6 perf-kwork - Tool to trace/measure kernel work properties (latencies)
15 There are several variants of 'perf kwork':
17 'perf kwork record <command>' to record the kernel work
18 of an arbitrary workload.
20 'perf kwork report' to report the per kwork runtime.
22 'perf kwork latency' to report the per kwork latencies.
24 'perf kwork timehist' provides an analysis of kernel work events.
27 perf kwork record -- sleep 1
34 By default it shows the individual work events such as irq, workqeueu,
35 including the run time and delay (time between raise and actually entry):
37 Runtime start Runtime end Cpu Kwork name Runtime Delaytime
38 (TYPE)NAME:NUM (msec) (msec)
39 ----------------- ----------------- ------ ------------------------- ---------- ----------
40 1811186.976062 1811186.976327 [0000] (s)RCU:9 0.266 0.114
41 1811186.978452 1811186.978547 [0000] (s)SCHED:7 0.095 0.171
42 1811186.980327 1811186.980490 [0000] (s)SCHED:7 0.162 0.083
43 1811186.981221 1811186.981271 [0000] (s)SCHED:7 0.050 0.077
44 1811186.984267 1811186.984318 [0000] (s)SCHED:7 0.051 0.075
45 1811186.987252 1811186.987315 [0000] (s)SCHED:7 0.063 0.081
46 1811186.987785 1811186.987843 [0006] (s)RCU:9 0.058 0.645
47 1811186.988319 1811186.988383 [0000] (s)SCHED:7 0.064 0.143
48 1811186.989404 1811186.989607 [0002] (s)TIMER:1 0.203 0.111
49 1811186.989660 1811186.989732 [0002] (s)SCHED:7 0.072 0.310
50 1811186.991295 1811186.991407 [0002] eth0:10 0.112
51 1811186.991639 1811186.991734 [0002] (s)NET_RX:3 0.095 0.277
52 1811186.989860 1811186.991826 [0002] (w)vmstat_shepherd 1.966 0.345
55 Times are in msec.usec.
61 Display verbose dump of the sched data.
65 Don't complain, do it.
69 List of kwork to profile (irq, softirq, workqueue, etc)
73 Be more verbose. (show symbol address, etc)
75 OPTIONS for 'perf kwork report'
76 ----------------------------
80 Use BPF to measure kwork runtime
84 Only show events for the given CPU(s) (comma separated list).
88 Input file name. (default: perf.data unless stdin is a fifo)
92 Only show events for the given name.
96 Sort by key(s): runtime, max, count
100 Show summary with statistics
103 Only analyze samples within given time window: <start>,<stop>. Times
104 have the format seconds.microseconds. If start is not given (i.e., time
105 string is ',x.y') then analysis starts at the beginning of the file. If
106 stop time is not given (i.e, time string is 'x.y,') then analysis goes
109 OPTIONS for 'perf kwork latency'
110 ----------------------------
114 Use BPF to measure kwork latency
118 Only show events for the given CPU(s) (comma separated list).
122 Input file name. (default: perf.data unless stdin is a fifo)
126 Only show events for the given name.
130 Sort by key(s): avg, max, count
133 Only analyze samples within given time window: <start>,<stop>. Times
134 have the format seconds.microseconds. If start is not given (i.e., time
135 string is ',x.y') then analysis starts at the beginning of the file. If
136 stop time is not given (i.e, time string is 'x.y,') then analysis goes
139 OPTIONS for 'perf kwork timehist'
140 ---------------------------------
144 Only show events for the given CPU(s) (comma separated list).
148 Display call chains if present (default off).
152 Input file name. (default: perf.data unless stdin is a fifo)
160 Only show events for the given name.
166 Maximum number of functions to display in backtrace, default 5.
168 --symfs=<directory>::
169 Look for files with symbols relative to this directory.
172 Only analyze samples within given time window: <start>,<stop>. Times
173 have the format seconds.microseconds. If start is not given (i.e., time
174 string is ',x.y') then analysis starts at the beginning of the file. If
175 stop time is not given (i.e, time string is 'x.y,') then analysis goes
180 linkperf:perf-record[1]