIFO thread  that  has  been  preempted  by  another
119          thread  of higher priority will stay at the head of the list for its
120          priority and will resume execution as soon as all threads of  higher
121          priority are blocked again.
122
123       •  When  a  blocked  SCHED_FIFO thread becomes runnable, it will be in‐
124          serted at the end of the list for its priority.
125
126       •  If a call  to  sched_setscheduler(2),  sched_setparam(2),  sched_se‐
127          tattr(2),   pthread_setschedparam(3),   or   pthread_setschedprio(3)
128          changes the priority of the running or  runnable  SCHED_FIFO  thread
129          identified  by  pid  the effect on the thread's position in the list
130          depends on the direction of the change to threads priority:
131
132          (a)  If the thread's priority is raised, it is placed at the end  of
133               the  list  for its new priority.  As a consequence, it may pre‐
134               empt a currently running thread with the same priority.
135
136          (b)  If the thread's priority is unchanged, its position in the  run
137               list is unchanged.
138
139          (c)  If  the thread's priority is lowered, it is placed at the front
140               of the list for its new priority.
141
142          According to POSIX.1-2008, changes to a thread's priority  (or  pol‐
143          icy)  using  any mechanism other than pthread_setschedprio(3) should
144          result in the thread being placed at the end of  the  list  for  its
145          priority.
146
147       •  A thread calling sched_yield(2) will be put at the end of the list.
148
149       No  other events will move a thread scheduled under the SCHED_FIFO pol‐
150       icy in the wait list of runnable threads with equal static priority.
151
152       A SCHED_FIFO thread runs until either it is blocked by an I/O  request,
153       it   is   preempted   by   a   higher  priority  thread,  or  it  calls
154       sched_yield(2).
155
156   SCHED_RR: Round-robin scheduling
157       SCHED_RR is a simple enhancement of SCHED_FIFO.   Everything  described
158       above  for SCHED_FIFO also applies to SCHED_RR, except that each thread
159       is allowed to run only for a  maximum  time  quantum.   If  a  SCHED_RR
160       thread  has  been running for a time period equal to or longer than the
161       time quantum, it will be put at the end of the list for  its  priority.
162       A  SCHED_RR  thread that has been preempted by a higher priority thread
163       and subsequently resumes execution as a running  thread  will  complete
164       the  unexpired  portion of its round-robin time quantum.  The length of
165       the time quantum can be retrieved using sched_rr_get_interval(2).
166
167   SCHED_DEADLINE: Sporadic task model deadline scheduling
168       Since  Linux  3.14,  Linux  provides  a  deadline   scheduling   policy
169       (SCHED_DEADLINE).   This  policy  is  currently  implemented using GEDF
170       (Global Earliest Deadline First)  in  conjunction  with  CBS  (Constant
171       Bandwidth  Server).   To  set  and fetch this policy and associated at‐
172       tributes,  one  must  use  the  Linux-specific   sched_setattr(2)   and
173       sched_getattr(2) system calls.
174
175       A  sporadic  task is one that has a sequence of jobs, where each job is
176       activated at most once per period.  Each job also has a relative  dead‐
177       line,  before which it should finish execution, and a computation time,
178       which is the CPU time necessary for executing the job.  The moment when
179       a  task wakes up because a new job has to be executed is called the ar‐
180       rival time (also referred to as the request time or release time).  The
181       start time is the time at which a task starts its execution.  The abso‐
182       lute deadline is thus obtained by adding the relative deadline  to  the
183       arrival time.
