real-time scheduler man page

man sche_setscheduler

实时调度的一些概念

SCHED_SETSCHEDULER(2) Linux Programmer's Manual      SCHED_SETSCHEDULER(2)

NAME

       sched_setscheduler, sched_getscheduler - set and get scheduling policy/parameters

SYNOPSIS

       #include <sched.h>


       int sched_setscheduler(pid_t pid, int policy,
     const struct sched_param *param);


       int sched_getscheduler(pid_t pid);


       struct sched_param {
  ...
  int sched_priority;
  ...
       };

DESCRIPTION

       sched_setscheduler() sets both the scheduling policy and the associated parameters for the process whose ID is specified in pid.  If pid equals zero, the scheduling
       policy and parameters of the calling process will be set.  The interpretation of the argument param depends on the selected policy.  Currently, Linux  supports the
       following "normal" (i.e., non-real-time) scheduling policies:


       SCHED_OTHER   the standard round-robin time-sharing policy;


       SCHED_BATCH   for "batch" style execution of processes; and


       SCHED_IDLE    for running very low priority background jobs.


       The following "real-time" policies are also supported, for special time-critical applications that need precise control over the way in which runnable processes are
       selected for execution:


       SCHED_FIFO    a first-in, first-out policy; and


       SCHED_RR      a round-robin policy.


       The semantics of each of these policies are detailed below.


       sched_getscheduler() queries the scheduling policy currently applied to the process identified by pid.  If pid equals zero, the policy of the calling  process  will
       be retrieved.

   Scheduling Policies

       The  scheduler is the kernel component that decides which runnable process will be executed by the CPU next.  Each process has an associated scheduling policy and a
       static scheduling priority, sched_priority; these are the settings that are modified by sched_setscheduler().  The scheduler makes it decisions based  on  knowledge
       of the scheduling policy and static priority of all processes on the system.


       For  processes  scheduled under one of the normal scheduling policies (SCHED_OTHER, SCHED_IDLE, SCHED_BATCH), sched_priority is not used in scheduling decisions (it
       must be specified as 0).


       Processes scheduled under one of the real-time policies (SCHED_FIFO, SCHED_RR) have a sched_priority value in the range 1 (low) to  99 (high).   (As  the  numbers
       imply, real-time processes always have higher priority than normal processes.)  Note well: POSIX.1-2001 only requires an implementation to support a minimum 32 dis‐
       tinct priority levels for the real-time policies, and  some  systems  supply  just  this  minimum.   Portable  programs should use  sched_get_priority_min(2) and
       sched_get_priority_max(2) to find the range of priorities supported for a particular policy.


       Conceptually,  the  scheduler  maintains  a  list  of runnable processes for each possible sched_priority value.  In order to determine which process runs next, the
       scheduler looks for the nonempty list with the highest static priority and selects the process at the head of this list.


       A process's scheduling policy determines where it will be inserted into the list of processes with equal static priority and how it will move inside this list.


       All scheduling is preemptive: if a process with a higher static priority becomes ready to run, the currently running process will be preempted and returned  to the
       wait list for its static priority level.   The scheduling policy only determines the ordering within the list of runnable processes with equal static priority.

   SCHED_FIFO: First In-First Out scheduling

       SCHED_FIFO  can only be used with static priorities higher than 0, which means that when a SCHED_FIFO processes becomes runnable, it will always immediately preempt
       any currently running SCHED_OTHER, SCHED_BATCH, or SCHED_IDLE process.  SCHED_FIFO is a simple scheduling algorithm without time slicing.  For  processes  scheduled
       under the SCHED_FIFO policy, the following rules apply:


       *  A  SCHED_FIFO  process that has been preempted by another process of higher priority will stay at the head of the list for its priority and will resume execution
 as soon as all processes of higher priority are blocked again.


       *  When a SCHED_FIFO process becomes runnable, it will be inserted at the end of the list for its priority.


       *   A call to sched_setscheduler() or sched_setparam(2) will put the SCHED_FIFO (or SCHED_RR) process identified by pid at the start of the list if it was  runnable.
 As a consequence, it may preempt the currently running process if it has the same priority.(会立即抢占当前同优先级的进程)  (POSIX.1-2001 specifies that the process should go to the end of the
 list.)


       *  A process calling sched_yield(2) will be put at the end of the list.


       No other events will move a process scheduled under the SCHED_FIFO policy in the wait list of runnable processes with equal static priority.


       A SCHED_FIFO process runs until either it is blocked by an I/O request, it is preempted by a higher priority process, or it calls sched_yield(2).

   SCHED_RR: Round Robin scheduling

        SCHED_RR is a simple enhancement of SCHED_FIFO. Everything described above for SCHED_FIFO also applies to SCHED_RR, except that each process is only allowed to run
       for  a  maximum time  quantum. If a SCHED_RR process has been running for a time period equal to or longer than the time quantum, it will be put at the end of the
       list for its priority.  A SCHED_RR process that has been preempted by a higher priority process and subsequently resumes execution as a running process  will  com‐
       plete the unexpired portion of its round robin time quantum.  The length of the time quantum can be retrieved using sched_rr_get_interval(2).

