1PERF-SCRIPT(1) perf Manual PERF-SCRIPT(1)
2
6 perf-script - Read perf.data (created by perf record) and display trace
7 output
8
10 perf script [<options>]
11 perf script [<options>] record <script> [<record-options>] <command>
12 perf script [<options>] report <script> [script-args]
13 perf script [<options>] <script> <required-script-args> [<record-options>] <command>
14 perf script [<options>] <top-script> [script-args]
15
17 This command reads the input file and displays the trace recorded.
18 There are several variants of perf script:
20 'perf script' to see a detailed trace of the workload that was
22 recorded.
23 You can also run a set of pre-canned scripts that aggregate and
25 summarize the raw trace data in various ways (the list of scripts is
26 available via 'perf script -l'). The following variants allow you to
27 record and run those scripts:
28 'perf script record <script> <command>' to record the events required
30 for 'perf script report'. <script> is the name displayed in the
31 output of 'perf script --list' i.e. the actual script name minus any
32 language extension. If <command> is not specified, the events are
33 recorded using the -a (system-wide) 'perf record' option.
34 'perf script report <script> [args]' to run and display the results
36 of <script>. <script> is the name displayed in the output of 'perf
37 script --list' i.e. the actual script name minus any language
38 extension. The perf.data output from a previous run of 'perf script
39 record <script>' is used and should be present for this command to
40 succeed. [args] refers to the (mainly optional) args expected by
41 the script.
42 'perf script <script> <required-script-args> <command>' to both
44 record the events required for <script> and to run the <script>
45 using 'live-mode' i.e. without writing anything to disk. <script>
46 is the name displayed in the output of 'perf script --list' i.e. the
47 actual script name minus any language extension. If <command> is
48 not specified, the events are recorded using the -a (system-wide)
49 'perf record' option. If <script> has any required args, they
50 should be specified before <command>. This mode doesn't allow for
51 optional script args to be specified; if optional script args are
52 desired, they can be specified using separate 'perf script record'
53 and 'perf script report' commands, with the stdout of the record step
54 piped to the stdin of the report script, using the '-o -' and '-i -'
55 options of the corresponding commands.
56 'perf script <top-script>' to both record the events required for
58 <top-script> and to run the <top-script> using 'live-mode'
59 i.e. without writing anything to disk. <top-script> is the name
60 displayed in the output of 'perf script --list' i.e. the actual
61 script name minus any language extension; a <top-script> is defined
62 as any script name ending with the string 'top'.
63 [<record-options>] can be passed to the record steps of 'perf script
65 record' and 'live-mode' variants; this isn't possible however for
66 <top-script> 'live-mode' or 'perf script report' variants.
67 See the 'SEE ALSO' section for links to language-specific
69 information on how to write and run your own trace scripts.
70
72 <command>...
73 Any command you can specify in a shell.
74 -D, --dump-raw-trace=
76 Display verbose dump of the trace data.
77 -L, --Latency=
79 Show latency attributes (irqs/preemption disabled, etc).
80 -l, --list=
82 Display a list of available trace scripts.
83 -s [lang], --script=
85 Process trace data with the given script ([lang]:script[.ext]). If
86 the string lang is specified in place of a script name, a list of
87 supported languages will be displayed instead.
88 -g, --gen-script=
90 Generate perf-script.[ext] starter script for given language, using
91 current perf.data.
92 -a
94 Force system-wide collection. Scripts run without a <command>
95 normally use -a by default, while scripts run with a <command>
96 normally don’t - this option allows the latter to be run in
97 system-wide mode.
98 -i, --input=
100 Input file name. (default: perf.data unless stdin is a fifo)
101 -d, --debug-mode
103 Do various checks like samples ordering and lost events.
104 -F, --fields
106 Comma separated list of fields to print. Options are: comm, tid,
107 pid, time, cpu, event, trace, ip, sym, dso, addr, symoff, srcline,
108 period, iregs, uregs, brstack, brstacksym, flags, bpf-output,
109 brstackinsn, brstackoff, callindent, insn, insnlen, synth,
110 phys_addr, metric, misc, srccode, ipc, data_page_size,
111 code_page_size. Field list can be prepended with the type, trace,
112 sw or hw, to indicate to which event type the field list applies.
113 e.g., -F sw:comm,tid,time,ip,sym and -F trace:time,cpu,trace
114 perf script -F <fields>
116 is equivalent to:
118 perf script -F trace:<fields> -F sw:<fields> -F hw:<fields>
120 i.e., the specified fields apply to all event types if the type string
122 is not given.
123 In addition to overriding fields, it is also possible to add or remove
125 fields from the defaults. For example
126 -F -cpu,+insn
128 removes the cpu field and adds the insn field. Adding/removing fields
130 cannot be mixed with normal overriding.
