.\"
.\" Worth looking at: http://rt.wiki.kernel.org/index.php
.\"
-.TH SCHED 7 2016-10-08 "Linux" "Linux Programmer's Manual"
+.TH SCHED 7 2019-08-02 "Linux" "Linux Programmer's Manual"
.SH NAME
sched \- overview of CPU scheduling
.SH DESCRIPTION
.IR sched_priority .
The scheduler makes its decisions based on knowledge of the scheduling
policy and static priority of all threads on the system.
-
+.PP
For threads scheduled under one of the normal scheduling policies
(\fBSCHED_OTHER\fP, \fBSCHED_IDLE\fP, \fBSCHED_BATCH\fP),
\fIsched_priority\fP is not used in scheduling
decisions (it must be specified as 0).
-
+.PP
Processes scheduled under one of the real-time policies
(\fBSCHED_FIFO\fP, \fBSCHED_RR\fP) have a
\fIsched_priority\fP value in the range 1 (low) to 99 (high).
and
.BR sched_get_priority_max (2)
to find the range of priorities supported for a particular policy.
-
+.PP
Conceptually, the scheduler maintains a list of runnable
threads for each possible \fIsched_priority\fP value.
In order to determine which thread runs next, the scheduler looks for
the nonempty list with the highest static priority and selects the
thread at the head of this list.
-
+.PP
A thread's scheduling policy determines
where it will be inserted into the list of threads
with equal static priority and how it will move inside this list.
-
+.PP
All scheduling is preemptive: if a thread with a higher static
priority becomes ready to run, the currently running thread
will be preempted and
priority.
.SS SCHED_FIFO: First in-first out scheduling
\fBSCHED_FIFO\fP can be used only with static priorities higher than
-0, which means that when a \fBSCHED_FIFO\fP threads becomes runnable,
+0, which means that when a \fBSCHED_FIFO\fP thread becomes runnable,
it will always immediately preempt any currently running
\fBSCHED_OTHER\fP, \fBSCHED_BATCH\fP, or \fBSCHED_IDLE\fP thread.
\fBSCHED_FIFO\fP is a simple scheduling
algorithm without time slicing.
For threads scheduled under the
\fBSCHED_FIFO\fP policy, the following rules apply:
-.IP * 3
-A \fBSCHED_FIFO\fP thread that has been preempted by another thread of
+.IP 1) 3
+A running \fBSCHED_FIFO\fP thread that has been preempted by another thread of
higher priority will stay at the head of the list for its priority and
will resume execution as soon as all threads of higher priority are
blocked again.
-.IP *
-When a \fBSCHED_FIFO\fP thread becomes runnable, it
+.IP 2)
+When a blocked \fBSCHED_FIFO\fP thread becomes runnable, it
will be inserted at the end of the list for its priority.
-.IP *
-A call to
+.IP 3)
+If a call to
.BR sched_setscheduler (2),
.BR sched_setparam (2),
+.BR sched_setattr (2),
+.BR pthread_setschedparam (3),
or
-.BR sched_setattr (2)
-will put the
-\fBSCHED_FIFO\fP (or \fBSCHED_RR\fP) thread identified by
-\fIpid\fP at the start of the list if it was runnable.
-As a consequence, it may preempt the currently running thread if
-it has the same priority.
-(POSIX.1 specifies that the thread should go to the end
-of the list.)
+.BR pthread_setschedprio (3)
+changes the priority of the running or runnable
+.B SCHED_FIFO
+thread identified by
+.I pid
+the effect on the thread's position in the list depends on
+the direction of the change to threads priority:
+.RS
+.IP \(bu 3
+If the thread's priority is raised,
+it is placed at the end of the list for its new priority.
+As a consequence,
+it may preempt a currently running thread with the same priority.
+.IP \(bu
+If the thread's priority is unchanged,
+its position in the run list is unchanged.
+.IP \(bu
+If the thread's priority is lowered,
+it is placed at the front of the list for its new priority.
+.RE
+.IP
+According to POSIX.1-2008,
+changes to a thread's priority (or policy) using any mechanism other than
+.BR pthread_setschedprio (3)
+should result in the thread being placed at the end of
+the list for its priority.
.\" In 2.2.x and 2.4.x, the thread is placed at the front of the queue
.\" In 2.0.x, the Right Thing happened: the thread went to the back -- MTK
-.IP *
+.IP 4)
A thread calling
.BR sched_yield (2)
will be put at the end of the list.
No other events will move a thread
scheduled under the \fBSCHED_FIFO\fP policy in the wait list of
runnable threads with equal static priority.
