1 .\" Copyright (C) 2014 Michael Kerrisk <mtk.manpages@gmail.com>
2 .\" and Copyright (C) 2014 Peter Zijlstra <peterz@infradead.org>
4 .\" %%%LICENSE_START(VERBATIM)
5 .\" Permission is granted to make and distribute verbatim copies of this
6 .\" manual provided the copyright notice and this permission notice are
7 .\" preserved on all copies.
9 .\" Permission is granted to copy and distribute modified versions of this
10 .\" manual under the conditions for verbatim copying, provided that the
11 .\" entire resulting derived work is distributed under the terms of a
12 .\" permission notice identical to this one.
14 .\" Since the Linux kernel and libraries are constantly changing, this
15 .\" manual page may be incorrect or out-of-date. The author(s) assume no
16 .\" responsibility for errors or omissions, or for damages resulting from
17 .\" the use of the information contained herein. The author(s) may not
18 .\" have taken the same level of care in the production of this manual,
19 .\" which is licensed free of charge, as they might when working
22 .\" Formatted or processed versions of this manual, if unaccompanied by
23 .\" the source, must acknowledge the copyright and authors of this work.
26 .TH SCHED_SETATTR 2 2019-03-06 "Linux" "Linux Programmer's Manual"
28 sched_setattr, sched_getattr \-
29 set and get scheduling policy and attributes
34 .BI "int sched_setattr(pid_t " pid ", struct sched_attr *" attr ,
35 .BI " unsigned int " flags );
37 .BI "int sched_getattr(pid_t " pid ", struct sched_attr *" attr ,
38 .BI " unsigned int " size ", unsigned int " flags );
40 .\" FIXME . Add feature test macro requirements
45 system call sets the scheduling policy and
46 associated attributes for the thread whose ID is specified in
51 the scheduling policy and attributes of the calling thread will be set.
53 Currently, Linux supports the following "normal"
54 (i.e., non-real-time) scheduling policies as values that may be specified in
58 the standard round-robin time-sharing policy;
59 .\" In the 2.6 kernel sources, SCHED_OTHER is actually called
63 for "batch" style execution of processes; and
68 low priority background jobs.
70 Various "real-time" policies are also supported,
71 for special time-critical applications that need precise control over
72 the way in which runnable threads are selected for execution.
73 For the rules governing when a process may use these policies, see
75 The real-time policies that may be specified in
80 a first-in, first-out policy; and
85 Linux also provides the following policy:
88 a deadline scheduling policy; see
94 argument is a pointer to a structure that defines
95 the new scheduling policy and attributes for the specified thread.
96 This structure has the following form:
101 u32 size; /* Size of this structure */
102 u32 sched_policy; /* Policy (SCHED_*) */
103 u64 sched_flags; /* Flags */
104 s32 sched_nice; /* Nice value (SCHED_OTHER,
106 u32 sched_priority; /* Static priority (SCHED_FIFO,
108 /* Remaining fields are for SCHED_DEADLINE */
116 The fields of this structure are as follows:
119 This field should be set to the size of the structure in bytes, as in
120 .IR "sizeof(struct sched_attr)" .
121 If the provided structure is smaller than the kernel structure,
122 any additional fields are assumed to be '0'.
123 If the provided structure is larger than the kernel structure,
124 the kernel verifies that all additional fields are 0;
131 to contain the size of the kernel structure.
133 The above behavior when the size of the user-space
135 structure does not match the size of the kernel structure
136 allows for future extensibility of the interface.
137 Malformed applications that pass oversize structures
138 won't break in the future if the size of the kernel
140 structure is increased.
142 it could also allow applications that know about a larger user-space
144 structure to determine whether they are running on an older kernel
145 that does not support the larger structure.
148 This field specifies the scheduling policy, as one of the
153 This field contains contains zero or more of the following flags
154 that are ORed together to control scheduling behavior:
157 .BR SCHED_FLAG_RESET_ON_FORK
160 do not inherit privileged scheduling policies.
165 .BR SCHED_FLAG_RECLAIM " (since Linux 4.13)"
166 .\" 2d4283e9d583a3ee8cfb1cbb9c1270614df4c29d
169 thread to reclaim bandwidth unused by other real-time threads.
170 .\" Bandwidth reclaim is done via the GRUB algorithm; see
171 .\" Documentation/scheduler/sched-deadline.txt
173 .BR SCHED_FLAG_DL_OVERRUN " (since Linux 4.16)"
174 .\" commit 34be39305a77b8b1ec9f279163c7cdb6cc719b91
175 This flag allows an application to get informed about run-time overruns in
178 Such overruns may be caused by (for example) coarse execution time accounting
179 or incorrect parameter assignment.
