| 1 | perf-script(1) |
| 2 | ============= |
| 3 | |
| 4 | NAME |
| 5 | ---- |
| 6 | perf-script - Read perf.data (created by perf record) and display trace output |
| 7 | |
| 8 | SYNOPSIS |
| 9 | -------- |
| 10 | [verse] |
| 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] |
| 16 | |
| 17 | DESCRIPTION |
| 18 | ----------- |
| 19 | This command reads the input file and displays the trace recorded. |
| 20 | |
| 21 | There are several variants of perf script: |
| 22 | |
| 23 | 'perf script' to see a detailed trace of the workload that was |
| 24 | recorded. |
| 25 | |
| 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: |
| 30 | |
| 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. |
| 36 | |
| 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 |
| 43 | the script. |
| 44 | |
| 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. |
| 58 | |
| 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'. |
| 65 | |
| 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. |
| 69 | |
| 70 | See the 'SEE ALSO' section for links to language-specific |
| 71 | information on how to write and run your own trace scripts. |
| 72 | |
| 73 | OPTIONS |
| 74 | ------- |
| 75 | <command>...:: |
| 76 | Any command you can specify in a shell. |
| 77 | |
| 78 | -D:: |
| 79 | --dump-raw-trace=:: |
| 80 | Display verbose dump of the trace data. |
| 81 | |
| 82 | --dump-unsorted-raw-trace=:: |
| 83 | Same as --dump-raw-trace but not sorted in time order. |
| 84 | |
| 85 | -L:: |
| 86 | --Latency=:: |
| 87 | Show latency attributes (irqs/preemption disabled, etc). |
| 88 | |
| 89 | -l:: |
| 90 | --list=:: |
| 91 | Display a list of available trace scripts. |
| 92 | |
| 93 | -s ['lang']:: |
| 94 | --script=:: |
| 95 | Process trace data with the given script ([lang]:script[.ext]). |
| 96 | If the string 'lang' is specified in place of a script name, a |
| 97 | list of supported languages will be displayed instead. |
| 98 | |
| 99 | -g:: |
| 100 | --gen-script=:: |
| 101 | Generate perf-script.[ext] starter script for given language, |
| 102 | using current perf.data. |
| 103 | |
| 104 | --dlfilter=<file>:: |
| 105 | Filter sample events using the given shared object file. |
| 106 | Refer linkperf:perf-dlfilter[1] |
| 107 | |
| 108 | --dlarg=<arg>:: |
| 109 | Pass 'arg' as an argument to the dlfilter. --dlarg may be repeated |
| 110 | to add more arguments. |
| 111 | |
| 112 | --list-dlfilters:: |
| 113 | Display a list of available dlfilters. Use with option -v (must come |
| 114 | before option --list-dlfilters) to show long descriptions. |
| 115 | |
| 116 | -a:: |
| 117 | Force system-wide collection. Scripts run without a <command> |
| 118 | normally use -a by default, while scripts run with a <command> |
| 119 | normally don't - this option allows the latter to be run in |
| 120 | system-wide mode. |
| 121 | |
| 122 | -i:: |
| 123 | --input=:: |
| 124 | Input file name. (default: perf.data unless stdin is a fifo) |
| 125 | |
| 126 | -d:: |
| 127 | --debug-mode:: |
| 128 | Do various checks like samples ordering and lost events. |
| 129 | |
| 130 | -F:: |
| 131 | --fields:: |
| 132 | Comma separated list of fields to print. Options are: |
| 133 | comm, tid, pid, time, cpu, event, trace, ip, sym, dso, dsoff, addr, symoff, |
| 134 | srcline, period, iregs, uregs, brstack, brstacksym, flags, bpf-output, |
| 135 | brstackinsn, brstackinsnlen, brstackdisasm, brstackoff, callindent, insn, disasm, |
| 136 | insnlen, synth, phys_addr, metric, misc, srccode, ipc, data_page_size, |
| 137 | code_page_size, ins_lat, machine_pid, vcpu, cgroup, retire_lat, brcntr, |
| 138 | |
| 139 | Field list can be prepended with the type, trace, sw or hw, |
| 140 | to indicate to which event type the field list applies. |
| 141 | e.g., -F sw:comm,tid,time,ip,sym and -F trace:time,cpu,trace |
| 142 | |
| 143 | perf script -F <fields> |
| 144 | |
| 145 | is equivalent to: |
| 146 | |
| 147 | perf script -F trace:<fields> -F sw:<fields> -F hw:<fields> |
| 148 | |
| 149 | i.e., the specified fields apply to all event types if the type string |
| 150 | is not given. |
| 151 | |
| 152 | In addition to overriding fields, it is also possible to add or remove |