   SCHED_OTHER: Default Linux time-sharing scheduling

       SCHED_OTHER  can only be used at static priority 0.  SCHED_OTHER is the standard Linux time-sharing scheduler that is intended for all processes that do not require
       the special real-time mechanisms.  The process to run is chosen from the static priority 0 list based on a dynamic priority that  is  determined  only  inside  this
       list.   The  dynamic  priority  is  based  on the nice value (set by nice(2) or setpriority(2)) and increased for each time quantum the process is ready to run, but
       denied to run by the scheduler. This ensures fair progress among all SCHED_OTHER processes.

   SCHED_BATCH: Scheduling batch processes

       (Since Linux 2.6.16.)  SCHED_BATCH can only be used at static priority 0.  This policy is similar to SCHED_OTHER in that it schedules the process according  to its
       dynamic priority (based on the nice value).  The difference is that this policy will cause the scheduler to always assume that the process is CPU-intensive.  Conse‐
       quently, the scheduler will apply a small scheduling penalty with respect to wakeup behaviour, so that this process is mildly disfavored in scheduling decisions.


       This policy is useful for workloads that are noninteractive, but do not want to lower their nice value, and for workloads that want a deterministic scheduling  pol‐
       icy without interactivity causing extra preemptions (between the workload's tasks).

   SCHED_IDLE: Scheduling very low priority jobs

       (Since Linux 2.6.23.)  SCHED_IDLE can only be used at static priority 0; the process nice value has no influence for this policy.


       This policy is intended for running jobs at extremely low priority (lower even than a +19 nice value with the SCHED_OTHER or SCHED_BATCH policies).


   Resetting scheduling policy for child processes
       Since  Linux  2.6.32, the SCHED_RESET_ON_FORK flag can be ORed in policy when calling sched_setscheduler().  As a result of including this flag, children created by
       fork(2) do not inherit privileged scheduling policies.  This feature is intended for media-playback applications, and can be used to  prevent  applications  evading
       the RLIMIT_RTTIME resource limit (see getrlimit(2)) by creating multiple child processes.


       More precisely, if the SCHED_RESET_ON_FORK flag is specified, the following rules apply for subsequently created children:


       *  If the calling process has a scheduling policy of SCHED_FIFO or SCHED_RR, the policy is reset to SCHED_OTHER in child processes.


       *  If the calling process has a negative nice value, the nice value is reset to zero in child processes.


       After  the SCHED_RESET_ON_FORK flag has been enabled, it can only be reset if the process has the CAP_SYS_NICE capability.  This flag is disabled in child processes
       created by fork(2).


       The SCHED_RESET_ON_FORK flag is visible in the policy value returned by sched_getscheduler()

   Privileges and resource limits

       In Linux kernels before 2.6.12, only privileged (CAP_SYS_NICE) processes can set a nonzero static priority (i.e., set a real-time  scheduling  policy).   The  only
       change  that  an unprivileged process can make is to set the SCHED_OTHER policy, and this can only be done if the effective user ID of the caller of sched_setsched‐
       uler() matches the real or effective user ID of the target process (i.e., the process specified by pid) whose policy is being changed.


       Since Linux 2.6.12, the RLIMIT_RTPRIO resource limit defines a ceiling on an unprivileged process's static priority for the SCHED_RR and SCHED_FIFO  policies. The
       rules for changing scheduling policy and priority are as follows:


       *  If  an  unprivileged process  has a nonzero RLIMIT_RTPRIO soft limit, then it can change its scheduling policy and priority, subject to the restriction that the
 priority cannot be set to a value higher than the maximum of its current priority and its RLIMIT_RTPRIO soft limit.


       *  If the RLIMIT_RTPRIO soft limit is 0, then the only permitted changes are to lower the priority, or to switch to a non-real-time policy.


       *  Subject to the same rules, another unprivileged process can also make these changes, as long as the effective user ID of the process making  the  change  matches
 the real or effective user ID of the target process.


       *  Special rules apply for the SCHED_IDLE.  In Linux kernels before 2.6.39, an unprivileged process operating under this policy cannot change its policy, regardless
 of the value of its RLIMIT_RTPRIO resource limit.  In Linux kernels since 2.6.39, an unprivileged process can switch to either the SCHED_BATCH or the  SCHED_NOR‐
 MAL policy so long as its nice value falls within the range permitted by its RLIMIT_NICE resource limit (see getrlimit(2)).


       Privileged  (CAP_SYS_NICE)  processes ignore the RLIMIT_RTPRIO limit; as with older kernels, they can make arbitrary changes to scheduling policy and priority. See
       getrlimit(2) for further information on RLIMIT_RTPRIO.

