1PERF-FTRACE(1) perf Manual PERF-FTRACE(1)
2
6 perf-ftrace - simple wrapper for kernel's ftrace functionality
7
9 perf ftrace {trace|latency} <command>
10
12 The perf ftrace command provides a collection of subcommands which use
13 kernel’s ftrace infrastructure.
14 'perf ftrace trace' is a simple wrapper of the ftrace. It only supports
16 single thread tracing currently and just reads trace_pipe in text and then
17 write it to stdout.
18 'perf ftrace latency' calculates execution latency of a given function
20 (optionally with BPF) and display it as a histogram.
21 The following options apply to perf ftrace.
23
25 -p, --pid=
26 Trace on existing process id (comma separated list).
27 --tid=
29 Trace on existing thread id (comma separated list).
30 -a, --all-cpus
32 Force system-wide collection. Scripts run without a <command>
33 normally use -a by default, while scripts run with a <command>
34 normally don’t - this option allows the latter to be run in
35 system-wide mode.
36 -C, --cpu=
38 Only trace for the list of CPUs provided. Multiple CPUs can be
39 provided as a comma separated list with no space like: 0,1. Ranges
40 of CPUs are specified with -: 0-2. Default is to trace on all
41 online CPUs.
42 -v, --verbose
44 Increase the verbosity level.
45
47 -t, --tracer=
48 Tracer to use when neither -G nor -F option is not specified:
49 function_graph or function.
50 -F, --funcs
52 List available functions to trace. It accepts a pattern to only
53 list interested functions.
54 -D, --delay
56 Time (ms) to wait before starting tracing after program start.
57 -m, --buffer-size
59 Set the size of per-cpu tracing buffer, <size> is expected to be a
60 number with appended unit character - B/K/M/G.
61 --inherit
63 Trace children processes spawned by our target.
64 -T, --trace-funcs=
66 Select function tracer and set function filter on the given
67 function (or a glob pattern). Multiple functions can be given by
68 using this option more than once. The function argument also can be
69 a glob pattern. It will be passed to set_ftrace_filter in tracefs.
70 -N, --notrace-funcs=
72 Select function tracer and do not trace functions given by the
73 argument. Like -T option, this can be used more than once to
74 specify multiple functions (or glob patterns). It will be passed to
75 set_ftrace_notrace in tracefs.
76 --func-opts
78 List of options allowed to set:
79 • call-graph - Display kernel stack trace for function tracer.
81 • irq-info - Display irq context info for function tracer.
83 -G, --graph-funcs=
85 Select function_graph tracer and set graph filter on the given
86 function (or a glob pattern). This is useful to trace for functions
87 executed from the given function. This can be used more than once
88 to specify multiple functions. It will be passed to
89 set_graph_function in tracefs.
90 -g, --nograph-funcs=
92 Select function_graph tracer and set graph notrace filter on the
93 given function (or a glob pattern). Like -G option, this is useful
94 for the function_graph tracer only and disables tracing for
95 function executed from the given function. This can be used more
96 than once to specify multiple functions. It will be passed to
97 set_graph_notrace in tracefs.
98 --graph-opts
100 List of options allowed to set:
101 • nosleep-time - Measure on-CPU time only for function_graph
103 tracer.
104 • noirqs - Ignore functions that happen inside interrupt.
106 • verbose - Show process names, PIDs, timestamps, etc.
108 • thresh=<n> - Setup trace duration threshold in microseconds.
110 • depth=<n> - Set max depth for function graph tracer to follow.
112
114 -T, --trace-funcs=
115 Set the function name to get the histogram. Unlike perf ftrace
116 trace, it only allows single function to calculate the histogram.
117 -b, --use-bpf
119 Use BPF to measure function latency instead of using the ftrace (it
120 uses function_graph tracer internally).
121 -n, --use-nsec
123 Use nano-second instead of micro-second as a base unit of the
124 histogram.
125
127 perf-record(1), perf-trace(1)
128perf 11/28/2023 PERF-FTRACE(1)