6 perf-script - Read perf.data (created by perf record) and display trace output
11 'perf script' [<options>]
12 'perf script' [<options>] record <script> [<record-options>] <command>
13 'perf script' [<options>] report <script> [script-args]
14 'perf script' [<options>] <script> <required-script-args> [<record-options>] <command>
15 'perf script' [<options>] <top-script> [script-args]
19 This command reads the input file and displays the trace recorded.
21 There are several variants of perf script:
23 'perf script' to see a detailed trace of the workload that was
26 You can also run a set of pre-canned scripts that aggregate and
27 summarize the raw trace data in various ways (the list of scripts is
28 available via 'perf script -l'). The following variants allow you to
29 record and run those scripts:
31 'perf script record <script> <command>' to record the events required
32 for 'perf script report'. <script> is the name displayed in the
33 output of 'perf script --list' i.e. the actual script name minus any
34 language extension. If <command> is not specified, the events are
35 recorded using the -a (system-wide) 'perf record' option.
37 'perf script report <script> [args]' to run and display the results
38 of <script>. <script> is the name displayed in the output of 'perf
39 script --list' i.e. the actual script name minus any language
40 extension. The perf.data output from a previous run of 'perf script
41 record <script>' is used and should be present for this command to
42 succeed. [args] refers to the (mainly optional) args expected by
45 'perf script <script> <required-script-args> <command>' to both
46 record the events required for <script> and to run the <script>
47 using 'live-mode' i.e. without writing anything to disk. <script>
48 is the name displayed in the output of 'perf script --list' i.e. the
49 actual script name minus any language extension. If <command> is
50 not specified, the events are recorded using the -a (system-wide)
51 'perf record' option. If <script> has any required args, they
52 should be specified before <command>. This mode doesn't allow for
53 optional script args to be specified; if optional script args are
54 desired, they can be specified using separate 'perf script record'
55 and 'perf script report' commands, with the stdout of the record step
56 piped to the stdin of the report script, using the '-o -' and '-i -'
57 options of the corresponding commands.
59 'perf script <top-script>' to both record the events required for
60 <top-script> and to run the <top-script> using 'live-mode'
61 i.e. without writing anything to disk. <top-script> is the name
62 displayed in the output of 'perf script --list' i.e. the actual
63 script name minus any language extension; a <top-script> is defined
64 as any script name ending with the string 'top'.
66 [<record-options>] can be passed to the record steps of 'perf script
67 record' and 'live-mode' variants; this isn't possible however for
68 <top-script> 'live-mode' or 'perf script report' variants.
70 See the 'SEE ALSO' section for links to language-specific
71 information on how to write and run your own trace scripts.
76 Any command you can specify in a shell.
80 Display verbose dump of the trace data.
84 Show latency attributes (irqs/preemption disabled, etc).
88 Display a list of available trace scripts.
92 Process trace data with the given script ([lang]:script[.ext]).
93 If the string 'lang' is specified in place of a script name, a
94 list of supported languages will be displayed instead.
98 Generate perf-script.[ext] starter script for given language,
99 using current perf.data.
102 Filter sample events using the given shared object file.
103 Refer linkperf:perf-dlfilter[1]
106 Force system-wide collection. Scripts run without a <command>
107 normally use -a by default, while scripts run with a <command>
108 normally don't - this option allows the latter to be run in
113 Input file name. (default: perf.data unless stdin is a fifo)
117 Do various checks like samples ordering and lost events.
121 Comma separated list of fields to print. Options are:
122 comm, tid, pid, time, cpu, event, trace, ip, sym, dso, addr, symoff,
123 srcline, period, iregs, uregs, brstack, brstacksym, flags, bpf-output,
124 brstackinsn, brstackoff, callindent, insn, insnlen, synth, phys_addr,
125 metric, misc, srccode, ipc, data_page_size, code_page_size.
126 Field list can be prepended with the type, trace, sw or hw,
127 to indicate to which event type the field list applies.
128 e.g., -F sw:comm,tid,time,ip,sym and -F trace:time,cpu,trace
130 perf script -F <fields>
134 perf script -F trace:<fields> -F sw:<fields> -F hw:<fields>
136 i.e., the specified fields apply to all event types if the type string
139 In addition to overriding fields, it is also possible to add or remove
140 fields from the defaults. For example
144 removes the cpu field and adds the insn field. Adding/removing fields
145 cannot be mixed with normal overriding.
147 The arguments are processed in the order received. A later usage can
148 reset a prior request. e.g.:
150 -F trace: -F comm,tid,time,ip,sym
152 The first -F suppresses trace events (field list is ""), but then the
153 second invocation sets the fields to comm,tid,time,ip,sym. In this case a
154 warning is given to the user:
156 "Overriding previous field request for all events."
158 Alternatively, consider the order:
160 -F comm,tid,time,ip,sym -F trace:
162 The first -F sets the fields for all events and the second -F
163 suppresses trace events. The user is given a warning message about
164 the override, and the result of the above is that only S/W and H/W
165 events are displayed with the given fields.
