1SCHED_SETSCHEDULER(2) Linux Programmer's Manual SCHED_SETSCHEDULER(2)
2
6 sched_setscheduler, sched_getscheduler - set and get scheduling pol‐
7 icy/parameters
8
10 #include <sched.h>
11 int sched_setscheduler(pid_t pid, int policy,
13 const struct sched_param *param);
14 int sched_getscheduler(pid_t pid);
16 struct sched_param {
18 ...
19 int sched_priority;
20 ...
21 };
22
24 sched_setscheduler() sets both the scheduling policy and the associated
25 parameters for the process whose ID is specified in pid. If pid equals
26 zero, the scheduling policy and parameters of the calling process will
27 be set. The interpretation of the argument param depends on the
28 selected policy. Currently, Linux supports the following "normal"
29 (i.e., non-real-time) scheduling policies:
30 SCHED_OTHER the standard round-robin time-sharing policy;
32 SCHED_BATCH for "batch" style execution of processes; and
34 SCHED_IDLE for running very low priority background jobs.
36 The following "real-time" policies are also supported, for special
38 time-critical applications that need precise control over the way in
39 which runnable processes are selected for execution:
40 SCHED_FIFO a first-in, first-out policy; and
42 SCHED_RR a round-robin policy.
44 The semantics of each of these policies are detailed below.
46 sched_getscheduler() queries the scheduling policy currently applied to
48 the process identified by pid. If pid equals zero, the policy of the
49 calling process will be retrieved.
50 Scheduling Policies
52 The scheduler is the kernel component that decides which runnable
53 process will be executed by the CPU next. Each process has an associ‐
54 ated scheduling policy and a static scheduling priority, sched_prior‐
55 ity; these are the settings that are modified by sched_setscheduler().
56 The scheduler makes it decisions based on knowledge of the scheduling
57 policy and static priority of all processes on the system.
58 For processes scheduled under one of the normal scheduling policies
60 (SCHED_OTHER, SCHED_IDLE, SCHED_BATCH), sched_priority is not used in
61 scheduling decisions (it must be specified as 0).
62 Processes scheduled under one of the real-time policies (SCHED_FIFO,
64 SCHED_RR) have a sched_priority value in the range 1 (low) to 99
65 (high). (As the numbers imply, real-time processes always have higher
66 priority than normal processes.) Note well: POSIX.1-2001 only requires
67 an implementation to support a minimum 32 distinct priority levels for
68 the real-time policies, and some systems supply just this minimum.
69 Portable programs should use sched_get_priority_min(2) and
70 sched_get_priority_max(2) to find the range of priorities supported for
71 a particular policy.
72 Conceptually, the scheduler maintains a list of runnable processes for
74 each possible sched_priority value. In order to determine which
75 process runs next, the scheduler looks for the nonempty list with the
76 highest static priority and selects the process at the head of this
77 list.
78 A process's scheduling policy determines where it will be inserted into
80 the list of processes with equal static priority and how it will move
81 inside this list.
82 All scheduling is preemptive: if a process with a higher static prior‐
84 ity becomes ready to run, the currently running process will be pre‐
85 empted and returned to the wait list for its static priority level.
86 The scheduling policy only determines the ordering within the list of
87 runnable processes with equal static priority.
88 SCHED_FIFO: First In-First Out scheduling
90 SCHED_FIFO can only be used with static priorities higher than 0, which
91 means that when a SCHED_FIFO processes becomes runnable, it will always
92 immediately preempt any currently running SCHED_OTHER, SCHED_BATCH, or
93 SCHED_IDLE process. SCHED_FIFO is a simple scheduling algorithm with‐
94 out time slicing. For processes scheduled under the SCHED_FIFO policy,
95 the following rules apply:
96 * A SCHED_FIFO process that has been preempted by another process of
98 higher priority will stay at the head of the list for its priority
99 and will resume execution as soon as all processes of higher prior‐
100 ity are blocked again.
101 * When a SCHED_FIFO process becomes runnable, it will be inserted at
103 the end of the list for its priority.
104 * A call to sched_setscheduler() or sched_setparam(2) will put the
106 SCHED_FIFO (or SCHED_RR) process identified by pid at the start of
107 the list if it was runnable. As a consequence, it may preempt the
108 currently running process if it has the same priority.
109 (POSIX.1-2001 specifies that the process should go to the end of the
110 list.)
111 * A process calling sched_yield(2) will be put at the end of the list.
113 No other events will move a process scheduled under the SCHED_FIFO pol‐
115 icy in the wait list of runnable processes with equal static priority.
116 A SCHED_FIFO process runs until either it is blocked by an I/O request,
118 it is preempted by a higher priority process, or it calls
119 sched_yield(2).
120 SCHED_RR: Round Robin scheduling
122 SCHED_RR is a simple enhancement of SCHED_FIFO. Everything described
123 above for SCHED_FIFO also applies to SCHED_RR, except that each process
124 is only allowed to run for a maximum time quantum. If a SCHED_RR
125 process has been running for a time period equal to or longer than the
126 time quantum, it will be put at the end of the list for its priority.
