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 --dump-unsorted-raw-trace=
79 Same as --dump-raw-trace but not sorted in time order.
80 -L, --Latency=
82 Show latency attributes (irqs/preemption disabled, etc).
83 -l, --list=
85 Display a list of available trace scripts.
86 -s [lang], --script=
88 Process trace data with the given script ([lang]:script[.ext]). If
89 the string lang is specified in place of a script name, a list of
90 supported languages will be displayed instead.
91 -g, --gen-script=
93 Generate perf-script.[ext] starter script for given language, using
94 current perf.data.
95 --dlfilter=<file>
97 Filter sample events using the given shared object file. Refer
98 perf-dlfilter(1)
99 --dlarg=<arg>
101 Pass arg as an argument to the dlfilter. --dlarg may be repeated to
102 add more arguments.
103 --list-dlfilters
105 Display a list of available dlfilters. Use with option -v (must
106 come before option --list-dlfilters) to show long descriptions.
107 -a
109 Force system-wide collection. Scripts run without a <command>
110 normally use -a by default, while scripts run with a <command>
111 normally don’t - this option allows the latter to be run in
112 system-wide mode.
113 -i, --input=
115 Input file name. (default: perf.data unless stdin is a fifo)
116 -d, --debug-mode
118 Do various checks like samples ordering and lost events.
119 -F, --fields
121 Comma separated list of fields to print. Options are: comm, tid,
122 pid, time, cpu, event, trace, ip, sym, dso, addr, symoff, srcline,
123 period, iregs, uregs, brstack, brstacksym, flags, bpf-output,
124 brstackinsn, brstackinsnlen, brstackoff, callindent, insn, insnlen,
125 synth, phys_addr, metric, misc, srccode, ipc, data_page_size,
126 code_page_size, ins_lat, machine_pid, vcpu. Field list can be
127 prepended with the type, trace, sw or hw, to indicate to which
128 event type the field list applies. e.g., -F sw:comm,tid,time,ip,sym
129 and -F trace:time,cpu,trace
130 perf script -F <fields>
132 is equivalent to:
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
138 is not given.
139 In addition to overriding fields, it is also possible to add or remove
141 fields from the defaults. For example
142 -F -cpu,+insn
144 removes the cpu field and adds the insn field. Adding/removing fields
146 cannot be mixed with normal overriding.
147 The arguments are processed in the order received. A later usage can
149 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
154 second invocation sets the fields to comm,tid,time,ip,sym. In this case a
155 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
164 suppresses trace events. The user is given a warning message about
165 the override, and the result of the above is that only S/W and H/W
166 events are displayed with the given fields.
167 It's possible tp add/remove fields only for specific event type:
169 -Fsw:-cpu,-period
171 removes cpu and period from software events.
173 For the 'wildcard' option if a user selected field is invalid for an
175 event type, a message is displayed to the user that the option is
176 ignored for that type. For example:
177 $ perf script -F comm,tid,trace
179 'trace' not valid for hardware events. Ignoring.
180 'trace' not valid for software events. Ignoring.
181 Alternatively, if the type is given an invalid field is specified it
183 is an error. For example:
184 perf script -v -F sw:comm,tid,trace
186 '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
191 Trace decoding. The flags are "bcrosyiABExghDt" which stand for branch,
192 call, return, conditional, system, asynchronous, interrupt,
193 transaction abort, trace begin, trace end, in transaction, VM-Entry,
194 VM-Exit, interrupt disabled and interrupt disable toggle respectively.
195 Known combinations of flags are printed more nicely e.g.
196 "call" for "bc", "return" for "br", "jcc" for "bo", "jmp" for "b",
197 "int" for "bci", "iret" for "bri", "syscall" for "bcs", "sysret" for "brs",
198 "async" for "by", "hw int" for "bcyi", "tx abrt" for "bA", "tr strt" for "bB",
199 "tr end" for "bE", "vmentry" for "bcg", "vmexit" for "bch".
200 However the "x", "D" and "t" flags will be displayed separately in those
201 cases e.g. "jcc (xD)" for a condition branch within a transaction
202 with interrupts disabled. Note, interrupts becoming disabled is "t",
203 whereas interrupts becoming enabled is "Dt".
204 The callindent field is synthesized and may have a value when
206 Instruction Trace decoding. For calls and returns, it will display the
207 name of the symbol indented with spaces to reflect the stack depth.
208 When doing instruction trace decoding insn and insnlen give the
210 instruction bytes and the instruction length of the current
211 instruction.
