man2/sched_setscheduler.2

   1 .\" Hey Emacs! This file is -*- nroff -*- source.
   2 .\"
   3 .\" Copyright (C) Tom Bjorkholm, Markus Kuhn & David A. Wheeler 1996-1999
   4 .\" and Copyright (C) 2007 Carsten Emde <Carsten.Emde@osadl.org>
   5 .\"
   6 .\" This is free documentation; you can redistribute it and/or
   7 .\" modify it under the terms of the GNU General Public License as
   8 .\" published by the Free Software Foundation; either version 2 of
   9 .\" the License, or (at your option) any later version.
  10 .\"
  11 .\" The GNU General Public License's references to "object code"
  12 .\" and "executables" are to be interpreted as the output of any
  13 .\" document formatting or typesetting system, including
  14 .\" intermediate and printed output.
  15 .\"
  16 .\" This manual is distributed in the hope that it will be useful,
  17 .\" but WITHOUT ANY WARRANTY; without even the implied warranty of
  18 .\" MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  19 .\" GNU General Public License for more details.
  20 .\"
  21 .\" You should have received a copy of the GNU General Public
  22 .\" License along with this manual; if not, write to the Free
  23 .\" Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111,
  24 .\" USA.
  25 .\"
  26 .\" 1996-04-01 Tom Bjorkholm <tomb@mydata.se>
  27 .\"            First version written
  28 .\" 1996-04-10 Markus Kuhn <mskuhn@cip.informatik.uni-erlangen.de>
  29 .\"            revision
  30 .\" 1999-08-18 David A. Wheeler <dwheeler@ida.org> added Note.
  31 .\" Modified, 25 Jun 2002, Michael Kerrisk <mtk.manpages@gmail.com>
  32 .\"     Corrected description of queue placement by sched_setparam() and
  33 .\"             sched_setscheduler()
  34 .\"     A couple of grammar clean-ups
  35 .\" Modified 2004-05-27 by Michael Kerrisk <mtk.manpages@gmail.com>
  36 .\" 2005-03-23, mtk, Added description of SCHED_BATCH.
  37 .\" 2007-07-10, Carsten Emde <Carsten.Emde@osadl.org>
  38 .\"     Add text on real-time features that are currently being
  39 .\"     added to the mainline kernel.
  40 .\"
  41 .TH SCHED_SETSCHEDULER 2 2007-11-25 "Linux" "Linux Programmer's Manual"
  42 .SH NAME
  43 sched_setscheduler, sched_getscheduler \-
  44 set and get scheduling algorithm/parameters
  45 .SH SYNOPSIS
  46 .nf
  47 .B #include <sched.h>
  48 .sp
  49 .BI "int sched_setscheduler(pid_t " pid ", int " policy ,
  50 .br
  51 .BI "                       const struct sched_param *" param );
  52 .sp
  53 .BI "int sched_getscheduler(pid_t " pid );
  54 .sp
  55 \fBstruct sched_param {
  56     ...
  57     int \fIsched_priority\fB;
  58     ...
  59 };
  60 .fi
  61 .SH DESCRIPTION
  62 .BR sched_setscheduler ()
  63 sets both the scheduling policy and the associated parameters for the
  64 process identified by \fIpid\fP.
  65 If \fIpid\fP equals zero, the
  66 scheduler of the calling process will be set.
  67 The interpretation of
  68 the parameter \fIparam\fP depends on the selected policy.
  69 Currently, the
  70 following three scheduling policies are supported under Linux:
  71 .BR SCHED_FIFO ,
  72 .BR SCHED_RR ,
  73 .BR SCHED_OTHER ,
  74 .\" In the 2.6 kernel sources, SCHED_OTHER is actually called
  75 .\" SCHED_NORMAL.
  76 and
  77 .BR SCHED_BATCH ;
  78 their respective semantics are described below.
  79
  80 .BR sched_getscheduler ()
  81 queries the scheduling policy currently applied to the process
  82 identified by \fIpid\fP.
  83 If \fIpid\fP equals zero, the policy of the
  84 calling process will be retrieved.
  85 .SS Scheduling Policies
  86 The scheduler is the kernel part that decides which runnable process
  87 will be executed by the CPU next.
  88 The Linux scheduler offers three
  89 different scheduling policies, one for normal processes and two for
  90 real-time applications.
  91 A static priority value \fIsched_priority\fP
  92 is assigned to each process and this value can be changed only via
  93 system calls.
  94 Conceptually, the scheduler maintains a list of runnable
  95 processes for each possible \fIsched_priority\fP value, and
  96 \fIsched_priority\fP can have a value in the range 0 to 99.