184
185       The following diagram clarifies these terms:
186
187           arrival/wakeup                    absolute deadline
188                |    start time                    |
189                |        |                         |
190                v        v                         v
191           -----x--------xooooooooooooooooo--------x--------x---
192                         |<- comp. time ->|
193                |<------- relative deadline ------>|
194                |<-------------- period ------------------->|
195
196       When  setting  a  SCHED_DEADLINE  policy  for  a thread using sched_se‐
197       tattr(2), one can specify three parameters: Runtime, Deadline, and  Pe‐
198       riod.   These parameters do not necessarily correspond to the aforemen‐
199       tioned terms: usual practice is to set Runtime to something bigger than
200       the  average  computation  time  (or worst-case execution time for hard
201       real-time tasks), Deadline to the relative deadline, and Period to  the
202       period of the task.  Thus, for SCHED_DEADLINE scheduling, we have:
203
204           arrival/wakeup                    absolute deadline
205                |    start time                    |
206                |        |                         |
207                v        v                         v
208           -----x--------xooooooooooooooooo--------x--------x---
209                         |<-- Runtime ------->|
210                |<----------- Deadline ----------->|
211                |<-------------- Period ------------------->|
212
213       The  three  deadline-scheduling parameters correspond to the sched_run‐
214       time, sched_deadline, and sched_period fields of the sched_attr  struc‐
215       ture;  see  sched_setattr(2).   These fields express values in nanosec‐
216       onds.  If sched_period is specified as 0, then it is made the  same  as
217       sched_deadline.
218
219       The kernel requires that:
220
221           sched_runtime <= sched_deadline <= sched_period
222
223       In  addition,  under  the  current implementation, all of the parameter
224       values must be at least 1024 (i.e., just over one microsecond, which is
225       the  resolution  of the implementation), and less than 2^63.  If any of
226       these checks fails, sched_setattr(2) fails with the error EINVAL.
227
228       The  CBS  guarantees  non-interference  between  tasks,  by  throttling
229       threads that attempt to over-run their specified Runtime.
230
231       To ensure deadline scheduling guarantees, the kernel must prevent situ‐
232       ations where the set of SCHED_DEADLINE threads is not feasible (schedu‐
233       lable)  within  the given constraints.  The kernel thus performs an ad‐
234       mittance test when setting or changing SCHED_DEADLINE  policy  and  at‐
235       tributes.   This admission test calculates whether the change is feasi‐
236       ble; if it is not, sched_setattr(2) fails with the error EBUSY.
237
238       For example, it is required (but not necessarily  sufficient)  for  the
239       total  utilization to be less than or equal to the total number of CPUs
240       available, where, since each thread can maximally run for  Runtime  per
241       Period, that thread's utilization is its Runtime divided by its Period.
242
243       In  order  to fulfill the guarantees that are made when a thread is ad‐
244       mitted to the SCHED_DEADLINE policy,  SCHED_DEADLINE  threads  are  the
245       highest  priority  (user  controllable)  threads  in the system; if any
246       SCHED_DEADLINE thread is runnable, it will preempt any thread scheduled
247       under one of the other policies.
248
249       A call to fork(2) by a thread scheduled under the SCHED_DEADLINE policy
250       fails with the error EAGAIN, unless the thread  has  its  reset-on-fork
251       flag set (see below).
252
253       A  SCHED_DEADLINE  thread that calls sched_yield(2) will yield the cur‐
254       rent job and wait for a new period to begin.
255
256   SCHED_OTHER: Default Linux time-sharing scheduling
257       SCHED_OTHER can be used at only static priority 0 (i.e., threads  under
258       real-time  policies  always  have priority over SCHED_OTHER processes).
259       SCHED_OTHER is the standard Linux time-sharing scheduler  that  is  in‐
260       tended for all threads that do not require the special real-time mecha‐
261       nisms.
262
263       The thread to run is chosen from the static priority 0 list based on  a
264       dynamic priority that is determined only inside this list.  The dynamic
265       priority is based on the nice value (see below) and  is  increased  for
266       each  time quantum the thread is ready to run, but denied to run by the
267       scheduler.  This ensures fair progress among all SCHED_OTHER threads.
268
269       In the Linux kernel source code, the  SCHED_OTHER  policy  is  actually
270       named SCHED_NORMAL.