-
+.PP
A \fBSCHED_FIFO\fP
thread runs until either it is blocked by an I/O request, it is
preempted by a higher priority thread, or it calls
and
.BR sched_getattr (2)
system calls.
-
+.PP
A sporadic task is one that has a sequence of jobs, where each
job is activated at most once per period.
Each job also has a
The
.I "absolute deadline"
is thus obtained by adding the relative deadline to the arrival time.
-
+.PP
The following diagram clarifies these terms:
-
+.PP
.in +4n
-.nf
+.EX
arrival/wakeup absolute deadline
| start time |
| | |
|<- comp. time ->|
|<------- relative deadline ------>|
|<-------------- period ------------------->|
-.fi
+.EE
.in
-
+.PP
When setting a
.B SCHED_DEADLINE
policy for a thread using
Thus, for
.BR SCHED_DEADLINE
scheduling, we have:
-
+.PP
.in +4n
-.nf
+.EX
arrival/wakeup absolute deadline
| start time |
| | |
|<-- Runtime ------->|
|<----------- Deadline ----------->|
|<-------------- Period ------------------->|
-.fi
+.EE
.in
-
+.PP
The three deadline-scheduling parameters correspond to the
.IR sched_runtime ,
.IR sched_deadline ,
.IR sched_period
is specified as 0, then it is made the same as
.IR sched_deadline .
-
+.PP
The kernel requires that:
-
+.PP
sched_runtime <= sched_deadline <= sched_period
-
+.PP
.\" See __checkparam_dl in kernel/sched/core.c
In addition, under the current implementation,
all of the parameter values must be at least 1024
.BR sched_setattr (2)
fails with the error
.BR EINVAL .
-
+.PP
The CBS guarantees non-interference between tasks, by throttling
threads that attempt to over-run their specified Runtime.
-
+.PP
To ensure deadline scheduling guarantees,
the kernel must prevent situations where the set of
.B SCHED_DEADLINE
.BR sched_setattr (2)
fails with the error
.BR EBUSY .
-
+.PP
For example, it is required (but not necessarily sufficient) for
the total utilization to be less than or equal to the total number of
CPUs available, where, since each thread can maximally run for
Runtime per Period, that thread's utilization is its
Runtime divided by its Period.
-
+.PP
In order to fulfill the guarantees that are made when
a thread is admitted to the
.BR SCHED_DEADLINE
.BR SCHED_DEADLINE
thread is runnable,
it will preempt any thread scheduled under one of the other policies.
-
+.PP
A call to
.BR fork (2)
by a thread scheduled under the
.B SCHED_DEADLINE
-policy will fail with the error
+policy fails with the error
.BR EAGAIN ,
unless the thread has its reset-on-fork flag set (see below).
-
+.PP
A
.B SCHED_DEADLINE
thread that calls
\fBSCHED_OTHER\fP is the standard Linux time-sharing scheduler that is
intended for all threads that do not require the special
real-time mechanisms.
-
+.PP
The thread to run is chosen from the static
priority 0 list based on a \fIdynamic\fP priority that is determined only
inside this list.
and is increased for each time quantum the thread is ready to run,
but denied to run by the scheduler.
This ensures fair progress among all \fBSCHED_OTHER\fP threads.
+.PP
+In the Linux kernel source code, the
+.B SCHED_OTHER
+policy is actually named
+.BR SCHED_NORMAL .
.\"
.SS The nice value
The nice value is an attribute
.BR setpriority (2),
or
.BR sched_setattr (2).
-
+.PP
According to POSIX.1, the nice value is a per-process attribute;
that is, the threads in a process should share a nice value.
However, on Linux, the nice value is a per-thread attribute:
different threads in the same process may have different nice values.
-
+.PP
The range of the nice value
varies across UNIX systems.
On modern Linux, the range is \-20 (high priority) to +19 (low priority).
Very early Linux kernels (Before Linux 2.0) had the range \-infinity..15.
.\" Linux before 1.3.36 had \-infinity..15.
.\" Since kernel 1.3.43, Linux has the range \-20..19.
-
+.PP
The degree to which the nice value affects the relative scheduling of
.BR SCHED_OTHER
processes likewise varies across UNIX systems and
across Linux kernel versions.
-
+.PP
With the advent of the CFS scheduler in kernel 2.6.23,
Linux adopted an algorithm that causes
relative differences in nice values to have a much stronger effect.
higher priority load on the system,
and makes high nice values (\-20) deliver most of the CPU to applications
that require it (e.g., some audio applications).