180 Notification takes the form of a
182 signal which is generated on each overrun.
190 rather than thread-directed.
191 This is probably a bug.
194 is being used to set a per-thread attribute.
195 On the other hand, if the process-directed signal is delivered to
196 a thread inside the process other than the one that had a run-time overrun,
197 the application has no way of knowing which thread overran.
201 This field specifies the nice value to be set when specifying
207 The nice value is a number in the range \-20 (high priority)
208 to +19 (low priority); see
212 This field specifies the static priority to be set when specifying
218 The allowed range of priorities for these policies can be determined using
219 .BR sched_get_priority_min (2)
221 .BR sched_get_priority_max (2).
222 For other policies, this field must be specified as 0.
225 This field specifies the "Runtime" parameter for deadline scheduling.
226 The value is expressed in nanoseconds.
227 This field, and the next two fields,
230 scheduling; for further details, see
234 This field specifies the "Deadline" parameter for deadline scheduling.
235 The value is expressed in nanoseconds.
238 This field specifies the "Period" parameter for deadline scheduling.
239 The value is expressed in nanoseconds.
243 argument is provided to allow for future extensions to the interface;
244 in the current implementation it must be specified as 0.
250 system call fetches the scheduling policy and the
251 associated attributes for the thread whose ID is specified in
256 the scheduling policy and attributes of the calling thread
261 argument should be set to the size of the
263 structure as known to user space.
264 The value must be at least as large as the size of the initially published
266 structure, or the call fails with the error
269 The retrieved scheduling attributes are placed in the fields of the
271 structure pointed to by
279 If the caller-provided
281 buffer is larger than the kernel's
284 the additional bytes in the user-space structure are not touched.
285 If the caller-provided structure is smaller than the kernel
287 structure and the kernel needs to return values outside the provided space,
292 .BR sched_setattr (),
293 these semantics allow for future extensibility of the interface.
297 argument is provided to allow for future extensions to the interface;
298 in the current implementation it must be specified as 0.
305 On error, \-1 is returned, and
307 is set to indicate the cause of the error.
312 can both fail for the following reasons:
323 The thread whose ID is
329 can fail for the following reasons:
332 The buffer specified by
340 is invalid; that is, it is smaller than the initial version of the
342 structure (48 bytes) or larger than the system page size.
346 can fail for the following reasons:
349 The buffer specified by
353 is larger than the kernel structure,
354 and one or more of the excess bytes is nonzero.
358 admission control failure, see
363 is not one of the recognized policies;
365 contains a flag other than
366 .BR SCHED_FLAG_RESET_ON_FORK ;
368 .I attr.sched_priority
373 and the deadline scheduling parameters in
378 The caller does not have appropriate privileges.
381 The CPU affinity mask of the thread specified by
383 does not include all CPUs in the system
385 .BR sched_setaffinity (2)).
387 These system calls first appeared in Linux 3.14.
388 .\" FIXME . Add glibc version
390 These system calls are nonstandard Linux extensions.
393 provides a superset of the functionality of
394 .BR sched_setscheduler (2),
395 .BR sched_setparam (2),
397 and (other than the ability to set the priority of all processes
398 belonging to a specified user or all processes in a specified group)
402 provides a superset of the functionality of
403 .BR sched_getscheduler (2),
404 .BR sched_getparam (2),
408 In Linux versions up to
409 .\" FIXME . patch sent to Peter Zijlstra
411 .BR sched_settattr ()
412 failed with the error
416 for the case described in ERRORS.
417 .\" In Linux versions up to up 3.15,
418 .\" FIXME . patch from Peter Zijlstra pending
419 .\" .BR sched_setattr ()
420 .\" allowed a negative
421 .\" .I attr.sched_policy
428 .BR sched_get_priority_max (2),
429 .BR sched_get_priority_min (2),
430 .BR sched_getaffinity (2),
431 .BR sched_getparam (2),
432 .BR sched_getscheduler (2),
433 .BR sched_rr_get_interval (2),
434 .BR sched_setaffinity (2),
435 .BR sched_setparam (2),
436 .BR sched_setscheduler (2),
439 .BR pthread_getschedparam (3),
440 .BR pthread_setschedparam (3),
441 .BR pthread_setschedprio (3),
442 .BR capabilities (7),