197 lines · plain
1perf-ftrace(1)2==============3 4NAME5----6perf-ftrace - simple wrapper for kernel's ftrace functionality7 8 9SYNOPSIS10--------11[verse]12'perf ftrace' {trace|latency|profile} <command>13 14DESCRIPTION15-----------16The 'perf ftrace' command provides a collection of subcommands which use17kernel's ftrace infrastructure.18 19 'perf ftrace trace' is a simple wrapper of the ftrace. It only supports20 single thread tracing currently and just reads trace_pipe in text and then21 write it to stdout.22 23 'perf ftrace latency' calculates execution latency of a given function24 (optionally with BPF) and display it as a histogram.25 26 'perf ftrace profile' show a execution profile for each function including27 total, average, max time and the number of calls.28 29The following options apply to perf ftrace.30 31COMMON OPTIONS32--------------33 34-p::35--pid=::36 Trace on existing process id (comma separated list).37 38--tid=::39 Trace on existing thread id (comma separated list).40 41-a::42--all-cpus::43 Force system-wide collection. Scripts run without a <command>44 normally use -a by default, while scripts run with a <command>45 normally don't - this option allows the latter to be run in46 system-wide mode.47 48-C::49--cpu=::50 Only trace for the list of CPUs provided. Multiple CPUs can51 be provided as a comma separated list with no space like: 0,1.52 Ranges of CPUs are specified with -: 0-2.53 Default is to trace on all online CPUs.54 55-v::56--verbose::57 Increase the verbosity level.58 59 60OPTIONS for 'perf ftrace trace'61-------------------------------62 63-t::64--tracer=::65 Tracer to use when neither -G nor -F option is not66 specified: function_graph or function.67 68-F::69--funcs::70 List available functions to trace. It accepts a pattern to71 only list interested functions.72 73-D::74--delay::75 Time (ms) to wait before starting tracing after program start.76 77-m::78--buffer-size::79 Set the size of per-cpu tracing buffer, <size> is expected to80 be a number with appended unit character - B/K/M/G.81 82--inherit::83 Trace children processes spawned by our target.84 85-T::86--trace-funcs=::87 Select function tracer and set function filter on the given88 function (or a glob pattern). Multiple functions can be given89 by using this option more than once. The function argument also90 can be a glob pattern. It will be passed to 'set_ftrace_filter'91 in tracefs.92 93-N::94--notrace-funcs=::95 Select function tracer and do not trace functions given by the96 argument. Like -T option, this can be used more than once to97 specify multiple functions (or glob patterns). It will be98 passed to 'set_ftrace_notrace' in tracefs.99 100--func-opts::101 List of options allowed to set:102 103 - call-graph - Display kernel stack trace for function tracer.104 - irq-info - Display irq context info for function tracer.105 106-G::107--graph-funcs=::108 Select function_graph tracer and set graph filter on the given109 function (or a glob pattern). This is useful to trace for110 functions executed from the given function. This can be used more111 than once to specify multiple functions. It will be passed to112 'set_graph_function' in tracefs.113 114-g::115--nograph-funcs=::116 Select function_graph tracer and set graph notrace filter on the117 given function (or a glob pattern). Like -G option, this is useful118 for the function_graph tracer only and disables tracing for function119 executed from the given function. This can be used more than once to120 specify multiple functions. It will be passed to 'set_graph_notrace'121 in tracefs.122 123--graph-opts::124 List of options allowed to set:125 126 - nosleep-time - Measure on-CPU time only for function_graph tracer.127 - noirqs - Ignore functions that happen inside interrupt.128 - verbose - Show process names, PIDs, timestamps, etc.129 - thresh=<n> - Setup trace duration threshold in microseconds.130 - depth=<n> - Set max depth for function graph tracer to follow.131 - tail - Print function name at the end.132 133 134OPTIONS for 'perf ftrace latency'135---------------------------------136 137-T::138--trace-funcs=::139 Set the function name to get the histogram. Unlike perf ftrace trace,140 it only allows single function to calculate the histogram.141 142-b::143--use-bpf::144 Use BPF to measure function latency instead of using the ftrace (it145 uses function_graph tracer internally).146 147-n::148--use-nsec::149 Use nano-second instead of micro-second as a base unit of the histogram.150 151 152OPTIONS for 'perf ftrace profile'153---------------------------------154 155-T::156--trace-funcs=::157 Set function filter on the given function (or a glob pattern).158 Multiple functions can be given by using this option more than once.159 The function argument also can be a glob pattern. It will be passed160 to 'set_ftrace_filter' in tracefs.161 162-N::163--notrace-funcs=::164 Do not trace functions given by the argument. Like -T option, this165 can be used more than once to specify multiple functions (or glob166 patterns). It will be passed to 'set_ftrace_notrace' in tracefs.167 168-G::169--graph-funcs=::170 Set graph filter on the given function (or a glob pattern). This is171 useful to trace for functions executed from the given function. This172 can be used more than once to specify multiple functions. It will be173 passed to 'set_graph_function' in tracefs.174 175-g::176--nograph-funcs=::177 Set graph notrace filter on the given function (or a glob pattern).178 Like -G option, this is useful for the function_graph tracer only and179 disables tracing for function executed from the given function. This180 can be used more than once to specify multiple functions. It will be181 passed to 'set_graph_notrace' in tracefs.182 183-m::184--buffer-size::185 Set the size of per-cpu tracing buffer, <size> is expected to186 be a number with appended unit character - B/K/M/G.187 188-s::189--sort=::190 Sort the result by the given field. Available values are:191 total, avg, max, count, name. Default is 'total'.192 193 194SEE ALSO195--------196linkperf:perf-record[1], linkperf:perf-trace[1]197