1 .\" This is _*_ nroff _*_ source. Emacs, gimme all those colors :)
3 .\" Copyright (c) International Business Machines orp., 2006
5 .\" This program is free software; you can redistribute it and/or
6 .\" modify it under the terms of the GNU General Public License as
7 .\" published by the Free Software Foundation; either version 2 of
8 .\" the License, or (at your option) any later version.
10 .\" This program is distributed in the hope that it will be useful,
11 .\" but WITHOUT ANY WARRANTY; without even the implied warranty of
12 .\" MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
13 .\" the GNU General Public License for more details.
15 .\" You should have received a copy of the GNU General Public License
16 .\" along with this program; if not, write to the Free Software
17 .\" Foundation, Inc., 59 Temple Place, Suite 330, Boston,
21 .\" 2006-04-27, created by Eduardo M. Fleury <efleury@br.ibm.com>
22 .\" with various additions by Michael Kerrisk <mtk-manpages@gmx.net>
25 .TH IOPRIO_GET 2 2007-06-01 "Linux" "Linux Programmer's Manual"
27 ioprio_get, ioprio_set \- get/set I/O scheduling class and priority
30 .BI "int ioprio_get(int " which ", int " who );
31 .BI "int ioprio_set(int " which ", int " who ", int " ioprio );
38 system calls respectively get and set the I/O scheduling class and
39 priority of one or more processes.
45 arguments identify the process(es) on which the system
49 argument determines how
51 is interpreted, and has one of the following values:
55 is a process ID identifying a single process.
59 is a process group ID identifying all the members of a process group.
63 is a user ID identifying all of the processes that
64 have a matching real UID.
74 and more than one process matches
76 then the returned priority will be the highest one found among
77 all of the matching processes.
78 One priority is said to be
79 higher than another one if it belongs to a higher priority
82 is the highest priority class;
85 or if it belongs to the same priority class as the other process but
86 has a higher priority level (a lower priority number means a
87 higher priority level).
93 is a bit mask that specifies both the scheduling class and the
94 priority to be assigned to the target process(es).
95 The following macros are used for assembling and dissecting
99 .BI IOPRIO_PRIO_VALUE( class ", " data )
104 this macro combines the two values to produce an
106 value, which is returned as the result of the macro.
108 .BI IOPRIO_PRIO_CLASS( mask )
113 value), this macro returns its I/O class component, that is,
115 .BR IOPRIO_CLASS_RT ,
116 .BR IOPRIO_CLASS_BE ,
118 .BR IOPRIO_CLASS_IDLE .
120 .BI IOPRIO_PRIO_DATA( mask )
125 value), this macro returns its priority
129 See the NOTES section for more
130 information on scheduling classes and priorities.
132 I/O priorities are supported for reads and for synchronous (O_DIRECT,
134 I/O priorities are not supported for asynchronous
135 writes because they are issued outside the context of the program
136 dirtying the memory, and thus program-specific priorities do not apply.
142 value of the process with highest I/O priority of any of the processes
143 that match the criteria specified in
147 On error, \-1 is returned, and
149 is set to indicate the error.
154 On error, \-1 is returned, and
156 is set to indicate the error.
160 The calling process does not have the privilege needed to assign this
162 to the specified process(es).
163 See the NOTES section for more information on required
168 No process(es) could be found that matched the specification in
178 Refer to the NOTES section for available scheduler
179 classes and priority levels for
182 These system calls have been available on Linux since
185 These system calls are Linux specific.
187 Glibc does not provide wrapper for these system calls; call them using
190 These system calls only have an effect when used
191 in conjunction with an I/O scheduler that supports I/O priorities.
192 As at kernel 2.6.17 the only such scheduler is the Completely Fair Queuing
194 .SS "Selecting an I/O Scheduler"
195 I/O Schedulers are selected on a per-device basis via the special
197 .IR /sys/block/<device>/queue/scheduler .
199 One can view the current I/O scheduler via the
202 For example, the following command
203 displays a list of all schedulers currently loaded in the kernel:
207 $ cat /sys/block/hda/queue/scheduler
208 noop anticipatory deadline [cfq]
212 The scheduler surrounded by brackets is the one actually
213 in use for the device
216 Setting another scheduler is done by writing the name of the
217 new scheduler to this file.
218 For example, the following command will set the
228 # echo cfq > /sys/block/hda/queue/scheduler
231 .SS "The Completely Fair Queuing (CFQ) I/O Scheduler"
232 Since v3 (aka CFQ Time Sliced) CFQ implements
233 I/O nice levels similar to those
235 These nice levels are grouped in three scheduling classes
236 each one containing one or more priority levels:
238 .BR IOPRIO_CLASS_RT " (1)"
239 This is the real-time I/O class.
240 This scheduling class is given
241 higher priority than any other class:
242 processes from this class are
243 given first access to the disk every time.
244 Thus this I/O class needs to be used with some
245 care: one I/O real-time process can starve the entire system.
246 Within the real-time class,
247 there are 8 levels of class data (priority) that determine exactly
248 how much time this process needs the disk for on each service.
249 The highest real-time priority level is 0; the lowest is 7.
250 In the future this might change to be more directly mappable to
251 performance, by passing in a desired data rate instead.
253 .BR IOPRIO_CLASS_BE " (2)"
254 This is the best-effort scheduling class,
255 which is the default for any process
256 that hasn't set a specific I/O priority.
257 The class data (priority) determines how much
258 I/O bandwidth the process will get.
259 Best-effort priority levels are analogous to CPU nice values
261 .BR getpriority (2)).
262 The priority level determines a priority relative
263 to other processes in the best-effort scheduling class.
264 Priority levels range from 0 (highest) to 7 (lowest).
266 .BR IOPRIO_CLASS_IDLE " (3)"
267 This is the idle scheduling class.
268 Processes running at this level only get I/O
269 time when no one else needs the disk.
270 The idle class has no class
272 Attention is required when assigning this priority class to a process,
273 since it may become starved if higher priority processes are
274 constantly accessing the disk.
277 .I Documentation/block/ioprio.txt
278 for more information on the CFQ I/O Scheduler and an example program.
279 .SS "Required permissions to set I/O priorities"
280 Permission to change a process's priority is granted or denied based
283 .B "Process ownership"
284 An unprivileged process may only set the I/O priority of a process
286 matches the real or effective UID of the calling process.
287 A process which has the
289 capability can change the priority of any process.
291 .B "What is the desired priority"
292 Attempts to set very high priorities
293 .RB ( IOPRIO_CLASS_RT )
295 .RB ( IOPRIO_CLASS_IDLE )
302 must follow both rules, or the call will fail with the error
305 .\" 6 May 07: Bug report raised:
306 .\" http://sources.redhat.com/bugzilla/show_bug.cgi?id=4464
307 .\" Ulriich Drepper replied that he wasn't going to add these
309 Glibc does not yet provide a suitable header file defining
310 the function prototypes and macros described on this page.
311 Suitable definitions can be found in
314 .BR getpriority "(2), " open "(2), " capabilities (7)
316 Documentation/block/ioprio.txt in the kernel source tree.