212 The synth field is used by synthesized events which may be created when
214 Instruction Trace decoding.
215 The ipc (instructions per cycle) field is synthesized and may have a value when
217 Instruction Trace decoding.
218 The machine_pid and vcpu fields are derived from data resulting from using
220 perf inject to insert a perf.data file recorded inside a virtual machine into
221 a perf.data file recorded on the host at the same time.
222 Finally, a user may not set fields to none for all event types.
224 i.e., -F "" is not allowed.
225 The brstack output includes branch related information with raw addresses using the
227 /v/v/v/v/cycles syntax in the following order:
228 FROM: branch source instruction
229 TO : branch target instruction
230 M/P/-: M=branch target mispredicted or branch direction was mispredicted, P=target predicted or direction predicted, -=not supported
231 X/- : X=branch inside a transactional region, -=not in transaction region or not supported
232 A/- : A=TSX abort entry, -=not aborted region or not supported
233 cycles
234 The brstacksym is identical to brstack, except that the FROM and TO addresses are printed in a symbolic form if possible.
236 When brstackinsn is specified the full assembler sequences of branch sequences for each sample
238 is printed. This is the full execution path leading to the sample. This is only supported when the
239 sample was recorded with perf record -b or -j any.
240 Use brstackinsnlen to print the brstackinsn lenght. For example, you
242 can’t know the next sequential instruction after an unconditional branch unless
243 you calculate that based on its length.
244 The brstackoff field will print an offset into a specific dso/binary.
246 With the metric option perf script can compute metrics for
248 sampling periods, similar to perf stat. This requires
249 specifying a group with multiple events defining metrics with the :S option
250 for perf record. perf will sample on the first event, and
251 print computed metrics for all the events in the group. Please note
252 that the metric computed is averaged over the whole sampling
253 period (since the last sample), not just for the sample point.
254 For sample events it's possible to display misc field with -F +misc option,
256 following letters are displayed for each bit:
257 PERF_RECORD_MISC_KERNEL K
259 PERF_RECORD_MISC_USER U
260 PERF_RECORD_MISC_HYPERVISOR H
261 PERF_RECORD_MISC_GUEST_KERNEL G
262 PERF_RECORD_MISC_GUEST_USER g
263 PERF_RECORD_MISC_MMAP_DATA* M
264 PERF_RECORD_MISC_COMM_EXEC E
265 PERF_RECORD_MISC_SWITCH_OUT S
266 PERF_RECORD_MISC_SWITCH_OUT_PREEMPT Sp
267 $ perf script -F +misc ...
269 sched-messaging 1414 K 28690.636582: 4590 cycles ...
270 sched-messaging 1407 U 28690.636600: 325620 cycles ...
271 sched-messaging 1414 K 28690.636608: 19473 cycles ...
272 misc field ___________/
273 -k, --vmlinux=<file>
275 vmlinux pathname
276 --kallsyms=<file>
278 kallsyms pathname
279 --symfs=<directory>
281 Look for files with symbols relative to this directory.
282 -G, --hide-call-graph
284 When printing symbols do not display call chain.
285 --stop-bt
287 Stop display of callgraph at these symbols
288 -C, --cpu
290 Only report samples for the list of CPUs provided. Multiple CPUs
291 can be provided as a comma-separated list with no space: 0,1.
292 Ranges of CPUs are specified with -: 0-2. Default is to report
293 samples on all CPUs.
294 -c, --comms=
296 Only display events for these comms. CSV that understands
297 file://filename entries.
298 --pid=
300 Only show events for given process ID (comma separated list).
301 --tid=
303 Only show events for given thread ID (comma separated list).
304 -I, --show-info
306 Display extended information about the perf.data file. This adds
307 information which may be very large and thus may clutter the
308 display. It currently includes: cpu and numa topology of the host
309 system. It can only be used with the perf script report mode.
310 --show-kernel-path
312 Try to resolve the path of [kernel.kallsyms]
313 --show-task-events Display task related events (e.g. FORK, COMM, EXIT).
315 --show-mmap-events Display mmap related events (e.g. MMAP, MMAP2).