| 153 | fields from the defaults. For example |
| 154 | |
| 155 | -F -cpu,+insn |
| 156 | |
| 157 | removes the cpu field and adds the insn field. Adding/removing fields |
| 158 | cannot be mixed with normal overriding. |
| 159 | |
| 160 | The arguments are processed in the order received. A later usage can |
| 161 | reset a prior request. e.g.: |
| 162 | |
| 163 | -F trace: -F comm,tid,time,ip,sym |
| 164 | |
| 165 | The first -F suppresses trace events (field list is ""), but then the |
| 166 | second invocation sets the fields to comm,tid,time,ip,sym. In this case a |
| 167 | warning is given to the user: |
| 168 | |
| 169 | "Overriding previous field request for all events." |
| 170 | |
| 171 | Alternatively, consider the order: |
| 172 | |
| 173 | -F comm,tid,time,ip,sym -F trace: |
| 174 | |
| 175 | The first -F sets the fields for all events and the second -F |
| 176 | suppresses trace events. The user is given a warning message about |
| 177 | the override, and the result of the above is that only S/W and H/W |
| 178 | events are displayed with the given fields. |
| 179 | |
| 180 | It's possible tp add/remove fields only for specific event type: |
| 181 | |
| 182 | -Fsw:-cpu,-period |
| 183 | |
| 184 | removes cpu and period from software events. |
| 185 | |
| 186 | For the 'wildcard' option if a user selected field is invalid for an |
| 187 | event type, a message is displayed to the user that the option is |
| 188 | ignored for that type. For example: |
| 189 | |
| 190 | $ perf script -F comm,tid,trace |
| 191 | 'trace' not valid for hardware events. Ignoring. |
| 192 | 'trace' not valid for software events. Ignoring. |
| 193 | |
| 194 | Alternatively, if the type is given an invalid field is specified it |
| 195 | is an error. For example: |
| 196 | |
| 197 | perf script -v -F sw:comm,tid,trace |
| 198 | 'trace' not valid for software events. |
| 199 | |
| 200 | At this point usage is displayed, and perf-script exits. |
| 201 | |
| 202 | The flags field is synthesized and may have a value when Instruction |
| 203 | Trace decoding. The flags are "bcrosyiABExghDt" which stand for branch, |
| 204 | call, return, conditional, system, asynchronous, interrupt, |
| 205 | transaction abort, trace begin, trace end, in transaction, VM-Entry, |
| 206 | VM-Exit, interrupt disabled and interrupt disable toggle respectively. |
| 207 | Known combinations of flags are printed more nicely e.g. |
| 208 | "call" for "bc", "return" for "br", "jcc" for "bo", "jmp" for "b", |
| 209 | "int" for "bci", "iret" for "bri", "syscall" for "bcs", "sysret" for "brs", |
| 210 | "async" for "by", "hw int" for "bcyi", "tx abrt" for "bA", "tr strt" for "bB", |
| 211 | "tr end" for "bE", "vmentry" for "bcg", "vmexit" for "bch". |
| 212 | However the "x", "D" and "t" flags will be displayed separately in those |
| 213 | cases e.g. "jcc (xD)" for a condition branch within a transaction |
| 214 | with interrupts disabled. Note, interrupts becoming disabled is "t", |
| 215 | whereas interrupts becoming enabled is "Dt". |
| 216 | |
| 217 | The callindent field is synthesized and may have a value when |
| 218 | Instruction Trace decoding. For calls and returns, it will display the |
| 219 | name of the symbol indented with spaces to reflect the stack depth. |
| 220 | |
| 221 | When doing instruction trace decoding, insn, disasm and insnlen give the |
| 222 | instruction bytes, disassembled instructions (requires libcapstone support) |
| 223 | and the instruction length of the current instruction respectively. |
| 224 | |
| 225 | The synth field is used by synthesized events which may be created when |
| 226 | Instruction Trace decoding. |
| 227 | |
| 228 | The ipc (instructions per cycle) field is synthesized and may have a value when |
| 229 | Instruction Trace decoding. |
| 230 | |
| 231 | The machine_pid and vcpu fields are derived from data resulting from using |
| 232 | perf inject to insert a perf.data file recorded inside a virtual machine into |
| 233 | a perf.data file recorded on the host at the same time. |
| 234 | |
| 235 | The cgroup fields requires sample having the cgroup id which is saved |
| 236 | when "--all-cgroups" option is passed to 'perf record'. |