  97 In order
  98 to determine the process that runs next, the Linux scheduler looks for
  99 the non-empty list with the highest static priority and takes the
 100 process at the head of this list.
 101 The scheduling policy determines for
 102 each process, where it will be inserted into the list of processes
 103 with equal static priority and how it will move inside this list.
 104
 105 \fBSCHED_OTHER\fP is the default universal time-sharing scheduler
 106 policy used by most processes.
 107 \fBSCHED_BATCH\fP is intended for "batch" style execution of processes.
 108 \fBSCHED_FIFO\fP and \fBSCHED_RR\fP are
 109 intended for special time-critical applications that need precise
 110 control over the way in which runnable processes are selected for
 111 execution.
 112
 113 Processes scheduled with \fBSCHED_OTHER\fP or \fBSCHED_BATCH\fP
 114 must be assigned the static priority 0.
 115 Processes scheduled under \fBSCHED_FIFO\fP or
 116 \fBSCHED_RR\fP can have a static priority in the range 1 to 99.
 117 The system calls
 118 .BR sched_get_priority_min (2)
 119 and
 120 .BR sched_get_priority_max (2)
 121 can be used to find out the valid
 122 priority range for a scheduling policy in a portable way on all
 123 POSIX.1-2001 conforming systems.
 124
 125 All scheduling is preemptive: If a process with a higher static
 126 priority gets ready to run, the current process will be preempted and
 127 returned into its wait list.
 128 The scheduling policy only determines the
 129 ordering within the list of runnable processes with equal static
 130 priority.
 131 .SS SCHED_FIFO: First In-First Out scheduling
 132 \fBSCHED_FIFO\fP can only be used with static priorities higher than
 133 0, which means that when a \fBSCHED_FIFO\fP processes becomes runnable,
 134 it will always immediately preempt any currently running
 135 \fBSCHED_OTHER\fP or \fBSCHED_BATCH\fP process.
 136 \fBSCHED_FIFO\fP is a simple scheduling
 137 algorithm without time slicing.
 138 For processes scheduled under the
 139 \fBSCHED_FIFO\fP policy, the following rules are applied: A
 140 \fBSCHED_FIFO\fP process that has been preempted by another process of
 141 higher priority will stay at the head of the list for its priority and
 142 will resume execution as soon as all processes of higher priority are
 143 blocked again.
 144 When a \fBSCHED_FIFO\fP process becomes runnable, it
 145 will be inserted at the end of the list for its priority.
 146 A call to
 147 .BR sched_setscheduler ()
 148 or
 149 .BR sched_setparam (2)
 150 will put the
 151 \fBSCHED_FIFO\fP (or \fBSCHED_RR\fP) process identified by
 152 \fIpid\fP at the start of the list if it was runnable.
 153 As a consequence, it may preempt the currently running process if
 154 it has the same priority.
 155 (POSIX.1-2001 specifies that the process should go to the end
 156 of the list.)
 157 .\" In 2.2.x and 2.4.x, the process is placed at the front of the queue
 158 .\" In 2.0.x, the Right Thing happened: the process went to the back -- MTK
 159 A process calling
 160 .BR sched_yield (2)
 161 will be
 162 put at the end of the list.
 163 No other events will move a process
 164 scheduled under the \fBSCHED_FIFO\fP policy in the wait list of
 165 runnable processes with equal static priority.
 166 A \fBSCHED_FIFO\fP
 167 process runs until either it is blocked by an I/O request, it is
 168 preempted by a higher priority process, or it calls
 169 .BR sched_yield (2).
 170 .SS SCHED_RR: Round Robin scheduling
 171 \fBSCHED_RR\fP is a simple enhancement of \fBSCHED_FIFO\fP.
 172 Everything
 173 described above for \fBSCHED_FIFO\fP also applies to \fBSCHED_RR\fP,
 174 except that each process is only allowed to run for a maximum time
 175 quantum.
 176 If a \fBSCHED_RR\fP process has been running for a time
 177 period equal to or longer than the time quantum, it will be put at the
 178 end of the list for its priority.
 179 A \fBSCHED_RR\fP process that has
 180 been preempted by a higher priority process and subsequently resumes
 181 execution as a running process will complete the unexpired portion of
 182 its round robin time quantum.
 183 The length of the time quantum can be
 184 retrieved using
 185 .BR sched_rr_get_interval (2).
 186 .\" On Linux 2.4, the length of the RR interval is influenced
 187 .\" by the process nice value -- MTK
 188 .\"
 189 .SS SCHED_OTHER: Default Linux time-sharing scheduling
 190 \fBSCHED_OTHER\fP can only be used at static priority 0.
