1SCHED_SETATTR(2) Linux Programmer's Manual SCHED_SETATTR(2)
2
6 sched_setattr, sched_getattr - set and get scheduling policy and at‐
7 tributes
8
10 #include <sched.h>
11 int sched_setattr(pid_t pid, struct sched_attr *attr,
13 unsigned int flags);
14 int sched_getattr(pid_t pid, struct sched_attr *attr,
16 unsigned int size, unsigned int flags);
17
19 sched_setattr()
20 The sched_setattr() system call sets the scheduling policy and associ‐
21 ated attributes for the thread whose ID is specified in pid. If pid
22 equals zero, the scheduling policy and attributes of the calling thread
23 will be set.
24 Currently, Linux supports the following "normal" (i.e., non-real-time)
26 scheduling policies as values that may be specified in policy:
27 SCHED_OTHER the standard round-robin time-sharing policy;
29 SCHED_BATCH for "batch" style execution of processes; and
31 SCHED_IDLE for running very low priority background jobs.
33 Various "real-time" policies are also supported, for special time-crit‐
35 ical applications that need precise control over the way in which
36 runnable threads are selected for execution. For the rules governing
37 when a process may use these policies, see sched(7). The real-time
38 policies that may be specified in policy are:
39 SCHED_FIFO a first-in, first-out policy; and
41 SCHED_RR a round-robin policy.
43 Linux also provides the following policy:
45 SCHED_DEADLINE
47 a deadline scheduling policy; see sched(7) for details.
48 The attr argument is a pointer to a structure that defines the new
50 scheduling policy and attributes for the specified thread. This struc‐
51 ture has the following form:
52 struct sched_attr {
54 u32 size; /* Size of this structure */
55 u32 sched_policy; /* Policy (SCHED_*) */
56 u64 sched_flags; /* Flags */
57 s32 sched_nice; /* Nice value (SCHED_OTHER,
58 SCHED_BATCH) */
59 u32 sched_priority; /* Static priority (SCHED_FIFO,
60 SCHED_RR) */
61 /* Remaining fields are for SCHED_DEADLINE */
62 u64 sched_runtime;
63 u64 sched_deadline;
64 u64 sched_period;
65 };
66 The fields of the sched_attr structure are as follows:
68 size This field should be set to the size of the structure in bytes,
70 as in sizeof(struct sched_attr). If the provided structure is
71 smaller than the kernel structure, any additional fields are as‐
72 sumed to be '0'. If the provided structure is larger than the
73 kernel structure, the kernel verifies that all additional fields
74 are 0; if they are not, sched_setattr() fails with the error
75 E2BIG and updates size to contain the size of the kernel struc‐
76 ture.
77 The above behavior when the size of the user-space sched_attr
79 structure does not match the size of the kernel structure allows
80 for future extensibility of the interface. Malformed applica‐
81 tions that pass oversize structures won't break in the future if
82 the size of the kernel sched_attr structure is increased. In
83 the future, it could also allow applications that know about a
84 larger user-space sched_attr structure to determine whether they
85 are running on an older kernel that does not support the larger
86 structure.
87 sched_policy
89 This field specifies the scheduling policy, as one of the
90 SCHED_* values listed above.
91 sched_flags
93 This field contains zero or more of the following flags that are
94 ORed together to control scheduling behavior:
95 SCHED_FLAG_RESET_ON_FORK
97 Children created by fork(2) do not inherit privileged
98 scheduling policies. See sched(7) for details.
99 SCHED_FLAG_RECLAIM (since Linux 4.13)
101 This flag allows a SCHED_DEADLINE thread to reclaim band‐
102 width unused by other real-time threads.
103 SCHED_FLAG_DL_OVERRUN (since Linux 4.16)
105 This flag allows an application to get informed about
106 run-time overruns in SCHED_DEADLINE threads. Such over‐
107 runs may be caused by (for example) coarse execution time
108 accounting or incorrect parameter assignment. Notifica‐
109 tion takes the form of a SIGXCPU signal which is gener‐
110 ated on each overrun.
111 This SIGXCPU signal is process-directed (see signal(7))
113 rather than thread-directed. This is probably a bug. On
114 the one hand, sched_setattr() is being used to set a per-
115 thread attribute. On the other hand, if the process-di‐
116 rected signal is delivered to a thread inside the process
117 other than the one that had a run-time overrun, the ap‐
118 plication has no way of knowing which thread overran.
119 sched_nice
121 This field specifies the nice value to be set when specifying
122 sched_policy as SCHED_OTHER or SCHED_BATCH. The nice value is a
123 number in the range -20 (high priority) to +19 (low priority);
124 see sched(7).
125 sched_priority
127 This field specifies the static priority to be set when specify‐
128 ing sched_policy as SCHED_FIFO or SCHED_RR. The allowed range
129 of priorities for these policies can be determined using
130 sched_get_priority_min(2) and sched_get_priority_max(2). For
131 other policies, this field must be specified as 0.