-
+.PP
On Linux, the
.BR RLIMIT_NICE
resource limit can be used to define a limit to which
an unprivileged process's nice value can be raised; see
.BR setrlimit (2)
for details.
+.PP
+For further details on the nice value, see the subsections on
+the autogroup feature and group scheduling, below.
.\"
.SS SCHED_BATCH: Scheduling batch processes
(Since Linux 2.6.16.)
Consequently, the scheduler will apply a small scheduling
penalty with respect to wakeup behavior,
so that this thread is mildly disfavored in scheduling decisions.
-
+.PP
.\" The following paragraph is drawn largely from the text that
.\" accompanied Ingo Molnar's patch for the implementation of
.\" SCHED_BATCH.
(Since Linux 2.6.23.)
\fBSCHED_IDLE\fP can be used only at static priority 0;
the process nice value has no influence for this policy.
-
+.PP
This policy is intended for running jobs at extremely low
priority (lower even than a +19 nice value with the
.B SCHED_OTHER
.BR sched_getscheduler (2)
and
.BR sched_getattr (2).
-
+.PP
The reset-on-fork feature is intended for media-playback applications,
and can be used to prevent applications evading the
.BR RLIMIT_RTTIME
resource limit (see
.BR getrlimit (2))
by creating multiple child processes.
-
+.PP
More precisely, if the reset-on-fork flag is set,
the following rules apply for subsequently created children:
.IP * 3
(i.e., the thread specified by
.IR pid )
whose policy is being changed.
-
+.PP
A thread must be privileged
.RB ( CAP_SYS_NICE )
in order to set or modify a
.BR SCHED_DEADLINE
policy.
-
+.PP
Since Linux 2.6.12, the
.B RLIMIT_RTPRIO
resource limit defines a ceiling on an unprivileged thread's
a shell scheduled under a higher static priority than the tested application.
This allows an emergency kill of tested
real-time applications that do not block or terminate as expected.
-
+.PP
Since Linux 2.6.25, there are other techniques for dealing with runaway
real-time and deadline processes.
One of these is to use the
See
.BR getrlimit (2)
for details.
-
+.PP
Since version 2.6.25, Linux also provides two
.I /proc
files that can be used to reserve a certain amount of CPU time
on the system.
The value in this file can range from \-1 to
.BR INT_MAX \-1.
-Specifying \-1 makes the runtime the same as the period;
+Specifying \-1 makes the run time the same as the period;
that is, no CPU time is set aside for non-real-time processes
(which was the Linux behavior before kernel 2.6.25).
The default value in this file is 950,000 (0.95 seconds),
.BR fork (2).
The scheduling policy and parameters are preserved across
.BR execve (2).
-
+.PP
Memory locking is usually needed for real-time processes to avoid
paging delays; this can be done with
.BR mlock (2)
or
.BR mlockall (2).
-.SH NOTES
-The
+.\"
+.SS The autogroup feature
+.\" commit 5091faa449ee0b7d73bc296a93bca9540fc51d0a
+Since Linux 2.6.38,
+the kernel provides a feature known as autogrouping to improve interactive
+desktop performance in the face of multiprocess, CPU-intensive
+workloads such as building the Linux kernel with large numbers of
+parallel build processes (i.e., the
+.BR make (1)
+.BR \-j
+flag).
+.PP
+This feature operates in conjunction with the
+CFS scheduler and requires a kernel that is configured with
+.BR CONFIG_SCHED_AUTOGROUP .
+On a running system, this feature is enabled or disabled via the file
+.IR /proc/sys/kernel/sched_autogroup_enabled ;
+a value of 0 disables the feature, while a value of 1 enables it.
+The default value in this file is 1, unless the kernel was booted with the
+.IR noautogroup
+parameter.
+.PP
+A new autogroup is created when a new session is created via
+.BR setsid (2);
+this happens, for example, when a new terminal window is started.
+A new process created by
+.BR fork (2)
+inherits its parent's autogroup membership.
+Thus, all of the processes in a session are members of the same autogroup.
+An autogroup is automatically destroyed when the last process
+in the group terminates.
+.PP
+When autogrouping is enabled, all of the members of an autogroup
+are placed in the same kernel scheduler "task group".
+The CFS scheduler employs an algorithm that equalizes the
+distribution of CPU cycles across task groups.
+The benefits of this for interactive desktop performance
+can be described via the following example.
+.PP
+Suppose that there are two autogroups competing for the same CPU
+(i.e., presume either a single CPU system or the use of
+.BR taskset (1)
+to confine all the processes to the same CPU on an SMP system).
