path: root/man2/ioprio_set.2

.\" This is _*_ nroff _*_ source. Emacs, gimme all those colors :)
.\"
.\" Copyright (c) International Business Machines orp., 2006
.\"
.\" This program is free software; you can redistribute it and/or
.\" modify it under the terms of the GNU General Public License as
.\" published by the Free Software Foundation; either version 2 of
.\" the License, or (at your option) any later version.
.\"
.\" This program is distributed in the hope that it will be useful,
.\" but WITHOUT ANY WARRANTY; without even the implied warranty of
.\" MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
.\" the GNU General Public License for more details.
.\"
.\" You should have received a copy of the GNU General Public License
.\" along with this program; if not, write to the Free Software
.\" Foundation, Inc., 59 Temple Place, Suite 330, Boston,
.\" MA 02111-1307 USA
.\"
.\" HISTORY:
.\" 2006-04-27, created by Eduardo M. Fleury <efleury@br.ibm.com>
.\" with various additions by Michael Kerrisk <mtk.manpages@gmail.com>
.\"
.\"
.TH IOPRIO_SET 2 2013-02-12 "Linux" "Linux Programmer's Manual"
.SH NAME
ioprio_get, ioprio_set \- get/set I/O scheduling class and priority
.SH SYNOPSIS
.nf
.BI "int ioprio_get(int " which ", int " who );
.BI "int ioprio_set(int " which ", int " who ", int " ioprio );
.fi

.IR Note :
There are no glibc wrappers for these system calls; see NOTES.
.SH DESCRIPTION
The
.BR ioprio_get ()
and
.BR ioprio_set ()
system calls respectively get and set the I/O scheduling class and
priority of one or more threads.

The
.I which
and
.I who
arguments identify the thread(s) on which the system
calls operate.
The
.I which
argument determines how
.I who
is interpreted, and has one of the following values:
.TP
.B IOPRIO_WHO_PROCESS
.I who
is a process ID or thread ID identifying a single process or thread.
If
.I who
is 0, then operate on the calling thread.
.TP
.B IOPRIO_WHO_PGRP
.I who
is a process group ID identifying all the members of a process group.
If
.I who
is 0, then operate on the process group of which the caller is a member.
.TP
.B IOPRIO_WHO_USER
.I who
is a user ID identifying all of the processes that
have a matching real UID.
.\" FIXME who==0 needs to be documented,
.\" See http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=652443
.PP
If
.I which
is specified as
.B IOPRIO_WHO_PGRP
or
.B IOPRIO_WHO_USER
when calling
.BR ioprio_get (),
and more than one process matches
.IR who ,
then the returned priority will be the highest one found among
all of the matching processes.
One priority is said to be
higher than another one if it belongs to a higher priority
class
.RB ( IOPRIO_CLASS_RT
is the highest priority class;
.B IOPRIO_CLASS_IDLE
is the lowest)
or if it belongs to the same priority class as the other process but
has a higher priority level (a lower priority number means a
higher priority level).

The
.I ioprio
argument given to
.BR ioprio_set ()
is a bit mask that specifies both the scheduling class and the
priority to be assigned to the target process(es).
The following macros are used for assembling and dissecting
.I ioprio
values:
.TP
.BI IOPRIO_PRIO_VALUE( class ", " data )
Given a scheduling
.I class
and priority
.RI ( data ),
this macro combines the two values to produce an
.I ioprio
value, which is returned as the result of the macro.
.TP
.BI IOPRIO_PRIO_CLASS( mask )
Given
.I mask
(an
.I ioprio
value), this macro returns its I/O class component, that is,
one of the values
.BR IOPRIO_CLASS_RT ,
.BR IOPRIO_CLASS_BE ,
or
.BR IOPRIO_CLASS_IDLE .
.TP
.BI IOPRIO_PRIO_DATA( mask )
Given
.I mask
(an
.I ioprio
value), this macro returns its priority
.RI ( data )
component.
.PP
See the NOTES section for more
information on scheduling classes and priorities.

I/O priorities are supported for reads and for synchronous
.RB ( O_DIRECT ,
.BR O_SYNC )
writes.
I/O priorities are not supported for asynchronous
writes because they are issued outside the context of the program
dirtying the memory, and thus program-specific priorities do not apply.
.SH "RETURN VALUE"
On success,
.BR ioprio_get ()
returns the
.I ioprio
value of the process with highest I/O priority of any of the processes
that match the criteria specified in
.I which
and
.IR who .
On error, \-1 is returned, and
.I errno
is set to indicate the error.
.PP
On success,
.BR ioprio_set ()
returns 0.
On error, \-1 is returned, and
.I errno
is set to indicate the error.
.SH ERRORS
.TP
.B EINVAL
Invalid value for
.I which
or
.IR ioprio .
Refer to the NOTES section for available scheduler
classes and priority levels for
.IR ioprio .
.TP
.B EPERM
The calling process does not have the privilege needed to assign this
.I ioprio
to the specified process(es).
See the NOTES section for more information on required
privileges for
.BR ioprio_set ().
.TP
.B ESRCH
No process(es) could be found that matched the specification in
.I which
and
.IR who .
.SH VERSIONS
These system calls have been available on Linux since
kernel 2.6.13.
.SH "CONFORMING TO"
These system calls are Linux-specific.
.SH NOTES
Glibc does not provide a wrapper for these system calls; call them using
.BR syscall (2).