271
272   The nice value
273       The  nice  value  is an attribute that can be used to influence the CPU
274       scheduler to favor or disfavor a process in scheduling  decisions.   It
275       affects  the scheduling of SCHED_OTHER and SCHED_BATCH (see below) pro‐
276       cesses.  The nice value can be modified using nice(2),  setpriority(2),
277       or sched_setattr(2).
278
279       According  to  POSIX.1, the nice value is a per-process attribute; that
280       is, the threads in a process should share a nice  value.   However,  on
281       Linux,  the  nice value is a per-thread attribute: different threads in
282       the same process may have different nice values.
283
284       The range of the nice value varies  across  UNIX  systems.   On  modern
285       Linux, the range is -20 (high priority) to +19 (low priority).  On some
286       other systems, the range is -20..20.  Very early Linux kernels  (before
287       Linux 2.0) had the range -infinity..15.
288
289       The  degree  to which the nice value affects the relative scheduling of
290       SCHED_OTHER processes likewise varies across UNIX  systems  and  across
291       Linux kernel versions.
292
293       With  the advent of the CFS scheduler in Linux 2.6.23, Linux adopted an
294       algorithm that causes relative differences in nice  values  to  have  a
295       much stronger effect.  In the current implementation, each unit of dif‐
296       ference in the nice values of two processes results in a factor of 1.25
297       in  the  degree  to  which  the  scheduler  favors  the higher priority
298       process.  This causes very low nice values (+19) to truly provide  lit‐
299       tle  CPU  to a process whenever there is any other higher priority load
300       on the system, and makes high nice values (-20) deliver most of the CPU
301       to applications that require it (e.g., some audio applications).
302
303       On  Linux, the RLIMIT_NICE resource limit can be used to define a limit
304       to which an unprivileged process's nice value can be raised; see  setr‐
305       limit(2) for details.
306
307       For further details on the nice value, see the subsections on the auto‐
308       group feature and group scheduling, below.
309
310   SCHED_BATCH: Scheduling batch processes
311       (Since Linux 2.6.16.)  SCHED_BATCH can be used only at static  priority
312       0.   This  policy  is  similar  to SCHED_OTHER in that it schedules the
313       thread according to its dynamic priority (based  on  the  nice  value).
314       The  difference  is that this policy will cause the scheduler to always
315       assume that the thread is CPU-intensive.  Consequently,  the  scheduler
316       will  apply a small scheduling penalty with respect to wakeup behavior,
317       so that this thread is mildly disfavored in scheduling decisions.
318
319       This policy is useful for workloads that are noninteractive, but do not
320       want to lower their nice value, and for workloads that want a determin‐
321       istic scheduling policy without interactivity causing extra preemptions
322       (between the workload's tasks).
323
324   SCHED_IDLE: Scheduling very low priority jobs
325       (Since  Linux  2.6.23.)  SCHED_IDLE can be used only at static priority
326       0; the process nice value has no influence for this policy.
327
328       This policy is intended for running  jobs  at  extremely  low  priority
329       (lower  even  than a +19 nice value with the SCHED_OTHER or SCHED_BATCH
330       policies).
331
332   Resetting scheduling policy for child processes
333       Each thread has a reset-on-fork scheduling flag.   When  this  flag  is
334       set,  children  created by fork(2) do not inherit privileged scheduling
335       policies.  The reset-on-fork flag can be set by either:
336
337       •  ORing the SCHED_RESET_ON_FORK flag into  the  policy  argument  when
338          calling sched_setscheduler(2) (since Linux 2.6.32); or
339
340       •  specifying  the  SCHED_FLAG_RESET_ON_FORK  flag  in attr.sched_flags
341          when calling sched_setattr(2).
342
343       Note that the constants used with these two APIs have different  names.
344       The  state of the reset-on-fork flag can analogously be retrieved using
345       sched_getscheduler(2) and sched_getattr(2).
