5 Tasks encounter delays in execution when they wait
6 for some kernel resource to become available e.g. a
7 runnable task may wait for a free CPU to run on.
9 The per-task delay accounting functionality measures
10 the delays experienced by a task while
12 a) waiting for a CPU (while being runnable)
13 b) completion of synchronous block I/O initiated by the task
20 and makes these statistics available to userspace through
21 the taskstats interface.
23 Such delays provide feedback for setting a task's cpu priority,
24 io priority and rss limit values appropriately. Long delays for
25 important tasks could be a trigger for raising its corresponding priority.
27 The functionality, through its use of the taskstats interface, also provides
28 delay statistics aggregated for all tasks (or threads) belonging to a
29 thread group (corresponding to a traditional Unix process). This is a commonly
30 needed aggregation that is more efficiently done by the kernel.
32 Userspace utilities, particularly resource management applications, can also
33 aggregate delay statistics into arbitrary groups. To enable this, delay
34 statistics of a task are available both during its lifetime as well as on its
35 exit, ensuring continuous and complete monitoring can be done.
41 Delay accounting uses the taskstats interface which is described
42 in detail in a separate document in this directory. Taskstats returns a
43 generic data structure to userspace corresponding to per-pid and per-tgid
44 statistics. The delay accounting functionality populates specific fields of
47 include/uapi/linux/taskstats.h
49 for a description of the fields pertaining to delay accounting.
50 It will generally be in the form of counters returning the cumulative
51 delay seen for cpu, sync block I/O, swapin, memory reclaim, thrash page
52 cache, direct compact, write-protect copy etc.
54 Taking the difference of two successive readings of a given
55 counter (say cpu_delay_total) for a task will give the delay
56 experienced by the task waiting for the corresponding resource
59 When a task exits, records containing the per-task statistics
60 are sent to userspace without requiring a command. If it is the last exiting
61 task of a thread group, the per-tgid statistics are also sent. More details
62 are given in the taskstats interface description.
64 The getdelays.c userspace utility in tools/accounting directory allows simple
65 commands to be run and the corresponding delay statistics to be displayed. It
66 also serves as an example of using the taskstats interface.
71 Compile the kernel with::
73 CONFIG_TASK_DELAY_ACCT=y
76 Delay accounting is disabled by default at boot up.
81 to the kernel boot options. The rest of the instructions below assume this has
82 been done. Alternatively, use sysctl kernel.task_delayacct to switch the state
83 at runtime. Note however that only tasks started after enabling it will have
84 delayacct information.
86 After the system has booted up, use a utility
87 similar to getdelays.c to access the delays
88 seen by a given task or a task group (tgid).
89 The utility also allows a given command to be
90 executed and the corresponding delays to be
93 General format of the getdelays command::
95 getdelays [-dilv] [-t tgid] [-p pid]
97 Get delays, since system boot, for pid 10::
99 # ./getdelays -d -p 10
100 (output similar to next case)
102 Get sum of delays, since system boot, for all pids with tgid 5::
104 # ./getdelays -d -t 5
105 print delayacct stats ON
109 CPU count real total virtual total delay total delay average
110 8 7000000 6872122 3382277 0.423ms
111 IO count delay total delay average
113 SWAP count delay total delay average
115 RECLAIM count delay total delay average
117 THRASHING count delay total delay average
119 COMPACT count delay total delay average
121 WPCOPY count delay total delay average
124 Get IO accounting for pid 1, it works only with -p::
126 # ./getdelays -i -p 1
127 printing IO accounting
128 linuxrc: read=65536, write=0, cancelled_write=0
130 The above command can be used with -v to get more debug information.