131 The arguments are processed in the order received. A later usage can
133 reset a prior request. e.g.:
134 -F trace: -F comm,tid,time,ip,sym
136 The first -F suppresses trace events (field list is ""), but then the
138 second invocation sets the fields to comm,tid,time,ip,sym. In this case a
139 warning is given to the user:
140 "Overriding previous field request for all events."
142 Alternatively, consider the order:
144 -F comm,tid,time,ip,sym -F trace:
146 The first -F sets the fields for all events and the second -F
148 suppresses trace events. The user is given a warning message about
149 the override, and the result of the above is that only S/W and H/W
150 events are displayed with the given fields.
151 It's possible tp add/remove fields only for specific event type:
153 -Fsw:-cpu,-period
155 removes cpu and period from software events.
157 For the 'wildcard' option if a user selected field is invalid for an
159 event type, a message is displayed to the user that the option is
160 ignored for that type. For example:
161 $ perf script -F comm,tid,trace
163 'trace' not valid for hardware events. Ignoring.
164 'trace' not valid for software events. Ignoring.
165 Alternatively, if the type is given an invalid field is specified it
167 is an error. For example:
168 perf script -v -F sw:comm,tid,trace
170 'trace' not valid for software events.
171 At this point usage is displayed, and perf-script exits.
173 The flags field is synthesized and may have a value when Instruction
175 Trace decoding. The flags are "bcrosyiABEx" which stand for branch,
176 call, return, conditional, system, asynchronous, interrupt,
177 transaction abort, trace begin, trace end, and in transaction,
178 respectively. Known combinations of flags are printed more nicely e.g.
179 "call" for "bc", "return" for "br", "jcc" for "bo", "jmp" for "b",
180 "int" for "bci", "iret" for "bri", "syscall" for "bcs", "sysret" for "brs",
181 "async" for "by", "hw int" for "bcyi", "tx abrt" for "bA", "tr strt" for "bB",
182 "tr end" for "bE". However the "x" flag will be display separately in those
183 cases e.g. "jcc (x)" for a condition branch within a transaction.
184 The callindent field is synthesized and may have a value when
186 Instruction Trace decoding. For calls and returns, it will display the
187 name of the symbol indented with spaces to reflect the stack depth.
188 When doing instruction trace decoding insn and insnlen give the
190 instruction bytes and the instruction length of the current
191 instruction.
192 The synth field is used by synthesized events which may be created when
194 Instruction Trace decoding.
195 The ipc (instructions per cycle) field is synthesized and may have a value when
197 Instruction Trace decoding.
198 Finally, a user may not set fields to none for all event types.
200 i.e., -F "" is not allowed.
201 The brstack output includes branch related information with raw addresses using the
203 /v/v/v/v/cycles syntax in the following order:
204 FROM: branch source instruction
205 TO : branch target instruction
206 M/P/-: M=branch target mispredicted or branch direction was mispredicted, P=target predicted or direction predicted, -=not supported
207 X/- : X=branch inside a transactional region, -=not in transaction region or not supported
208 A/- : A=TSX abort entry, -=not aborted region or not supported
209 cycles
210 The brstacksym is identical to brstack, except that the FROM and TO addresses are printed in a symbolic form if possible.
212 When brstackinsn is specified the full assembler sequences of branch sequences for each sample
214 is printed. This is the full execution path leading to the sample. This is only supported when the
215 sample was recorded with perf record -b or -j any.
216 The brstackoff field will print an offset into a specific dso/binary.
218 With the metric option perf script can compute metrics for
220 sampling periods, similar to perf stat. This requires
221 specifying a group with multiple events defining metrics with the :S option
222 for perf record. perf will sample on the first event, and
223 print computed metrics for all the events in the group. Please note
224 that the metric computed is averaged over the whole sampling
225 period (since the last sample), not just for the sample point.
226 For sample events it's possible to display misc field with -F +misc option,
228 following letters are displayed for each bit:
229 PERF_RECORD_MISC_KERNEL K
231 PERF_RECORD_MISC_USER U
232 PERF_RECORD_MISC_HYPERVISOR H
233 PERF_RECORD_MISC_GUEST_KERNEL G
234 PERF_RECORD_MISC_GUEST_USER g
235 PERF_RECORD_MISC_MMAP_DATA* M
236 PERF_RECORD_MISC_COMM_EXEC E
237 PERF_RECORD_MISC_SWITCH_OUT S
238 PERF_RECORD_MISC_SWITCH_OUT_PREEMPT Sp
239 $ perf script -F +misc ...
241 sched-messaging 1414 K 28690.636582: 4590 cycles ...