346
347       The reset-on-fork feature is intended for media-playback  applications,
348       and  can  be used to prevent applications evading the RLIMIT_RTTIME re‐
349       source limit (see getrlimit(2)) by creating multiple child processes.
350
351       More precisely, if the reset-on-fork flag is set, the  following  rules
352       apply for subsequently created children:
353
354       •  If  the  calling  thread  has  a  scheduling policy of SCHED_FIFO or
355          SCHED_RR, the policy is reset to SCHED_OTHER in child processes.
356
357       •  If the calling process has a negative nice value, the nice value  is
358          reset to zero in child processes.
359
360       After  the reset-on-fork flag has been enabled, it can be reset only if
361       the thread has the CAP_SYS_NICE capability.  This flag is  disabled  in
362       child processes created by fork(2).
363
364   Privileges and resource limits
365       Before  Linux  2.6.12, only privileged (CAP_SYS_NICE) threads can set a
366       nonzero static priority (i.e., set a real-time scheduling policy).  The
367       only  change  that  an  unprivileged  thread  can  make  is  to set the
368       SCHED_OTHER policy, and this can be done only if the effective user  ID
369       of  the  caller  matches  the  real  or effective user ID of the target
370       thread (i.e., the thread  specified  by  pid)  whose  policy  is  being
371       changed.
372
373       A  thread must be privileged (CAP_SYS_NICE) in order to set or modify a
374       SCHED_DEADLINE policy.
375
376       Since Linux 2.6.12, the RLIMIT_RTPRIO resource limit defines a  ceiling
377       on  an  unprivileged  thread's  static  priority  for  the SCHED_RR and
378       SCHED_FIFO policies.  The rules for changing scheduling policy and pri‐
379       ority are as follows:
380
381       •  If  an  unprivileged  thread has a nonzero RLIMIT_RTPRIO soft limit,
382          then it can change its scheduling policy and  priority,  subject  to
383          the  restriction  that  the priority cannot be set to a value higher
384          than the maximum of its current priority and its RLIMIT_RTPRIO  soft
385          limit.
386
387       •  If  the  RLIMIT_RTPRIO  soft  limit  is  0,  then the only permitted
388          changes are to lower the priority, or to switch to  a  non-real-time
389          policy.
390
391       •  Subject to the same rules, another unprivileged thread can also make
392          these changes, as long as the effective user ID of the thread making
393          the  change  matches  the  real  or  effective user ID of the target
394          thread.
395
396       •  Special rules apply for the SCHED_IDLE policy.  Before Linux 2.6.39,
397          an unprivileged thread operating under this policy cannot change its
398          policy, regardless of the value of its RLIMIT_RTPRIO resource limit.
399          Since  Linux 2.6.39, an unprivileged thread can switch to either the
400          SCHED_BATCH or the SCHED_OTHER policy so  long  as  its  nice  value
401          falls  within  the range permitted by its RLIMIT_NICE resource limit
402          (see getrlimit(2)).
403
404       Privileged (CAP_SYS_NICE) threads ignore the  RLIMIT_RTPRIO  limit;  as
405       with  older kernels, they can make arbitrary changes to scheduling pol‐
406       icy  and  priority.   See  getrlimit(2)  for  further  information   on
407       RLIMIT_RTPRIO.
408
409   Limiting the CPU usage of real-time and deadline processes
410       A nonblocking infinite loop in a thread scheduled under the SCHED_FIFO,
411       SCHED_RR, or SCHED_DEADLINE policy  can  potentially  block  all  other
412       threads  from accessing the CPU forever.  Before Linux 2.6.25, the only
413       way of preventing a runaway real-time process from freezing the  system
414       was  to  run  (at  the console) a shell scheduled under a higher static
415       priority than the tested application.  This allows an emergency kill of
416       tested  real-time  applications  that  do not block or terminate as ex‐
417       pected.