 191 \fBSCHED_OTHER\fP is the standard Linux time-sharing scheduler that is
 192 intended for all processes that do not require special static priority
 193 real-time mechanisms.
 194 The process to run is chosen from the static
 195 priority 0 list based on a dynamic priority that is determined only
 196 inside this list.
 197 The dynamic priority is based on the nice level (set
 198 by
 199 .BR nice (2)
 200 or
 201 .BR setpriority (2))
 202 and increased for
 203 each time quantum the process is ready to run, but denied to run by
 204 the scheduler.
 205 This ensures fair progress among all \fBSCHED_OTHER\fP
 206 processes.
 207 .SS SCHED_BATCH: Scheduling batch processes
 208 (Since Linux 2.6.16.)
 209 \fBSCHED_BATCH\fP can only be used at static priority 0.
 210 This policy is similar to \fBSCHED_OTHER\fP, except that
 211 this policy will cause the scheduler to always assume
 212 that the process is CPU-intensive.
 213 Consequently, the scheduler will apply a small scheduling
 214 penalty so that this process is mildly disfavored in scheduling
 215 decisions.
 216 .\" The following paragraph is drawn largely from the text that
 217 .\" accompanied Ingo Molnar's patch for the implementation of
 218 .\" SCHED_BATCH.
 219 This policy is useful for workloads that are non-interactive,
 220 but do not want to lower their nice value,
 221 and for workloads that want a deterministic scheduling policy without
 222 interactivity causing extra preemptions (between the workload's tasks).
 223 .SS Privileges and resource limits
 224 In Linux kernels before 2.6.12, only privileged
 225 .RB ( CAP_SYS_NICE )
 226 processes can set a non-zero static priority.
 227 The only change that an unprivileged process can make is to set the
 228 .B SCHED_OTHER
 229 policy, and this can only be done if the effective user ID of the caller of
 230 .BR sched_setscheduler ()
 231 matches the real or effective user ID of the target process
 232 (i.e., the process specified by
 233 .IR pid )
 234 whose policy is being changed.
 235
 236 Since Linux 2.6.12, the
 237 .B RLIMIT_RTPRIO
 238 resource limit defines a ceiling on an unprivileged process's
 239 priority for the
 240 .B SCHED_RR
 241 and
 242 .B SCHED_FIFO
 243 policies.
 244 If an unprivileged process has a non-zero
 245 .B RLIMIT_RTPRIO
 246 soft limit, then it can change its scheduling policy and priority,
 247 subject to the restriction that the priority cannot be set to a
 248 value higher than the
 249 .B RLIMIT_RTPRIO
 250 soft limit.
 251 If the
 252 .B RLIMIT_RTPRIO
 253 soft limit is 0, then the only permitted change is to lower the priority.
 254 Subject to the same rules,
 255 another unprivileged process can also make these changes,
 256 as long as the effective user ID of the process making the change
 257 matches the real or effective user ID of the target process.
 258 See
 259 .BR getrlimit (2)
 260 for further information on
 261 .BR RLIMIT_RTPRIO .
 262 Privileged
 263 .RB ( CAP_SYS_NICE )
 264 processes ignore this limit; as with older kernels,
 265 they can make arbitrary changes to scheduling policy and priority.
 266 .SS Response time
 267 A blocked high priority process waiting for the I/O has a certain
 268 response time before it is scheduled again.
 269 The device driver writer
 270 can greatly reduce this response time by using a "slow interrupt"
 271 interrupt handler.
 272 .\" as described in
 273 .\" .BR request_irq (9).
 274 .SS Miscellaneous
 275 Child processes inherit the scheduling algorithm and parameters across a
 276 .BR fork (2).
 277 The scheduling algorithm and parameters are preserved across
 278 .BR execve (2).
 279
 280 Memory locking is usually needed for real-time processes to avoid
 281 paging delays, this can be done with
 282 .BR mlock (2)
 283 or
 284 .BR mlockall (2).
 285
 286 As a non-blocking end-less loop in a process scheduled under
 287 \fBSCHED_FIFO\fP or \fBSCHED_RR\fP will block all processes with lower
 288 priority forever, a software developer should always keep available on
 289 the console a shell scheduled under a higher static priority than the
 290 tested application.
 291 This will allow an emergency kill of tested
 292 real-time applications that do not block or terminate as expected.
 293
 294 POSIX systems on which
 295 .BR sched_setscheduler ()
 296 and
 297 .BR sched_getscheduler ()
 298 are available define
 299 .B _POSIX_PRIORITY_SCHEDULING
 300 in \fI<unistd.h>\fP.
 301 .SH "RETURN VALUE"
 302 On success,
 303 .BR sched_setscheduler ()
 304 returns zero.