   Response time

       A blocked high priority process waiting for the I/O has a certain response time before it is scheduled again.  The device driver  writer  can  greatly  reduce  this
       response time by using a "slow interrupt" interrupt handler.

   Miscellaneous

       Child processes inherit the scheduling policy and parameters across a fork(2).  The scheduling policy and parameters are preserved across execve(2).


       Memory locking is usually needed for real-time processes to avoid paging delays; this can be done with mlock(2) or mlockall(2).


       Since  a  nonblocking  infinite loop in a process scheduled under SCHED_FIFO or SCHED_RR will block all processes with lower priority forever, a software developer
       should always keep available on the console a shell scheduled under a higher static priority than the tested application.  This will  allow  an emergency  kill  of
       tested real-time applications that do not block or terminate as expected.  See also the description of the RLIMIT_RTTIME resource limit in getrlimit(2).


       POSIX systems on which sched_setscheduler() and sched_getscheduler() are available define _POSIX_PRIORITY_SCHEDULING in <unistd.h>.

RETURN VALUE

       On  success,  sched_setscheduler()  returns  zero.   On success,  sched_getscheduler() returns the policy for the process (a nonnegative integer).  On error, -1 is
       returned, and errno is set appropriately.

ERRORS

       EINVAL The scheduling policy is not one of the recognized policies, param is NULL, or param does not make sense for the policy.


       EPERM  The calling process does not have appropriate privileges.


       ESRCH  The process whose ID is pid could not be found.

CONFORMING TO

       POSIX.1-2001 (but see BUGS below).  The SCHED_BATCH and SCHED_IDLE policies are Linux-specific.

NOTES

       POSIX.1 does not detail the permissions that an unprivileged process requires in order to call sched_setscheduler(), and details vary across systems.  For  example,
       the Solaris 7 manual page says that the real or effective user ID of the calling process must match the real user ID or the save set-user-ID of the target process.


       Originally,  Standard Linux was intended as a general-purpose operating system being able to handle background processes, interactive applications, and less demand‐
       ing real-time applications (applications that need to usually meet timing deadlines).  Although the Linux kernel 2.6 allowed for kernel preemption  and  the  newly
       introduced  O(1)  scheduler ensures that the time needed to schedule is fixed and deterministic irrespective of the number of active tasks, true real-time computing
       was not possible up to kernel version 2.6.17.

   Real-time features in the mainline Linux kernel

       From kernel version 2.6.18 onward, however, Linux is gradually becoming equipped with real-time capabilities, most of which are derived from  the  former  realtime-
       preempt patches developed by Ingo Molnar, Thomas Gleixner, Steven Rostedt, and others.  Until the patches have been completely merged into the mainline kernel (this
       is expected to be around kernel version 2.6.30), they must be installed to achieve the best real-time performance.  These patches are named:


  patch-kernelversion-rtpatchversion


       and can be downloaded from ⟨http://www.kernel.org/pub/linux/kernel/projects/rt/⟩.


       Without the patches and prior to their full inclusion into the mainline kernel, the kernel configuration  offers  only  the  three  preemption  classes CONFIG_PRE‐
       EMPT_NONE,  CONFIG_PREEMPT_VOLUNTARY,  and  CONFIG_PREEMPT_DESKTOP  which  respectively provide  no,  some, and considerable reduction of the worst-case scheduling
       latency.


       With the patches applied or after their full inclusion into the mainline kernel, the additional configuration item CONFIG_PREEMPT_RT becomes available. If this  is
       selected,  Linux  is transformed into a regular real-time operating system.  The FIFO and RR scheduling policies that can be selected using sched_setscheduler() are
       then used to run a process with true real-time priority and a minimum worst-case scheduling latency.

BUGS

       POSIX says that on success, sched_setscheduler() should return the previous scheduling policy.  Linux sched_setscheduler() does not  conform  to  this  requirement,
       since it always returns 0 on success.

SEE ALSO

       getpriority(2), mlock(2), mlockall(2), munlock(2), munlockall(2), nice(2), sched_get_priority_max(2), sched_get_priority_min(2), sched_getaffinity(2),
       sched_getparam(2), sched_rr_get_interval(2), sched_setaffinity(2), sched_setparam(2), sched_yield(2), setpriority(2), capabilities(7), cpuset(7)


       Programming for the real world - POSIX.4 by Bill O. Gallmeister, O'Reilly & Associates, Inc., ISBN 1-56592-074-0.


       Documentation/scheduler/sched-rt-group.txt in the Linux kernel source tree (since kernel 2.6.25).

COLOPHON

       This page is part of release 3.44 of the Linux man-pages project.   A  description  of  the  project,  and  information about  reporting  bugs,  can  be  found  at
       http://www.kernel.org/doc/man-pages/.






Linux 2012-08-05      SCHED_SETSCHEDULER(2)