167 It's possible tp add/remove fields only for specific event type:
171 removes cpu and period from software events.
173 For the 'wildcard' option if a user selected field is invalid for an
174 event type, a message is displayed to the user that the option is
175 ignored for that type. For example:
177 $ perf script -F comm,tid,trace
178 'trace' not valid for hardware events. Ignoring.
179 'trace' not valid for software events. Ignoring.
181 Alternatively, if the type is given an invalid field is specified it
182 is an error. For example:
184 perf script -v -F sw:comm,tid,trace
185 'trace' not valid for software events.
187 At this point usage is displayed, and perf-script exits.
189 The flags field is synthesized and may have a value when Instruction
190 Trace decoding. The flags are "bcrosyiABExgh" which stand for branch,
191 call, return, conditional, system, asynchronous, interrupt,
192 transaction abort, trace begin, trace end, in transaction, VM-Entry, and VM-Exit
193 respectively. Known combinations of flags are printed more nicely e.g.
194 "call" for "bc", "return" for "br", "jcc" for "bo", "jmp" for "b",
195 "int" for "bci", "iret" for "bri", "syscall" for "bcs", "sysret" for "brs",
196 "async" for "by", "hw int" for "bcyi", "tx abrt" for "bA", "tr strt" for "bB",
197 "tr end" for "bE", "vmentry" for "bcg", "vmexit" for "bch".
198 However the "x" flag will be displayed separately in those
199 cases e.g. "jcc (x)" for a condition branch within a transaction.
201 The callindent field is synthesized and may have a value when
202 Instruction Trace decoding. For calls and returns, it will display the
203 name of the symbol indented with spaces to reflect the stack depth.
205 When doing instruction trace decoding insn and insnlen give the
206 instruction bytes and the instruction length of the current
209 The synth field is used by synthesized events which may be created when
210 Instruction Trace decoding.
212 The ipc (instructions per cycle) field is synthesized and may have a value when
213 Instruction Trace decoding.
215 Finally, a user may not set fields to none for all event types.
216 i.e., -F "" is not allowed.
218 The brstack output includes branch related information with raw addresses using the
219 /v/v/v/v/cycles syntax in the following order:
220 FROM: branch source instruction
221 TO : branch target instruction
222 M/P/-: M=branch target mispredicted or branch direction was mispredicted, P=target predicted or direction predicted, -=not supported
223 X/- : X=branch inside a transactional region, -=not in transaction region or not supported
224 A/- : A=TSX abort entry, -=not aborted region or not supported
227 The brstacksym is identical to brstack, except that the FROM and TO addresses are printed in a symbolic form if possible.
229 When brstackinsn is specified the full assembler sequences of branch sequences for each sample
230 is printed. This is the full execution path leading to the sample. This is only supported when the
231 sample was recorded with perf record -b or -j any.
233 The brstackoff field will print an offset into a specific dso/binary.
235 With the metric option perf script can compute metrics for
236 sampling periods, similar to perf stat. This requires
237 specifying a group with multiple events defining metrics with the :S option
238 for perf record. perf will sample on the first event, and
239 print computed metrics for all the events in the group. Please note
240 that the metric computed is averaged over the whole sampling
241 period (since the last sample), not just for the sample point.
243 For sample events it's possible to display misc field with -F +misc option,
244 following letters are displayed for each bit:
246 PERF_RECORD_MISC_KERNEL K
247 PERF_RECORD_MISC_USER U
248 PERF_RECORD_MISC_HYPERVISOR H
249 PERF_RECORD_MISC_GUEST_KERNEL G
250 PERF_RECORD_MISC_GUEST_USER g
251 PERF_RECORD_MISC_MMAP_DATA* M
252 PERF_RECORD_MISC_COMM_EXEC E
253 PERF_RECORD_MISC_SWITCH_OUT S
254 PERF_RECORD_MISC_SWITCH_OUT_PREEMPT Sp
256 $ perf script -F +misc ...
257 sched-messaging 1414 K 28690.636582: 4590 cycles ...
258 sched-messaging 1407 U 28690.636600: 325620 cycles ...
259 sched-messaging 1414 K 28690.636608: 19473 cycles ...
260 misc field ___________/
269 --symfs=<directory>::
270 Look for files with symbols relative to this directory.
274 When printing symbols do not display call chain.
277 Stop display of callgraph at these symbols
280 --cpu:: Only report samples for the list of CPUs provided. Multiple CPUs can
281 be provided as a comma-separated list with no space: 0,1. Ranges of
282 CPUs are specified with -: 0-2. Default is to report samples on all
287 Only display events for these comms. CSV that understands
288 file://filename entries.
291 Only show events for given process ID (comma separated list).
294 Only show events for given thread ID (comma separated list).
298 Display extended information about the perf.data file. This adds
299 information which may be very large and thus may clutter the display.