418
419       Since Linux 2.6.25, there are other techniques for dealing with runaway
420       real-time  and  deadline  processes.   One  of  these  is  to  use  the
421       RLIMIT_RTTIME resource limit to set a ceiling on the CPU  time  that  a
422       real-time process may consume.  See getrlimit(2) for details.
423
424       Since  Linux  2.6.25,  Linux  also provides two /proc files that can be
425       used to reserve a certain amount of CPU time to be  used  by  non-real-
426       time  processes.   Reserving  CPU  time in this fashion allows some CPU
427       time to be allocated to (say) a root shell that can be used to  kill  a
428       runaway  process.  Both of these files specify time values in microsec‐
429       onds:
430
431       /proc/sys/kernel/sched_rt_period_us
432              This file specifies a scheduling period that  is  equivalent  to
433              100%  CPU bandwidth.  The value in this file can range from 1 to
434              INT_MAX, giving an operating range of 1 microsecond to around 35
435              minutes.   The  default  value in this file is 1,000,000 (1 sec‐
436              ond).
437
438       /proc/sys/kernel/sched_rt_runtime_us
439              The value in this file specifies how much of the  "period"  time
440              can be used by all real-time and deadline scheduled processes on
441              the system.  The value  in  this  file  can  range  from  -1  to
442              INT_MAX-1.  Specifying -1 makes the run time the same as the pe‐
443              riod; that is, no CPU time is set aside for  non-real-time  pro‐
444              cesses  (which  was  the behavior before Linux 2.6.25).  The de‐
445              fault value in this file is 950,000 (0.95 seconds), meaning that
446              5%  of the CPU time is reserved for processes that don't run un‐
447              der a real-time or deadline scheduling policy.
448
449   Response time
450       A blocked high priority thread waiting for I/O has a  certain  response
451       time  before  it  is  scheduled  again.   The  device driver writer can
452       greatly reduce this response time by using a "slow interrupt" interrupt
453       handler.
454
455   Miscellaneous
456       Child  processes  inherit the scheduling policy and parameters across a
457       fork(2).  The scheduling policy and parameters are preserved across ex‐
458       ecve(2).
459
460       Memory  locking is usually needed for real-time processes to avoid pag‐
461       ing delays; this can be done with mlock(2) or mlockall(2).
462
463   The autogroup feature
464       Since Linux 2.6.38, the kernel provides a feature known as autogrouping
465       to improve interactive desktop performance in the face of multiprocess,
466       CPU-intensive workloads such as building the Linux  kernel  with  large
467       numbers of parallel build processes (i.e., the make(1) -j flag).
468
469       This  feature  operates  in  conjunction with the CFS scheduler and re‐
470       quires a kernel that is configured with CONFIG_SCHED_AUTOGROUP.   On  a
471       running  system,  this  feature  is  enabled  or  disabled via the file
472       /proc/sys/kernel/sched_autogroup_enabled; a value  of  0  disables  the
473       feature, while a value of 1 enables it.  The default value in this file
474       is 1, unless the kernel was booted with the noautogroup parameter.
475
476       A new autogroup is created when a new session is created via setsid(2);
477       this  happens,  for  example, when a new terminal window is started.  A
478       new process created by fork(2) inherits its parent's autogroup  member‐
479       ship.   Thus, all of the processes in a session are members of the same
480       autogroup.  An autogroup  is  automatically  destroyed  when  the  last
481       process in the group terminates.
482
483       When  autogrouping  is  enabled, all of the members of an autogroup are
484       placed in the same kernel scheduler "task group".   The  CFS  scheduler
485       employs  an  algorithm  that  equalizes  the distribution of CPU cycles
486       across task groups.  The benefits of this for interactive desktop  per‐
487       formance can be described via the following example.