127 A SCHED_RR process that has been preempted by a higher priority process
128 and subsequently resumes execution as a running process will complete
129 the unexpired portion of its round robin time quantum. The length of
130 the time quantum can be retrieved using sched_rr_get_interval(2).
131 SCHED_OTHER: Default Linux time-sharing scheduling
133 SCHED_OTHER can only be used at static priority 0. SCHED_OTHER is the
134 standard Linux time-sharing scheduler that is intended for all pro‐
135 cesses that do not require the special real-time mechanisms. The
136 process to run is chosen from the static priority 0 list based on a
137 dynamic priority that is determined only inside this list. The dynamic
138 priority is based on the nice value (set by nice(2) or setpriority(2))
139 and increased for each time quantum the process is ready to run, but
140 denied to run by the scheduler. This ensures fair progress among all
141 SCHED_OTHER processes.
142 SCHED_BATCH: Scheduling batch processes
144 (Since Linux 2.6.16.) SCHED_BATCH can only be used at static priority
145 0. This policy is similar to SCHED_OTHER in that it schedules the
146 process according to its dynamic priority (based on the nice value).
147 The difference is that this policy will cause the scheduler to always
148 assume that the process is CPU-intensive. Consequently, the scheduler
149 will apply a small scheduling penalty with respect to wakeup behaviour,
150 so that this process is mildly disfavored in scheduling decisions.
151 This policy is useful for workloads that are noninteractive, but do not
153 want to lower their nice value, and for workloads that want a determin‐
154 istic scheduling policy without interactivity causing extra preemptions
155 (between the workload's tasks).
156 SCHED_IDLE: Scheduling very low priority jobs
158 (Since Linux 2.6.23.) SCHED_IDLE can only be used at static priority
159 0; the process nice value has no influence for this policy.
160 This policy is intended for running jobs at extremely low priority
162 (lower even than a +19 nice value with the SCHED_OTHER or SCHED_BATCH
163 policies).
164 Resetting scheduling policy for child processes
166 Since Linux 2.6.32, the SCHED_RESET_ON_FORK flag can be ORed in policy
167 when calling sched_setscheduler(). As a result of including this flag,
168 children created by fork(2) do not inherit privileged scheduling poli‐
169 cies. This feature is intended for media-playback applications, and
170 can be used to prevent applications evading the RLIMIT_RTTIME resource
171 limit (see getrlimit(2)) by creating multiple child processes.
172 More precisely, if the SCHED_RESET_ON_FORK flag is specified, the fol‐
174 lowing rules apply for subsequently created children:
175 * If the calling process has a scheduling policy of SCHED_FIFO or
177 SCHED_RR, the policy is reset to SCHED_OTHER in child processes.
178 * If the calling processes has a negative nuce value, the nice value
180 is reset to zero in child processes.
181 After the SCHED_RESET_ON_FORK flag has been enabled, it can only be
183 reset if the process has the CAP_SYS_NICE capability. This flag is
184 disabled in child processes created by fork(2).
185 The SCHED_RESET_ON_FORK flag is visible in the policy value returned by
187 sched_getscheduler()
188 Privileges and resource limits
190 In Linux kernels before 2.6.12, only privileged (CAP_SYS_NICE) pro‐
191 cesses can set a nonzero static priority (i.e., set a real-time sched‐
192 uling policy). The only change that an unprivileged process can make
193 is to set the SCHED_OTHER policy, and this can only be done if the
194 effective user ID of the caller of sched_setscheduler() matches the
195 real or effective user ID of the target process (i.e., the process
196 specified by pid) whose policy is being changed.
197 Since Linux 2.6.12, the RLIMIT_RTPRIO resource limit defines a ceiling
199 on an unprivileged process's static priority for the SCHED_RR and
200 SCHED_FIFO policies. The rules for changing scheduling policy and pri‐
201 ority are as follows:
202 * If an unprivileged process has a nonzero RLIMIT_RTPRIO soft limit,
204 then it can change its scheduling policy and priority, subject to
205 the restriction that the priority cannot be set to a value higher
206 than the maximum of its current priority and its RLIMIT_RTPRIO soft
207 limit.
208 * If the RLIMIT_RTPRIO soft limit is 0, then the only permitted
210 changes are to lower the priority, or to switch to a non-real-time
211 policy.
212 * Subject to the same rules, another unprivileged process can also
214 make these changes, as long as the effective user ID of the process
215 making the change matches the real or effective user ID of the tar‐
216 get process.
217 * Special rules apply for the SCHED_IDLE: an unprivileged process
219 operating under this policy cannot change its policy, regardless of
220 the value of its RLIMIT_RTPRIO resource limit.
221 Privileged (CAP_SYS_NICE) processes ignore the RLIMIT_RTPRIO limit; as
223 with older kernels, they can make arbitrary changes to scheduling pol‐
224 icy and priority. See getrlimit(2) for further information on
225 RLIMIT_RTPRIO.
226 Response time
228 A blocked high priority process waiting for the I/O has a certain
229 response time before it is scheduled again. The device driver writer
230 can greatly reduce this response time by using a "slow interrupt"
231 interrupt handler.