 305 On success,
 306 .BR sched_getscheduler ()
 307 returns the policy for the process (a non-negative integer).
 308 On error, \-1 is returned, and
 309 .I errno
 310 is set appropriately.
 311 .SH ERRORS
 312 .TP
 313 .B EINVAL
 314 The scheduling \fIpolicy\fP is not one of the recognized policies,
 315 or the parameter \fIparam\fP does not make sense for the \fIpolicy\fP.
 316 .TP
 317 .B EPERM
 318 The calling process does not have appropriate privileges.
 319 .TP
 320 .B ESRCH
 321 The process whose ID is \fIpid\fP could not be found.
 322 .SH "CONFORMING TO"
 323 POSIX.1-2001 (but see BUGS below).
 324 The \fBSCHED_BATCH\fP policy is Linux specific.
 325 .SH NOTES
 326 POSIX.1 does not detail the permissions that an unprivileged
 327 process requires in order to call
 328 .BR sched_setscheduler (),
 329 and details vary across systems.
 330 For example, the Solaris 7 manual page says that
 331 the real or effective user ID of the calling process must
 332 match the real user ID or the save set-user-ID of the target process.
 333 .PP
 334 Originally, Standard Linux was intended as a general-purpose operating
 335 system being able to handle background processes, interactive
 336 applications, and less demanding real-time applications (applications that
 337 need to usually meet timing deadlines).
 338 Although the Linux kernel 2.6
 339 allowed for kernel preemption and the newly introduced O(1) scheduler
 340 ensures that the time needed to schedule is fixed and deterministic
 341 irrespective of the number of active tasks, true real-time computing
 342 was not possible up to kernel version 2.6.17.
 343 .SS Real-time features in the mainline Linux kernel
 344 .\" FIXME . Probably this text will need some minor tweaking
 345 .\" by about the time of 2.6.25; ask Carsten Emde about this then.
 346 From kernel version 2.6.18 onwards, however, Linux is gradually
 347 becoming equipped with real-time capabilities,
 348 most of which are derived from the former
 349 realtime-preempt patches developed by Ingo Molnar, Thomas Gleixner and
 350 others.
 351 Until the patches have been completely merged into the
 352 mainline kernel
 353 (this is expected to be around kernel version 2.6.24 or 2.6.25),
 354 the realtime-preempt patches must be installed to achieve the best
 355 realtime performance.
 356 These patches are named:
 357 .in +4n
 358 .nf
 359
 360 patch-\fIkernelversion\fP-rt\fIpatchversion\fP
 361 .fi
 362 .in
 363 .PP
 364 and can be downloaded from
 365 .IR http://people.redhat.com/mingo/realtime-preempt/ .
 366
 367 Without the patches and prior to their full inclusion into the mainline
 368 kernel, the kernel configuration offers only the three preemption classes
 369 .BR CONFIG_PREEMPT_NONE ,
 370 .BR CONFIG_PREEMPT_VOLUNTARY ,
 371 and
 372 .B CONFIG_PREEMPT_DESKTOP
 373 which respectively provide no, some, and considerable
 374 reduction of the worst-case scheduling latency.
 375
 376 With the patches applied or after their full inclusion into the mainline
 377 kernel, the additional configuration item
 378 .B CONFIG_PREEMPT_RT
 379 becomes available.
 380 If this is selected, Linux is transformed into a regular
 381 real-time operating system.
 382 The FIFO and RR scheduling policies that can be selected using
 383 .BR sched_setscheduler ()
 384 are then used to run a process
 385 with true real-time priority and a minimum worst-case scheduling latency.
 386 .SH BUGS
 387 POSIX says that on success,
 388 .BR sched_setscheduler ()
 389 should return the previous scheduling policy.
 390 Linux
 391 .BR sched_setscheduler ()
 392 does not conform to this requirement,
 393 since it always returns 0 on success.
 394 .SH "SEE ALSO"
 395 .BR getpriority (2),
 396 .BR mlock (2),
 397 .BR mlockall (2),
 398 .BR munlock (2),
 399 .BR munlockall (2),
 400 .BR nice (2),
 401 .BR sched_get_priority_max (2),
 402 .BR sched_get_priority_min (2),
 403 .BR sched_getaffinity (2),
 404 .BR sched_getparam (2),
 405 .BR sched_rr_get_interval (2),
 406 .BR sched_setaffinity (2),
 407 .BR sched_setparam (2),
 408 .BR sched_yield (2),
 409 .BR setpriority (2),
 410 .BR capabilities (7)
 411 .PP
 412 .I Programming for the real world \- POSIX.4
 413 by Bill O. Gallmeister, O'Reilly & Associates, Inc., ISBN 1-56592-074-0