488
489       Suppose that there are two autogroups competing for the same CPU (i.e.,
490       presume either a single CPU system or the use of taskset(1) to  confine
491       all  the  processes to the same CPU on an SMP system).  The first group
492       contains ten CPU-bound processes  from  a  kernel  build  started  with
493       make  -j10.   The  other  contains  a single CPU-bound process: a video
494       player.  The effect of autogrouping is that the two  groups  will  each
495       receive half of the CPU cycles.  That is, the video player will receive
496       50% of the CPU cycles, rather than just 9% of the cycles,  which  would
497       likely lead to degraded video playback.  The situation on an SMP system
498       is more complex, but the general effect is the same: the scheduler dis‐
499       tributes CPU cycles across task groups such that an autogroup that con‐
500       tains a large number of CPU-bound processes does not end up hogging CPU
501       cycles at the expense of the other jobs on the system.
502
503       A  process's  autogroup  (task  group) membership can be viewed via the
504       file /proc/pid/autogroup:
505
506           $ cat /proc/1/autogroup
507           /autogroup-1 nice 0
508
509       This file can also be used to modify the CPU bandwidth allocated to  an
510       autogroup.  This is done by writing a number in the "nice" range to the
511       file to set the autogroup's nice value.  The allowed range is from  +19
512       (low priority) to -20 (high priority).  (Writing values outside of this
513       range causes write(2) to fail with the error EINVAL.)
514
515       The autogroup nice setting has the same meaning  as  the  process  nice
516       value,  but applies to distribution of CPU cycles to the autogroup as a
517       whole, based on the relative nice values of other  autogroups.   For  a
518       process  inside an autogroup, the CPU cycles that it receives will be a
519       product of the autogroup's nice value (compared  to  other  autogroups)
520       and  the  process's nice value (compared to other processes in the same
521       autogroup.
522
523       The use of the cgroups(7) CPU controller to place processes in  cgroups
524       other than the root CPU cgroup overrides the effect of autogrouping.
525
526       The  autogroup  feature groups only processes scheduled under non-real-
527       time policies (SCHED_OTHER, SCHED_BATCH, and SCHED_IDLE).  It does  not
528       group processes scheduled under real-time and deadline policies.  Those
529       processes are scheduled according to the rules described earlier.
530
531   The nice value and group scheduling
532       When scheduling non-real-time processes (i.e.,  those  scheduled  under
533       the  SCHED_OTHER, SCHED_BATCH, and SCHED_IDLE policies), the CFS sched‐
534       uler employs a technique known as "group scheduling", if the kernel was
535       configured with the CONFIG_FAIR_GROUP_SCHED option (which is typical).
536
537       Under  group  scheduling, threads are scheduled in "task groups".  Task
538       groups have a hierarchical relationship, rooted under the initial  task
539       group  on  the system, known as the "root task group".  Task groups are
540       formed in the following circumstances:
541
542       •  All of the threads in a CPU cgroup form a task group.  The parent of
543          this  task  group  is  the  task  group  of the corresponding parent
544          cgroup.
545
546       •  If autogrouping is enabled, then all of the threads  that  are  (im‐
547          plicitly) placed in an autogroup (i.e., the same session, as created
548          by setsid(2)) form a task group.  Each new autogroup is thus a sepa‐
549          rate  task group.  The root task group is the parent of all such au‐
550          togroups.
551
552       •  If autogrouping is enabled, then the root task group consists of all
553          processes  in the root CPU cgroup that were not otherwise implicitly
554          placed into a new autogroup.
555
556       •  If autogrouping is disabled, then the root task  group  consists  of
557          all processes in the root CPU cgroup.
558
559       •  If  group  scheduling  was disabled (i.e., the kernel was configured
560          without CONFIG_FAIR_GROUP_SCHED), then all of the processes  on  the
561          system are notionally placed in a single task group.
562
563       Under  group scheduling, a thread's nice value has an effect for sched‐
564       uling decisions only relative to other threads in the same task  group.
565       This  has  some surprising consequences in terms of the traditional se‐
566       mantics of the nice value on UNIX systems.   In  particular,  if  auto‐
567       grouping  is  enabled  (which is the default in various distributions),
568       then employing setpriority(2) or nice(1) on a  process  has  an  effect
569       only  for  scheduling  relative to other processes executed in the same
570       session (typically: the same terminal window).
