0001 perf-kowrk(1)
0002 =============
0003
0004 NAME
0005 ----
0006 perf-kwork - Tool to trace/measure kernel work properties (latencies)
0007
0008 SYNOPSIS
0009 --------
0010 [verse]
0011 'perf kwork' {record}
0012
0013 DESCRIPTION
0014 -----------
0015 There are several variants of 'perf kwork':
0016
0017 'perf kwork record <command>' to record the kernel work
0018 of an arbitrary workload.
0019
0020 'perf kwork report' to report the per kwork runtime.
0021
0022 'perf kwork latency' to report the per kwork latencies.
0023
0024 'perf kwork timehist' provides an analysis of kernel work events.
0025
0026 Example usage:
0027 perf kwork record -- sleep 1
0028 perf kwork report
0029 perf kwork report -b
0030 perf kwork latency
0031 perf kwork latency -b
0032 perf kwork timehist
0033
0034 By default it shows the individual work events such as irq, workqeueu,
0035 including the run time and delay (time between raise and actually entry):
0036
0037 Runtime start Runtime end Cpu Kwork name Runtime Delaytime
0038 (TYPE)NAME:NUM (msec) (msec)
0039 ----------------- ----------------- ------ ------------------------- ---------- ----------
0040 1811186.976062 1811186.976327 [0000] (s)RCU:9 0.266 0.114
0041 1811186.978452 1811186.978547 [0000] (s)SCHED:7 0.095 0.171
0042 1811186.980327 1811186.980490 [0000] (s)SCHED:7 0.162 0.083
0043 1811186.981221 1811186.981271 [0000] (s)SCHED:7 0.050 0.077
0044 1811186.984267 1811186.984318 [0000] (s)SCHED:7 0.051 0.075
0045 1811186.987252 1811186.987315 [0000] (s)SCHED:7 0.063 0.081
0046 1811186.987785 1811186.987843 [0006] (s)RCU:9 0.058 0.645
0047 1811186.988319 1811186.988383 [0000] (s)SCHED:7 0.064 0.143
0048 1811186.989404 1811186.989607 [0002] (s)TIMER:1 0.203 0.111
0049 1811186.989660 1811186.989732 [0002] (s)SCHED:7 0.072 0.310
0050 1811186.991295 1811186.991407 [0002] eth0:10 0.112
0051 1811186.991639 1811186.991734 [0002] (s)NET_RX:3 0.095 0.277
0052 1811186.989860 1811186.991826 [0002] (w)vmstat_shepherd 1.966 0.345
0053 ...
0054
0055 Times are in msec.usec.
0056
0057 OPTIONS
0058 -------
0059 -D::
0060 --dump-raw-trace=::
0061 Display verbose dump of the sched data.
0062
0063 -f::
0064 --force::
0065 Don't complain, do it.
0066
0067 -k::
0068 --kwork::
0069 List of kwork to profile (irq, softirq, workqueue, etc)
0070
0071 -v::
0072 --verbose::
0073 Be more verbose. (show symbol address, etc)
0074
0075 OPTIONS for 'perf kwork report'
0076 ----------------------------
0077
0078 -b::
0079 --use-bpf::
0080 Use BPF to measure kwork runtime
0081
0082 -C::
0083 --cpu::
0084 Only show events for the given CPU(s) (comma separated list).
0085
0086 -i::
0087 --input::
0088 Input file name. (default: perf.data unless stdin is a fifo)
0089
0090 -n::
0091 --name::
0092 Only show events for the given name.
0093
0094 -s::
0095 --sort::
0096 Sort by key(s): runtime, max, count
0097
0098 -S::
0099 --with-summary::
0100 Show summary with statistics
0101
0102 --time::
0103 Only analyze samples within given time window: <start>,<stop>. Times
0104 have the format seconds.microseconds. If start is not given (i.e., time
0105 string is ',x.y') then analysis starts at the beginning of the file. If
0106 stop time is not given (i.e, time string is 'x.y,') then analysis goes
0107 to end of file.
0108
0109 OPTIONS for 'perf kwork latency'
0110 ----------------------------
0111
0112 -b::
0113 --use-bpf::
0114 Use BPF to measure kwork latency
0115
0116 -C::
0117 --cpu::
0118 Only show events for the given CPU(s) (comma separated list).
0119
0120 -i::
0121 --input::
0122 Input file name. (default: perf.data unless stdin is a fifo)
0123
0124 -n::
0125 --name::
0126 Only show events for the given name.
0127
0128 -s::
0129 --sort::
0130 Sort by key(s): avg, max, count
0131
0132 --time::
0133 Only analyze samples within given time window: <start>,<stop>. Times
0134 have the format seconds.microseconds. If start is not given (i.e., time
0135 string is ',x.y') then analysis starts at the beginning of the file. If
0136 stop time is not given (i.e, time string is 'x.y,') then analysis goes
0137 to end of file.
0138
0139 OPTIONS for 'perf kwork timehist'
0140 ---------------------------------
0141
0142 -C::
0143 --cpu::
0144 Only show events for the given CPU(s) (comma separated list).
0145
0146 -g::
0147 --call-graph::
0148 Display call chains if present (default off).
0149
0150 -i::
0151 --input::
0152 Input file name. (default: perf.data unless stdin is a fifo)
0153
0154 -k::
0155 --vmlinux=<file>::
0156 Vmlinux pathname
0157
0158 -n::
0159 --name::
0160 Only show events for the given name.
0161
0162 --kallsyms=<file>::
0163 Kallsyms pathname
0164
0165 --max-stack::
0166 Maximum number of functions to display in backtrace, default 5.
0167
0168 --symfs=<directory>::
0169 Look for files with symbols relative to this directory.
0170
0171 --time::
0172 Only analyze samples within given time window: <start>,<stop>. Times
0173 have the format seconds.microseconds. If start is not given (i.e., time
0174 string is ',x.y') then analysis starts at the beginning of the file. If
0175 stop time is not given (i.e, time string is 'x.y,') then analysis goes
0176 to end of file.
0177
0178 SEE ALSO
0179 --------
0180 linkperf:perf-record[1]