242 sched-messaging 1407 U 28690.636600: 325620 cycles ...
243 sched-messaging 1414 K 28690.636608: 19473 cycles ...
244 misc field ___________/
245 -k, --vmlinux=<file>
247 vmlinux pathname
248 --kallsyms=<file>
250 kallsyms pathname
251 --symfs=<directory>
253 Look for files with symbols relative to this directory.
254 -G, --hide-call-graph
256 When printing symbols do not display call chain.
257 --stop-bt
259 Stop display of callgraph at these symbols
260 -C, --cpu
262 Only report samples for the list of CPUs provided. Multiple CPUs
263 can be provided as a comma-separated list with no space: 0,1.
264 Ranges of CPUs are specified with -: 0-2. Default is to report
265 samples on all CPUs.
266 -c, --comms=
268 Only display events for these comms. CSV that understands
269 file://filename entries.
270 --pid=
272 Only show events for given process ID (comma separated list).
273 --tid=
275 Only show events for given thread ID (comma separated list).
276 -I, --show-info
278 Display extended information about the perf.data file. This adds
279 information which may be very large and thus may clutter the
280 display. It currently includes: cpu and numa topology of the host
281 system. It can only be used with the perf script report mode.
282 --show-kernel-path
284 Try to resolve the path of [kernel.kallsyms]
285 --show-task-events Display task related events (e.g. FORK, COMM, EXIT).
287 --show-mmap-events Display mmap related events (e.g. MMAP, MMAP2).
289 --show-namespace-events Display namespace events i.e. events of type
291 PERF_RECORD_NAMESPACES.
292 --show-switch-events Display context switch events i.e. events of type
294 PERF_RECORD_SWITCH or PERF_RECORD_SWITCH_CPU_WIDE.
295 --show-lost-events Display lost events i.e. events of type
297 PERF_RECORD_LOST.
298 --show-round-events Display finished round events i.e. events of type
300 PERF_RECORD_FINISHED_ROUND.
301 --show-bpf-events Display bpf events i.e. events of type
303 PERF_RECORD_KSYMBOL and PERF_RECORD_BPF_EVENT.
304 --show-cgroup-events Display cgroup events i.e. events of type
306 PERF_RECORD_CGROUP.
307 --show-text-poke-events Display text poke events i.e. events of type
309 PERF_RECORD_TEXT_POKE and PERF_RECORD_KSYMBOL.
310 --demangle
312 Demangle symbol names to human readable form. It’s enabled by
313 default, disable with --no-demangle.
314 --demangle-kernel
316 Demangle kernel symbol names to human readable form (for C++
317 kernels).
318 --header Show perf.data header.
320 --header-only Show only perf.data header.
322 --itrace
324 Options for decoding instruction tracing data. The options are:
325 i synthesize instructions events
327 b synthesize branches events (branch misses for Arm SPE)
328 c synthesize branches events (calls only)
329 r synthesize branches events (returns only)
330 x synthesize transactions events
331 w synthesize ptwrite events
332 p synthesize power events (incl. PSB events for Intel PT)
333 o synthesize other events recorded due to the use
334 of aux-output (refer to perf record)
335 e synthesize error events
336 d create a debug log
337 f synthesize first level cache events
338 m synthesize last level cache events
339 M synthesize memory events
340 t synthesize TLB events
341 a synthesize remote access events
342 g synthesize a call chain (use with i or x)
343 G synthesize a call chain on existing event records
344 l synthesize last branch entries (use with i or x)
345 L synthesize last branch entries on existing event records
346 s skip initial number of events
347 q quicker (less detailed) decoding
348 The default is all events i.e. the same as --itrace=ibxwpe,
350 except for perf script where it is --itrace=ce
351 In addition, the period (default 100000, except for perf script where it is 1)
353 for instructions events can be specified in units of:
354 i instructions
356 t ticks
357 ms milliseconds
358 us microseconds
359 ns nanoseconds (default)
360 Also the call chain size (default 16, max. 1024) for instructions or
362 transactions events can be specified.
363 Also the number of last branch entries (default 64, max. 1024) for
365 instructions or transactions events can be specified.
366 Similar to options g and l, size may also be specified for options G and L.
368 On x86, note that G and L work poorly when data has been recorded with
369 large PEBS. Refer linkperf:perf-intel-pt[1] man page for details.
370 It is also possible to skip events generated (instructions, branches, transactions,
372 ptwrite, power) at the beginning. This is useful to ignore initialization code.
373 --itrace=i0nss1000000
375 skips the first million instructions.
377 The 'e' option may be followed by flags which affect what errors will or
379 will not be reported. Each flag must be preceded by either '+' or '-'.