300 It currently includes: cpu and numa topology of the host system.
301 It can only be used with the perf script report mode.
304 Try to resolve the path of [kernel.kallsyms]
307 Display task related events (e.g. FORK, COMM, EXIT).
310 Display mmap related events (e.g. MMAP, MMAP2).
312 --show-namespace-events
313 Display namespace events i.e. events of type PERF_RECORD_NAMESPACES.
316 Display context switch events i.e. events of type PERF_RECORD_SWITCH or
317 PERF_RECORD_SWITCH_CPU_WIDE.
320 Display lost events i.e. events of type PERF_RECORD_LOST.
323 Display finished round events i.e. events of type PERF_RECORD_FINISHED_ROUND.
326 Display bpf events i.e. events of type PERF_RECORD_KSYMBOL and PERF_RECORD_BPF_EVENT.
329 Display cgroup events i.e. events of type PERF_RECORD_CGROUP.
331 --show-text-poke-events
332 Display text poke events i.e. events of type PERF_RECORD_TEXT_POKE and
336 Demangle symbol names to human readable form. It's enabled by default,
337 disable with --no-demangle.
340 Demangle kernel symbol names to human readable form (for C++ kernels).
343 Show perf.data header.
346 Show only perf.data header.
349 Options for decoding instruction tracing data. The options are:
351 include::itrace.txt[]
353 To disable decoding entirely, use --no-itrace.
356 Show the full path for source files for srcline output.
359 Set the stack depth limit when parsing the callchain, anything
360 beyond the specified depth will be ignored. This is a trade-off
361 between information loss and faster processing especially for
362 workloads that can have a very long callchain stack.
363 Note that when using the --itrace option the synthesized callchain size
364 will override this value if the synthesized callchain size is bigger.
369 Use 9 decimal places when displaying time (i.e. show the nanoseconds)
373 Don't do ownership validation.
376 Only analyze samples within given time window: <start>,<stop>. Times
377 have the format seconds.nanoseconds. If start is not given (i.e. time
378 string is ',x.y') then analysis starts at the beginning of the file. If
379 stop time is not given (i.e. time string is 'x.y,') then analysis goes
380 to end of file. Multiple ranges can be separated by spaces, which
381 requires the argument to be quoted e.g. --time "1234.567,1234.789 1235,"
383 Also support time percent with multiple time ranges. Time string is
384 'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.
387 Select the second 10% time slice:
388 perf script --time 10%/2
390 Select from 0% to 10% time slice:
391 perf script --time 0%-10%
393 Select the first and second 10% time slices:
394 perf script --time 10%/1,10%/2
396 Select from 0% to 10% and 30% to 40% slices:
397 perf script --time 0%-10%,30%-40%
400 Set the maximum number of program blocks to print with brstackinsn for
404 Print time stamps relative to trace start.
407 Print time stamps relative to previous event.
410 Create per event files with a "perf.data.EVENT.dump" name instead of
411 printing to stdout, useful, for instance, for generating flamegraphs.
414 If a callgraph address belongs to an inlined function, the inline stack
415 will be printed. Each entry has function name and file/line. Enabled by
416 default, disable with --no-inline.
419 Show instruction stream for intel_pt traces. Combine with --xed to
423 Run xed disassembler on output. Requires installing the xed disassembler.
426 --symbols=symbol[,symbol...]::
427 Only consider the listed symbols. Symbols are typically a name
428 but they may also be hexadecimal address.
430 The hexadecimal address may be the start address of a symbol or
431 any other address to filter the trace records
433 For example, to select the symbol noploop or the address 0x4007a0:
434 perf script --symbols=noploop,0x4007a0
436 Support filtering trace records by symbol name, start address of
437 symbol, any hexadecimal address and address range.
439 The comparison order is:
441 1. symbol name comparison
442 2. symbol start address comparison.
443 3. any hexadecimal address comparison.
444 4. address range comparison (see --addr-range).
447 Use with -S or --symbols to list traced records within address range.
449 For example, to list the traced records within the address range
450 [0x4007a0, 0x0x4007a9]:
451 perf script -S 0x4007a0 --addr-range 10
454 Only consider symbols in these DSOs.
457 Show call stream for intel_pt traces. The CPUs are interleaved, but
458 can be filtered with -C.
461 Show call and return stream for intel_pt traces.
464 For itrace only show specified functions and their callees for
465 itrace. Multiple functions can be separated by comma.
467 --switch-on EVENT_NAME::
468 Only consider events after this event is found.
470 --switch-off EVENT_NAME::
471 Stop considering events after this event is found.
473 --show-on-off-events::
474 Show the --switch-on/off events too.
477 Show callgraph with stitched LBRs, which may have more complete
478 callgraph. The perf.data file must have been obtained using
479 perf record --call-graph lbr.
480 Disabled by default. In common cases with call stack overflows,
481 it can recreate better call stacks than the default lbr call stack
482 output. But this approach is not full proof. There can be cases
483 where it creates incorrect call stacks from incorrect matches.
484 The known limitations include exception handing such as
485 setjmp/longjmp will have calls/returns not match.
489 linkperf:perf-record[1], linkperf:perf-script-perl[1],
490 linkperf:perf-script-python[1], linkperf:perf-intel-pt[1],
491 linkperf:perf-dlfilter[1]