232 Miscellaneous
234 Child processes inherit the scheduling policy and parameters across a
235 fork(2). The scheduling policy and parameters are preserved across
236 execve(2).
237 Memory locking is usually needed for real-time processes to avoid pag‐
239 ing delays; this can be done with mlock(2) or mlockall(2).
240 Since a nonblocking infinite loop in a process scheduled under
242 SCHED_FIFO or SCHED_RR will block all processes with lower priority
243 forever, a software developer should always keep available on the con‐
244 sole a shell scheduled under a higher static priority than the tested
245 application. This will allow an emergency kill of tested real-time
246 applications that do not block or terminate as expected. See also the
247 description of the RLIMIT_RTTIME resource limit in getrlimit(2).
248 POSIX systems on which sched_setscheduler() and sched_getscheduler()
250 are available define _POSIX_PRIORITY_SCHEDULING in <unistd.h>.
251
253 On success, sched_setscheduler() returns zero. On success,
254 sched_getscheduler() returns the policy for the process (a nonnegative
255 integer). On error, -1 is returned, and errno is set appropriately.
256
258 EINVAL The scheduling policy is not one of the recognized policies, or
259 param does not make sense for the policy.
260 EPERM The calling process does not have appropriate privileges.
262 ESRCH The process whose ID is pid could not be found.
264
266 POSIX.1-2001 (but see BUGS below). The SCHED_BATCH and SCHED_IDLE
267 policies are Linux-specific.
268
270 POSIX.1 does not detail the permissions that an unprivileged process
271 requires in order to call sched_setscheduler(), and details vary across
272 systems. For example, the Solaris 7 manual page says that the real or
273 effective user ID of the calling process must match the real user ID or
274 the save set-user-ID of the target process.
275 Originally, Standard Linux was intended as a general-purpose operating
277 system being able to handle background processes, interactive applica‐
278 tions, and less demanding real-time applications (applications that
279 need to usually meet timing deadlines). Although the Linux kernel 2.6
280 allowed for kernel preemption and the newly introduced O(1) scheduler
281 ensures that the time needed to schedule is fixed and deterministic
282 irrespective of the number of active tasks, true real-time computing
283 was not possible up to kernel version 2.6.17.
284 Real-time features in the mainline Linux kernel
286 From kernel version 2.6.18 onwards, however, Linux is gradually becom‐
287 ing equipped with real-time capabilities, most of which are derived
288 from the former realtime-preempt patches developed by Ingo Molnar,
289 Thomas Gleixner, Steven Rostedt, and others. Until the patches have
290 been completely merged into the mainline kernel (this is expected to be
291 around kernel version 2.6.30), they must be installed to achieve the
292 best real-time performance. These patches are named:
293 patch-kernelversion-rtpatchversion
295 and can be downloaded from http://www.kernel.org/pub/linux/ker‐
297 nel/projects/rt/.
298 Without the patches and prior to their full inclusion into the mainline
300 kernel, the kernel configuration offers only the three preemption
301 classes CONFIG_PREEMPT_NONE, CONFIG_PREEMPT_VOLUNTARY, and CONFIG_PRE‐
302 EMPT_DESKTOP which respectively provide no, some, and considerable
303 reduction of the worst-case scheduling latency.
304 With the patches applied or after their full inclusion into the main‐
306 line kernel, the additional configuration item CONFIG_PREEMPT_RT
307 becomes available. If this is selected, Linux is transformed into a
308 regular real-time operating system. The FIFO and RR scheduling poli‐
309 cies that can be selected using sched_setscheduler() are then used to
310 run a process with true real-time priority and a minimum worst-case
311 scheduling latency.
312
314 POSIX says that on success, sched_setscheduler() should return the pre‐
315 vious scheduling policy. Linux sched_setscheduler() does not conform
316 to this requirement, since it always returns 0 on success.
317
319 getpriority(2), mlock(2), mlockall(2), munlock(2), munlockall(2),
320 nice(2), sched_get_priority_max(2), sched_get_priority_min(2),
321 sched_getaffinity(2), sched_getparam(2), sched_rr_get_interval(2),
322 sched_setaffinity(2), sched_setparam(2), sched_yield(2), setprior‐
323 ity(2), capabilities(7), cpuset(7)
324 Programming for the real world - POSIX.4 by Bill O. Gallmeister,
326 O'Reilly & Associates, Inc., ISBN 1-56592-074-0
327 The kernel source file Documentation/scheduler/sched-rt-group.txt
329 (since kernel 2.6.25).
330
332 This page is part of release 3.25 of the Linux man-pages project. A
333 description of the project, and information about reporting bugs, can
334 be found at http://www.kernel.org/doc/man-pages/.
335Linux 2010-06-19 SCHED_SETSCHEDULER(2)