132 sched_runtime
134 This field specifies the "Runtime" parameter for deadline sched‐
135 uling. The value is expressed in nanoseconds. This field, and
136 the next two fields, are used only for SCHED_DEADLINE schedul‐
137 ing; for further details, see sched(7).
138 sched_deadline
140 This field specifies the "Deadline" parameter for deadline
141 scheduling. The value is expressed in nanoseconds.
142 sched_period
144 This field specifies the "Period" parameter for deadline sched‐
145 uling. The value is expressed in nanoseconds.
146 The flags argument is provided to allow for future extensions to the
148 interface; in the current implementation it must be specified as 0.
149 sched_getattr()
151 The sched_getattr() system call fetches the scheduling policy and the
152 associated attributes for the thread whose ID is specified in pid. If
153 pid equals zero, the scheduling policy and attributes of the calling
154 thread will be retrieved.
155 The size argument should be set to the size of the sched_attr structure
157 as known to user space. The value must be at least as large as the
158 size of the initially published sched_attr structure, or the call fails
159 with the error EINVAL.
160 The retrieved scheduling attributes are placed in the fields of the
162 sched_attr structure pointed to by attr. The kernel sets attr.size to
163 the size of its sched_attr structure.
164 If the caller-provided attr buffer is larger than the kernel's
166 sched_attr structure, the additional bytes in the user-space structure
167 are not touched. If the caller-provided structure is smaller than the
168 kernel sched_attr structure, the kernel will silently not return any
169 values which would be stored outside the provided space. As with
170 sched_setattr(), these semantics allow for future extensibility of the
171 interface.
172 The flags argument is provided to allow for future extensions to the
174 interface; in the current implementation it must be specified as 0.
175
177 On success, sched_setattr() and sched_getattr() return 0. On error, -1
178 is returned, and errno is set to indicate the cause of the error.
179
181 sched_getattr() and sched_setattr() can both fail for the following
182 reasons:
183 EINVAL attr is NULL; or pid is negative; or flags is not zero.
185 ESRCH The thread whose ID is pid could not be found.
187 In addition, sched_getattr() can fail for the following reasons:
189 E2BIG The buffer specified by size and attr is too small.
191 EINVAL size is invalid; that is, it is smaller than the initial version
193 of the sched_attr structure (48 bytes) or larger than the system
194 page size.
195 In addition, sched_setattr() can fail for the following reasons:
197 E2BIG The buffer specified by size and attr is larger than the kernel
199 structure, and one or more of the excess bytes is nonzero.
200 EBUSY SCHED_DEADLINE admission control failure, see sched(7).
202 EINVAL attr.sched_policy is not one of the recognized policies;
204 attr.sched_flags contains a flag other than SCHED_FLAG_RE‐
205 SET_ON_FORK; or attr.sched_priority is invalid; or
206 attr.sched_policy is SCHED_DEADLINE and the deadline scheduling
207 parameters in attr are invalid.
208 EPERM The caller does not have appropriate privileges.
210 EPERM The CPU affinity mask of the thread specified by pid does not
212 include all CPUs in the system (see sched_setaffinity(2)).
213
215 These system calls first appeared in Linux 3.14.
216
218 These system calls are nonstandard Linux extensions.
219
221 sched_setattr() provides a superset of the functionality of
222 sched_setscheduler(2), sched_setparam(2), nice(2), and (other than the
223 ability to set the priority of all processes belonging to a specified
224 user or all processes in a specified group) setpriority(2). Analo‐
225 gously, sched_getattr() provides a superset of the functionality of
226 sched_getscheduler(2), sched_getparam(2), and (partially) getprior‐
227 ity(2).
228
230 In Linux versions up to 3.15, sched_setattr() failed with the error
231 EFAULT instead of E2BIG for the case described in ERRORS.
232 In Linux versions up to 5.3, sched_getattr() failed with the error EF‐
234 BIG if the in-kernel sched_attr structure was larger than the size
235 passed by user space.
236
238 chrt(1), nice(2), sched_get_priority_max(2), sched_get_priority_min(2),
239 sched_getaffinity(2), sched_getparam(2), sched_getscheduler(2),
240 sched_rr_get_interval(2), sched_setaffinity(2), sched_setparam(2),
241 sched_setscheduler(2), sched_yield(2), setpriority(2),
242 pthread_getschedparam(3), pthread_setschedparam(3),
243 pthread_setschedprio(3), capabilities(7), cpuset(7), sched(7)
244
246 This page is part of release 5.10 of the Linux man-pages project. A
247 description of the project, information about reporting bugs, and the
248 latest version of this page, can be found at
249 https://www.kernel.org/doc/man-pages/.
250Linux 2020-11-01 SCHED_SETATTR(2)