Two or more processes or threads can share an I/O context.
This will be the case when
.BR clone (2)
was called with the
.B CLONE_IO
flag.
However, by default, the distinct threads of a process will
.B not
share the same I/O context.
This means that if you want to change the I/O
priority of all threads in a process, you may need to call
.BR ioprio_set ()
on each of the threads.
The thread ID that you would need for this operation
is the one that is returned by
.BR gettid (2)
or
.BR clone (2).

These system calls only have an effect when used
in conjunction with an I/O scheduler that supports I/O priorities.
As at kernel 2.6.17 the only such scheduler is the Completely Fair Queuing
(CFQ) I/O scheduler.
.SS "Selecting an I/O Scheduler"
I/O Schedulers are selected on a per-device basis via the special
file
.IR /sys/block/<device>/queue/scheduler .

One can view the current I/O scheduler via the
.I /sys
file system.
For example, the following command
displays a list of all schedulers currently loaded in the kernel:
.sp
.RS
.nf
.RB "$" " cat /sys/block/hda/queue/scheduler"
noop anticipatory deadline [cfq]
.fi
.RE
.sp
The scheduler surrounded by brackets is the one actually
in use for the device
.RI ( hda
in the example).
Setting another scheduler is done by writing the name of the
new scheduler to this file.
For example, the following command will set the
scheduler for the
.I hda
device to
.IR cfq :
.sp
.RS
.nf
.RB "$" " su"
Password:
.RB "#" " echo cfq > /sys/block/hda/queue/scheduler"
.fi
.RE
.SS "The Completely Fair Queuing (CFQ) I/O Scheduler"
Since v3 (aka CFQ Time Sliced) CFQ implements
I/O nice levels similar to those
of CPU scheduling.
These nice levels are grouped in three scheduling classes
each one containing one or more priority levels:
.TP
.BR IOPRIO_CLASS_RT " (1)"
This is the real-time I/O class.
This scheduling class is given
higher priority than any other class:
processes from this class are
given first access to the disk every time.
Thus this I/O class needs to be used with some
care: one I/O real-time process can starve the entire system.
Within the real-time class,
there are 8 levels of class data (priority) that determine exactly
how much time this process needs the disk for on each service.
The highest real-time priority level is 0; the lowest is 7.
In the future this might change to be more directly mappable to
performance, by passing in a desired data rate instead.
.TP
.BR IOPRIO_CLASS_BE " (2)"
This is the best-effort scheduling class,
which is the default for any process
that hasn't set a specific I/O priority.
The class data (priority) determines how much
I/O bandwidth the process will get.
Best-effort priority levels are analogous to CPU nice values
(see
.BR getpriority (2)).
The priority level determines a priority relative
to other processes in the best-effort scheduling class.
Priority levels range from 0 (highest) to 7 (lowest).
.TP
.BR IOPRIO_CLASS_IDLE " (3)"
This is the idle scheduling class.
Processes running at this level only get I/O
time when no-one else needs the disk.
The idle class has no class data.
Attention is required when assigning this priority class to a process,
since it may become starved if higher priority processes are
constantly accessing the disk.
.PP
Refer to
.I Documentation/block/ioprio.txt
for more information on the CFQ I/O Scheduler and an example program.
.SS "Required permissions to set I/O priorities"
Permission to change a process's priority is granted or denied based
on two assertions:
.TP
.B "Process ownership"
An unprivileged process may only set the I/O priority of a process
whose real UID
matches the real or effective UID of the calling process.
A process which has the
.B CAP_SYS_NICE
capability can change the priority of any process.
.TP
.B "What is the desired priority"
Attempts to set very high priorities
.RB ( IOPRIO_CLASS_RT )
require the
.B CAP_SYS_ADMIN
capability.
Kernel versions up to 2.6.24 also required
.B CAP_SYS_ADMIN
to set a very low priority
.RB ( IOPRIO_CLASS_IDLE ),
but since Linux 2.6.25, this is no longer required.
.PP
A call to
.BR ioprio_set ()
must follow both rules, or the call will fail with the error
.BR EPERM .
.SH BUGS
.\" 6 May 07: Bug report raised:
.\" http://sources.redhat.com/bugzilla/show_bug.cgi?id=4464
.\" Ulrich Drepper replied that he wasn't going to add these
.\" to glibc.
Glibc does not yet provide a suitable header file defining
the function prototypes and macros described on this page.
Suitable definitions can be found in
.IR linux/ioprio.h .
.SH "SEE ALSO"
.BR ionice (1),
.BR getpriority (2),
.BR open (2),
.BR capabilities (7)

.I Documentation/block/ioprio.txt
in the Linux kernel source tree