+The first group contains ten CPU-bound processes from
+a kernel build started with
+.IR "make\ \-j10" .
+The other contains a single CPU-bound process: a video player.
+The effect of autogrouping is that the two groups will
+each receive half of the CPU cycles.
+That is, the video player will receive 50% of the CPU cycles,
+rather than just 9% of the cycles,
+which would likely lead to degraded video playback.
+The situation on an SMP system is more complex,
+.\" Mike Galbraith, 25 Nov 2016:
+.\" I'd say something more wishy-washy here, like cycles are
+.\" distributed fairly across groups and leave it at that, as your
+.\" detailed example is incorrect due to SMP fairness (which I don't
+.\" like much because [very unlikely] worst case scenario
+.\" renders a box sized group incapable of utilizing more that
+.\" a single CPU total). For example, if a group of NR_CPUS
+.\" size competes with a singleton, load balancing will try to give
+.\" the singleton a full CPU of its very own. If groups intersect for
+.\" whatever reason on say my quad lappy, distribution is 80/20 in
+.\" favor of the singleton.
+but the general effect is the same:
+the scheduler distributes CPU cycles across task groups such that
+an autogroup that contains a large number of CPU-bound processes
+does not end up hogging CPU cycles at the expense of the other
+jobs on the system.
+.PP
+A process's autogroup (task group) membership can be viewed via the file
+.IR /proc/[pid]/autogroup :
+.PP
+.in +4n
+.EX
+$ \fBcat /proc/1/autogroup\fP
+/autogroup-1 nice 0
+.EE
+.in
+.PP
+This file can also be used to modify the CPU bandwidth allocated
+to an autogroup.
+This is done by writing a number in the "nice" range to the file
+to set the autogroup's nice value.
+The allowed range is from +19 (low priority) to \-20 (high priority).
+(Writing values outside of this range causes
+.BR write (2)
+to fail with the error
+.BR EINVAL .)
+.\" FIXME .
+.\" Because of a bug introduced in Linux 4.7
+.\" (commit 2159197d66770ec01f75c93fb11dc66df81fd45b made changes
+.\" that exposed the fact that autogroup didn't call scale_load()),
+.\" it happened that *all* values in this range caused a task group
+.\" to be further disfavored by the scheduler, with \-20 resulting
+.\" in the scheduler mildly disfavoring the task group and +19 greatly
+.\" disfavoring it.
+.\"
+.\" A patch was posted on 23 Nov 2016
+.\" ("sched/autogroup: Fix 64bit kernel nice adjustment";
+.\" check later to see in which kernel version it lands.
+.PP
+The autogroup nice setting has the same meaning as the process nice value,
+but applies to distribution of CPU cycles to the autogroup as a whole,
+based on the relative nice values of other autogroups.
+For a process inside an autogroup, the CPU cycles that it receives
+will be a product of the autogroup's nice value
+(compared to other autogroups)
+and the process's nice value
+(compared to other processes in the same autogroup.
+.PP
+The use of the
.BR cgroups (7)
-CPU controller can be used to limit the CPU consumption of
-groups of processes.
+CPU controller to place processes in cgroups other than the
+root CPU cgroup overrides the effect of autogrouping.
.PP
-Originally, Standard Linux was intended as a general-purpose operating
-system being able to handle background processes, interactive
-applications, and less demanding 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.
+The autogroup feature groups only processes scheduled under
+non-real-time policies
+.RB ( SCHED_OTHER ,
+.BR SCHED_BATCH ,
+and
+.BR SCHED_IDLE ).
+It does not group processes scheduled under real-time and
+deadline policies.
+Those processes are scheduled according to the rules described earlier.
+.\"
+.SS The nice value and group scheduling
+When scheduling non-real-time processes (i.e., those scheduled under the
+.BR SCHED_OTHER ,
+.BR SCHED_BATCH ,
+and
+.BR SCHED_IDLE
+policies), the CFS scheduler employs a technique known as "group scheduling",
+if the kernel was configured with the
+.BR CONFIG_FAIR_GROUP_SCHED
+option (which is typical).
+.PP
+Under group scheduling, threads are scheduled in "task groups".
+Task groups have a hierarchical relationship,
+rooted under the initial task group on the system,
+known as the "root task group".
+Task groups are formed in the following circumstances:
+.IP * 3
+All of the threads in a CPU cgroup form a task group.
+The parent of this task group is the task group of the
+corresponding parent cgroup.