317 --show-namespace-events Display namespace events i.e. events of type
319 PERF_RECORD_NAMESPACES.
320 --show-switch-events Display context switch events i.e. events of type
322 PERF_RECORD_SWITCH or PERF_RECORD_SWITCH_CPU_WIDE.
323 --show-lost-events Display lost events i.e. events of type
325 PERF_RECORD_LOST.
326 --show-round-events Display finished round events i.e. events of type
328 PERF_RECORD_FINISHED_ROUND.
329 --show-bpf-events Display bpf events i.e. events of type
331 PERF_RECORD_KSYMBOL and PERF_RECORD_BPF_EVENT.
332 --show-cgroup-events Display cgroup events i.e. events of type
334 PERF_RECORD_CGROUP.
335 --show-text-poke-events Display text poke events i.e. events of type
337 PERF_RECORD_TEXT_POKE and PERF_RECORD_KSYMBOL.
338 --demangle
340 Demangle symbol names to human readable form. It’s enabled by
341 default, disable with --no-demangle.
342 --demangle-kernel
344 Demangle kernel symbol names to human readable form (for C++
345 kernels).
346 --header Show perf.data header.
348 --header-only Show only perf.data header.
350 --itrace
352 Options for decoding instruction tracing data. The options are:
353 i synthesize instructions events
355 b synthesize branches events (branch misses for Arm SPE)
356 c synthesize branches events (calls only)
357 r synthesize branches events (returns only)
358 x synthesize transactions events
359 w synthesize ptwrite events
360 p synthesize power events (incl. PSB events for Intel PT)
361 o synthesize other events recorded due to the use
362 of aux-output (refer to perf record)
363 I synthesize interrupt or similar (asynchronous) events
364 (e.g. Intel PT Event Trace)
365 e synthesize error events
366 d create a debug log
367 f synthesize first level cache events
368 m synthesize last level cache events
369 M synthesize memory events
370 t synthesize TLB events
371 a synthesize remote access events
372 g synthesize a call chain (use with i or x)
373 G synthesize a call chain on existing event records
374 l synthesize last branch entries (use with i or x)
375 L synthesize last branch entries on existing event records
376 s skip initial number of events
377 q quicker (less detailed) decoding
378 A approximate IPC
379 Z prefer to ignore timestamps (so-called "timeless" decoding)
380 The default is all events i.e. the same as --itrace=ibxwpe,
382 except for perf script where it is --itrace=ce
383 In addition, the period (default 100000, except for perf script where it is 1)
385 for instructions events can be specified in units of:
386 i instructions
388 t ticks
389 ms milliseconds
390 us microseconds
391 ns nanoseconds (default)
392 Also the call chain size (default 16, max. 1024) for instructions or
394 transactions events can be specified.
395 Also the number of last branch entries (default 64, max. 1024) for
397 instructions or transactions events can be specified.
398 Similar to options g and l, size may also be specified for options G and L.
400 On x86, note that G and L work poorly when data has been recorded with
401 large PEBS. Refer linkperf:perf-intel-pt[1] man page for details.
402 It is also possible to skip events generated (instructions, branches, transactions,
404 ptwrite, power) at the beginning. This is useful to ignore initialization code.
405 --itrace=i0nss1000000
407 skips the first million instructions.
409 The 'e' option may be followed by flags which affect what errors will or
411 will not be reported. Each flag must be preceded by either '+' or '-'.
412 The flags are:
413 o overflow
414 l trace data lost
415 If supported, the 'd' option may be followed by flags which affect what
417 debug messages will or will not be logged. Each flag must be preceded
418 by either '+' or '-'. The flags are:
419 a all perf events
420 e output only on errors (size configurable - see linkperf:perf-config[1])
421 o output to stdout
422 If supported, the 'q' option may be repeated to increase the effect.
424 To disable decoding entirely, use --no-itrace.
426 --full-source-path
428 Show the full path for source files for srcline output.
429 --max-stack
431 Set the stack depth limit when parsing the callchain, anything
432 beyond the specified depth will be ignored. This is a trade-off
433 between information loss and faster processing especially for
434 workloads that can have a very long callchain stack. Note that when
435 using the --itrace option the synthesized callchain size will
436 override this value if the synthesized callchain size is bigger.
437 Default: 127
439 --ns
441 Use 9 decimal places when displaying time (i.e. show the
442 nanoseconds)
443 -f, --force
445 Don’t do ownership validation.
446 --time
448 Only analyze samples within given time window: <start>,<stop>.
449 Times have the format seconds.nanoseconds. If start is not given
450 (i.e. time string is ,x.y) then analysis starts at the beginning of
451 the file. If stop time is not given (i.e. time string is x.y,) then
452 analysis goes to end of file. Multiple ranges can be separated by
453 spaces, which requires the argument to be quoted e.g. --time
454 "1234.567,1234.789 1235,"
455 Also support time percent with multiple time ranges. Time string is
457 'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.