380 The flags are:
381 o overflow
382 l trace data lost
383 If supported, the 'd' option may be followed by flags which affect what
385 debug messages will or will not be logged. Each flag must be preceded
386 by either '+' or '-'. The flags are:
387 a all perf events
388 If supported, the 'q' option may be repeated to increase the effect.
390 To disable decoding entirely, use --no-itrace.
392 --full-source-path
394 Show the full path for source files for srcline output.
395 --max-stack
397 Set the stack depth limit when parsing the callchain, anything
398 beyond the specified depth will be ignored. This is a trade-off
399 between information loss and faster processing especially for
400 workloads that can have a very long callchain stack. Note that when
401 using the --itrace option the synthesized callchain size will
402 override this value if the synthesized callchain size is bigger.
403 Default: 127
405 --ns
407 Use 9 decimal places when displaying time (i.e. show the
408 nanoseconds)
409 -f, --force
411 Don’t do ownership validation.
412 --time
414 Only analyze samples within given time window: <start>,<stop>.
415 Times have the format seconds.nanoseconds. If start is not given
416 (i.e. time string is ,x.y) then analysis starts at the beginning of
417 the file. If stop time is not given (i.e. time string is x.y,) then
418 analysis goes to end of file. Multiple ranges can be separated by
419 spaces, which requires the argument to be quoted e.g. --time
420 "1234.567,1234.789 1235,"
421 Also support time percent with multiple time ranges. Time string is
423 'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.
424 For example:
426 Select the second 10% time slice:
427 perf script --time 10%/2
428 Select from 0% to 10% time slice:
430 perf script --time 0%-10%
431 Select the first and second 10% time slices:
433 perf script --time 10%/1,10%/2
434 Select from 0% to 10% and 30% to 40% slices:
436 perf script --time 0%-10%,30%-40%
437 --max-blocks
439 Set the maximum number of program blocks to print with brstackinsn
440 for each sample.
441 --reltime
443 Print time stamps relative to trace start.
444 --deltatime
446 Print time stamps relative to previous event.
447 --per-event-dump
449 Create per event files with a "perf.data.EVENT.dump" name instead
450 of printing to stdout, useful, for instance, for generating
451 flamegraphs.
452 --inline
454 If a callgraph address belongs to an inlined function, the inline
455 stack will be printed. Each entry has function name and file/line.
456 Enabled by default, disable with --no-inline.
457 --insn-trace
459 Show instruction stream for intel_pt traces. Combine with --xed to
460 show disassembly.
461 --xed
463 Run xed disassembler on output. Requires installing the xed
464 disassembler.
465 -S, --symbols=symbol[,symbol...]
467 Only consider the listed symbols. Symbols are typically a name but
468 they may also be hexadecimal address.
469 The hexadecimal address may be the start address of a symbol or
471 any other address to filter the trace records
472 For example, to select the symbol noploop or the address 0x4007a0:
474 perf script --symbols=noploop,0x4007a0
475 Support filtering trace records by symbol name, start address of
477 symbol, any hexadecimal address and address range.
478 The comparison order is:
480 1. symbol name comparison
482 2. symbol start address comparison.
484 3. any hexadecimal address comparison.
486 4. address range comparison (see --addr-range).
488 --addr-range
490 Use with -S or --symbols to list traced records within address
491 range.
492 For example, to list the traced records within the address range
494 [0x4007a0, 0x0x4007a9]:
495 perf script -S 0x4007a0 --addr-range 10
496 --dsos=
498 Only consider symbols in these DSOs.
499 --call-trace
501 Show call stream for intel_pt traces. The CPUs are interleaved, but
502 can be filtered with -C.
503 --call-ret-trace
505 Show call and return stream for intel_pt traces.
506 --graph-function
508 For itrace only show specified functions and their callees for
509 itrace. Multiple functions can be separated by comma.
510 --switch-on EVENT_NAME
512 Only consider events after this event is found.
513 --switch-off EVENT_NAME
515 Stop considering events after this event is found.
516 --show-on-off-events
518 Show the --switch-on/off events too.
519 --stitch-lbr
521 Show callgraph with stitched LBRs, which may have more complete
522 callgraph. The perf.data file must have been obtained using perf
523 record --call-graph lbr. Disabled by default. In common cases with
524 call stack overflows, it can recreate better call stacks than the
525 default lbr call stack output. But this approach is not full proof.
526 There can be cases where it creates incorrect call stacks from
527 incorrect matches. The known limitations include exception handing
528 such as setjmp/longjmp will have calls/returns not match.
529
531 perf-record(1), perf-script-perl(1), perf-script-python(1), perf-intel-
532 pt(1)
533perf 06/03/2021 PERF-SCRIPT(1)