+.IP *
+If autogrouping is enabled,
+then all of the threads that are (implicitly) placed in an autogroup
+(i.e., the same session, as created by
+.BR setsid (2))
+form a task group.
+Each new autogroup is thus a separate task group.
+The root task group is the parent of all such autogroups.
+.IP *
+If autogrouping is enabled, then the root task group consists of
+all processes in the root CPU cgroup that were not
+otherwise implicitly placed into a new autogroup.
+.IP *
+If autogrouping is disabled, then the root task group consists of
+all processes in the root CPU cgroup.
+.IP *
+If group scheduling was disabled (i.e., the kernel was configured without
+.BR CONFIG_FAIR_GROUP_SCHED ),
+then all of the processes on the system are notionally placed
+in a single task group.
+.PP
+Under group scheduling,
+a thread's nice value has an effect for scheduling decisions
+.IR "only relative to other threads in the same task group" .
+This has some surprising consequences in terms of the traditional semantics
+of the nice value on UNIX systems.
+In particular, if autogrouping
+is enabled (which is the default in various distributions), then employing
+.BR setpriority (2)
+or
+.BR nice (1)
+on a process has an effect only for scheduling relative
+to other processes executed in the same session
+(typically: the same terminal window).
+.PP
+Conversely, for two processes that are (for example)
+the sole CPU-bound processes in different sessions
+(e.g., different terminal windows,
+each of whose jobs are tied to different autogroups),
+.IR "modifying the nice value of the process in one of the sessions"
+.IR "has no effect"
+in terms of the scheduler's decisions relative to the
+process in the other session.
+.\" More succinctly: the nice(1) command is in many cases a no-op since
+.\" Linux 2.6.38.
+.\"
+A possibly useful workaround here is to use a command such as
+the following to modify the autogroup nice value for
+.I all
+of the processes in a terminal session:
+.PP
+.in +4n
+.EX
+$ \fBecho 10 > /proc/self/autogroup\fP
+.EE
+.in
.SS Real-time features in the mainline Linux kernel
.\" FIXME . Probably this text will need some minor tweaking
.\" ask Carsten Emde about this.
-From kernel version 2.6.18 onward, however, Linux is gradually
+Since kernel version 2.6.18, Linux is gradually
becoming equipped with real-time capabilities,
most of which are derived from the former
.I realtime-preempt
mainline kernel,
they must be installed to achieve the best real-time performance.
These patches are named:
+.PP
.in +4n
-.nf
-
+.EX
patch-\fIkernelversion\fP-rt\fIpatchversion\fP
-.fi
+.EE
.in
.PP
and can be downloaded from
.UR http://www.kernel.org\:/pub\:/linux\:/kernel\:/projects\:/rt/
.UE .
-
+.PP
Without the patches and prior to their full inclusion into the mainline
kernel, the kernel configuration offers only the three preemption classes
.BR CONFIG_PREEMPT_NONE ,
.B CONFIG_PREEMPT_DESKTOP
which respectively provide no, some, and considerable
reduction of the worst-case scheduling latency.
-
+.PP
With the patches applied or after their full inclusion into the mainline
kernel, the additional configuration item
.B CONFIG_PREEMPT_RT
real-time operating system.
The FIFO and RR scheduling policies are then used to run a thread
with true real-time priority and a minimum worst-case scheduling latency.
+.SH NOTES
+The
+.BR cgroups (7)
+CPU controller can be used to limit the CPU consumption of
+groups of processes.
+.PP
+Originally, Standard Linux was intended as a general-purpose operating
+system being able to handle background processes, interactive
+applications, and less demanding 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.
.SH SEE ALSO
.ad l
.nh
+.BR chcpu (1),
.BR chrt (1),
+.BR lscpu (1),
+.BR ps (1),
.BR taskset (1),
+.BR top (1),
.BR getpriority (2),
.BR mlock (2),
.BR mlockall (2),
.BR sched_yield (2),
.BR setpriority (2),
.BR pthread_getaffinity_np (3),
+.BR pthread_getschedparam (3),
.BR pthread_setaffinity_np (3),
.BR sched_getcpu (3),
.BR capabilities (7),
.ad
.PP
.I Programming for the real world \- POSIX.4
-by Bill O. Gallmeister, O'Reilly & Associates, Inc., ISBN 1-56592-074-0.
+by Bill O.\& Gallmeister, O'Reilly & Associates, Inc., ISBN 1-56592-074-0.
.PP
The Linux kernel source files
.IR Documentation/scheduler/sched-deadline.txt ,
projects / thirdparty / man-pages.git / blobdiff