458 For example:
460 Select the second 10% time slice:
461 perf script --time 10%/2
462 Select from 0% to 10% time slice:
464 perf script --time 0%-10%
465 Select the first and second 10% time slices:
467 perf script --time 10%/1,10%/2
468 Select from 0% to 10% and 30% to 40% slices:
470 perf script --time 0%-10%,30%-40%
471 --max-blocks
473 Set the maximum number of program blocks to print with brstackinsn
474 for each sample.
475 --reltime
477 Print time stamps relative to trace start.
478 --deltatime
480 Print time stamps relative to previous event.
481 --per-event-dump
483 Create per event files with a "perf.data.EVENT.dump" name instead
484 of printing to stdout, useful, for instance, for generating
485 flamegraphs.
486 --inline
488 If a callgraph address belongs to an inlined function, the inline
489 stack will be printed. Each entry has function name and file/line.
490 Enabled by default, disable with --no-inline.
491 --insn-trace
493 Show instruction stream for intel_pt traces. Combine with --xed to
494 show disassembly.
495 --xed
497 Run xed disassembler on output. Requires installing the xed
498 disassembler.
499 -S, --symbols=symbol[,symbol...]
501 Only consider the listed symbols. Symbols are typically a name but
502 they may also be hexadecimal address.
503 The hexadecimal address may be the start address of a symbol or
505 any other address to filter the trace records
506 For example, to select the symbol noploop or the address 0x4007a0:
508 perf script --symbols=noploop,0x4007a0
509 Support filtering trace records by symbol name, start address of
511 symbol, any hexadecimal address and address range.
512 The comparison order is:
514 1. symbol name comparison
516 2. symbol start address comparison.
518 3. any hexadecimal address comparison.
520 4. address range comparison (see --addr-range).
522 --addr-range
524 Use with -S or --symbols to list traced records within address
525 range.
526 For example, to list the traced records within the address range
528 [0x4007a0, 0x0x4007a9]:
529 perf script -S 0x4007a0 --addr-range 10
530 --dsos=
532 Only consider symbols in these DSOs.
533 --call-trace
535 Show call stream for intel_pt traces. The CPUs are interleaved, but
536 can be filtered with -C.
537 --call-ret-trace
539 Show call and return stream for intel_pt traces.
540 --graph-function
542 For itrace only show specified functions and their callees for
543 itrace. Multiple functions can be separated by comma.
544 --switch-on EVENT_NAME
546 Only consider events after this event is found.
547 --switch-off EVENT_NAME
549 Stop considering events after this event is found.
550 --show-on-off-events
552 Show the --switch-on/off events too.
553 --stitch-lbr
555 Show callgraph with stitched LBRs, which may have more complete
556 callgraph. The perf.data file must have been obtained using perf
557 record --call-graph lbr. Disabled by default. In common cases with
558 call stack overflows, it can recreate better call stacks than the
559 default lbr call stack output. But this approach is not full proof.
560 There can be cases where it creates incorrect call stacks from
561 incorrect matches. The known limitations include exception handing
562 such as setjmp/longjmp will have calls/returns not match.
563 --guestmount=<path>
565 Guest OS root file system mount directory. Users mount guest OS
566 root directories under <path> by a specific filesystem access
567 method, typically, sshfs. For example, start 2 guest OS, one’s pid
568 is 8888 and the other’s is 9999:
569 $ mkdir ~/guestmount
571 $ cd ~/guestmount
572 $ sshfs -o allow_other,direct_io -p 5551 localhost:/ 8888/
573 $ sshfs -o allow_other,direct_io -p 5552 localhost:/ 9999/
574 $ perf script --guestmount=~/guestmount
575 --guestkallsyms=<path>
577 Guest OS /proc/kallsyms file copy. perf reads it to get guest
578 kernel symbols. Users copy it out from guest OS.
579 --guestmodules=<path>
581 Guest OS /proc/modules file copy. perf reads it to get guest kernel
582 module information. Users copy it out from guest OS.
583 --guestvmlinux=<path>
585 Guest OS kernel vmlinux.
586 --guest-code
588 Indicate that guest code can be found in the hypervisor process,
589 which is a common case for KVM test programs.
590
592 perf-record(1), perf-script-perl(1), perf-script-python(1), perf-intel-
593 pt(1), perf-dlfilter(1)
594perf 01/12/2023 PERF-SCRIPT(1)