6 perf-ftrace - simple wrapper for kernel's ftrace functionality
12 'perf ftrace' {trace|latency} <command>
16 The 'perf ftrace' command provides a collection of subcommands which use
17 kernel's ftrace infrastructure.
19 'perf ftrace trace' is a simple wrapper of the ftrace. It only supports
20 single thread tracing currently and just reads trace_pipe in text and then
23 'perf ftrace latency' calculates execution latency of a given function
24 (optionally with BPF) and display it as a histogram.
26 The following options apply to perf ftrace.
33 Trace on existing process id (comma separated list).
36 Trace on existing thread id (comma separated list).
40 Force system-wide collection. Scripts run without a <command>
41 normally use -a by default, while scripts run with a <command>
42 normally don't - this option allows the latter to be run in
47 Only trace for the list of CPUs provided. Multiple CPUs can
48 be provided as a comma separated list with no space like: 0,1.
49 Ranges of CPUs are specified with -: 0-2.
50 Default is to trace on all online CPUs.
54 Increase the verbosity level.
57 OPTIONS for 'perf ftrace trace'
58 -------------------------------
62 Tracer to use when neither -G nor -F option is not
63 specified: function_graph or function.
67 List available functions to trace. It accepts a pattern to
68 only list interested functions.
72 Time (ms) to wait before starting tracing after program start.
76 Set the size of per-cpu tracing buffer, <size> is expected to
77 be a number with appended unit character - B/K/M/G.
80 Trace children processes spawned by our target.
84 Select function tracer and set function filter on the given
85 function (or a glob pattern). Multiple functions can be given
86 by using this option more than once. The function argument also
87 can be a glob pattern. It will be passed to 'set_ftrace_filter'
92 Select function tracer and do not trace functions given by the
93 argument. Like -T option, this can be used more than once to
94 specify multiple functions (or glob patterns). It will be
95 passed to 'set_ftrace_notrace' in tracefs.
98 List of options allowed to set:
100 - call-graph - Display kernel stack trace for function tracer.
101 - irq-info - Display irq context info for function tracer.
105 Select function_graph tracer and set graph filter on the given
106 function (or a glob pattern). This is useful to trace for
107 functions executed from the given function. This can be used more
108 than once to specify multiple functions. It will be passed to
109 'set_graph_function' in tracefs.
113 Select function_graph tracer and set graph notrace filter on the
114 given function (or a glob pattern). Like -G option, this is useful
115 for the function_graph tracer only and disables tracing for function
116 executed from the given function. This can be used more than once to
117 specify multiple functions. It will be passed to 'set_graph_notrace'
121 List of options allowed to set:
123 - nosleep-time - Measure on-CPU time only for function_graph tracer.
124 - noirqs - Ignore functions that happen inside interrupt.
125 - verbose - Show process names, PIDs, timestamps, etc.
126 - thresh=<n> - Setup trace duration threshold in microseconds.
127 - depth=<n> - Set max depth for function graph tracer to follow.
130 OPTIONS for 'perf ftrace latency'
131 ---------------------------------
135 Set the function name to get the histogram. Unlike perf ftrace trace,
136 it only allows single function to calculate the histogram.
140 Use BPF to measure function latency instead of using the ftrace (it
141 uses function_graph tracer internally).
145 Use nano-second instead of micro-second as a base unit of the histogram.
150 linkperf:perf-record[1], linkperf:perf-trace[1]