571
572       Conversely, for two processes that are (for example) the sole CPU-bound
573       processes in different sessions (e.g., different terminal windows, each
574       of whose jobs are tied to different  autogroups),  modifying  the  nice
575       value  of  the process in one of the sessions has no effect in terms of
576       the scheduler's decisions relative to the process in the other session.
577       A  possibly useful workaround here is to use a command such as the fol‐
578       lowing to modify the autogroup nice value for all of the processes in a
579       terminal session:
580
581           $ echo 10 > /proc/self/autogroup
582
583   Real-time features in the mainline Linux kernel
584       Since Linux 2.6.18, Linux is gradually becoming equipped with real-time
585       capabilities, most of which are derived from the  former  realtime-pre‐
586       empt patch set.  Until the patches have been completely merged into the
587       mainline kernel, they must be installed to achieve the  best  real-time
588       performance.  These patches are named:
589
590           patch-kernelversion-rtpatchversion
591
592       and  can  be  downloaded  from  ⟨http://www.kernel.org/pub/linux/kernel
593       /projects/rt/⟩.
594
595       Without the patches and prior to their full inclusion into the mainline
596       kernel,  the  kernel  configuration  offers  only  the three preemption
597       classes CONFIG_PREEMPT_NONE, CONFIG_PREEMPT_VOLUNTARY, and  CONFIG_PRE‐
598       EMPT_DESKTOP  which respectively provide no, some, and considerable re‐
599       duction of the worst-case scheduling latency.
600
601       With the patches applied or after their full inclusion into  the  main‐
602       line  kernel,  the  additional configuration item CONFIG_PREEMPT_RT be‐
603       comes available.  If this is selected, Linux is transformed into a reg‐
604       ular  real-time  operating system.  The FIFO and RR scheduling policies
605       are then used to run a thread with true real-time priority and a  mini‐
606       mum worst-case scheduling latency.
607

NOTES

609       The  cgroups(7) CPU controller can be used to limit the CPU consumption
610       of groups of processes.
611
612       Originally, Standard Linux was intended as a general-purpose  operating
613       system  being able to handle background processes, interactive applica‐
614       tions, and less demanding  real-time  applications  (applications  that
615       need to usually meet timing deadlines).  Although the Linux 2.6 allowed
616       for kernel preemption and the newly introduced O(1)  scheduler  ensures
617       that  the  time needed to schedule is fixed and deterministic irrespec‐
618       tive of the number of active tasks, true real-time  computing  was  not
619       possible up to Linux 2.6.17.
620

SEE ALSO

622       chcpu(1), chrt(1), lscpu(1), ps(1), taskset(1), top(1), getpriority(2),
623       mlock(2), mlockall(2), munlock(2), munlockall(2), nice(2),
624       sched_get_priority_max(2), sched_get_priority_min(2),
625       sched_getaffinity(2), sched_getparam(2), sched_getscheduler(2),
626       sched_rr_get_interval(2), sched_setaffinity(2), sched_setparam(2),
627       sched_setscheduler(2), sched_yield(2), setpriority(2),
628       pthread_getaffinity_np(3), pthread_getschedparam(3),
629       pthread_setaffinity_np(3), sched_getcpu(3), capabilities(7), cpuset(7)
630
631       Programming for the real  world  -  POSIX.4  by  Bill  O.  Gallmeister,
632       O'Reilly & Associates, Inc., ISBN 1-56592-074-0.
633
634       The  Linux  kernel  source files Documentation/scheduler/sched-deadline
635       .txt,    Documentation/scheduler/sched-rt-group.txt,     Documentation/
636       scheduler/sched-design-CFS.txt,       and      Documentation/scheduler/
637       sched-nice-design.txt
638
639
640
641Linux man-pages 6.04              2023-02-10                          sched(7)
Impressum