| 237 | |
| 238 | Finally, a user may not set fields to none for all event types. |
| 239 | i.e., -F "" is not allowed. |
| 240 | |
| 241 | The brstack output includes branch related information with raw addresses using the |
| 242 | /v/v/v/v/cycles syntax in the following order: |
| 243 | FROM: branch source instruction |
| 244 | TO : branch target instruction |
| 245 | M/P/-: M=branch target mispredicted or branch direction was mispredicted, P=target predicted or direction predicted, -=not supported |
| 246 | X/- : X=branch inside a transactional region, -=not in transaction region or not supported |
| 247 | A/- : A=TSX abort entry, -=not aborted region or not supported |
| 248 | cycles |
| 249 | |
| 250 | The brstacksym is identical to brstack, except that the FROM and TO addresses are printed in a symbolic form if possible. |
| 251 | |
| 252 | When brstackinsn is specified the full assembler sequences of branch sequences for each sample |
| 253 | is printed. This is the full execution path leading to the sample. This is only supported when the |
| 254 | sample was recorded with perf record -b or -j any. |
| 255 | |
| 256 | Use brstackinsnlen to print the brstackinsn lenght. For example, you |
| 257 | can’t know the next sequential instruction after an unconditional branch unless |
| 258 | you calculate that based on its length. |
| 259 | |
| 260 | brstackdisasm acts like brstackinsn, but will print disassembled instructions if |
| 261 | perf is built with the capstone library. |
| 262 | |
| 263 | The brstackoff field will print an offset into a specific dso/binary. |
| 264 | |
| 265 | With the metric option perf script can compute metrics for |
| 266 | sampling periods, similar to perf stat. This requires |
| 267 | specifying a group with multiple events defining metrics with the :S option |
| 268 | for perf record. perf will sample on the first event, and |
| 269 | print computed metrics for all the events in the group. Please note |
| 270 | that the metric computed is averaged over the whole sampling |
| 271 | period (since the last sample), not just for the sample point. |
| 272 | |
| 273 | For sample events it's possible to display misc field with -F +misc option, |
| 274 | following letters are displayed for each bit: |
| 275 | |
| 276 | PERF_RECORD_MISC_KERNEL K |
| 277 | PERF_RECORD_MISC_USER U |
| 278 | PERF_RECORD_MISC_HYPERVISOR H |
| 279 | PERF_RECORD_MISC_GUEST_KERNEL G |
| 280 | PERF_RECORD_MISC_GUEST_USER g |
| 281 | PERF_RECORD_MISC_MMAP_DATA* M |
| 282 | PERF_RECORD_MISC_COMM_EXEC E |
| 283 | PERF_RECORD_MISC_SWITCH_OUT S |
| 284 | PERF_RECORD_MISC_SWITCH_OUT_PREEMPT Sp |
| 285 | |
| 286 | $ perf script -F +misc ... |
| 287 | sched-messaging 1414 K 28690.636582: 4590 cycles ... |
| 288 | sched-messaging 1407 U 28690.636600: 325620 cycles ... |
| 289 | sched-messaging 1414 K 28690.636608: 19473 cycles ... |
| 290 | misc field ___________/ |
| 291 | |
| 292 | -k:: |
| 293 | --vmlinux=<file>:: |
| 294 | vmlinux pathname |
| 295 | |
| 296 | --kallsyms=<file>:: |
| 297 | kallsyms pathname |
| 298 | |
| 299 | --symfs=<directory>:: |
| 300 | Look for files with symbols relative to this directory. |
| 301 | |
| 302 | -G:: |
| 303 | --hide-call-graph:: |
| 304 | When printing symbols do not display call chain. |
| 305 | |
| 306 | --stop-bt:: |
| 307 | Stop display of callgraph at these symbols |
| 308 | |
| 309 | -C:: |
| 310 | --cpu:: Only report samples for the list of CPUs provided. Multiple CPUs can |
| 311 | be provided as a comma-separated list with no space: 0,1. Ranges of |
| 312 | CPUs are specified with -: 0-2. Default is to report samples on all |
| 313 | CPUs. |
| 314 | |
| 315 | -c:: |
| 316 | --comms=:: |
| 317 | Only display events for these comms. CSV that understands |
| 318 | file://filename entries. |
| 319 | |
| 320 | --pid=:: |
| 321 | Only show events for given process ID (comma separated list). |
| 322 | |
| 323 | --tid=:: |
| 324 | Only show events for given thread ID (comma separated list). |
| 325 | |
| 326 | -I:: |
| 327 | --show-info:: |
| 328 | Display extended information about the perf.data file. This adds |
| 329 | information which may be very large and thus may clutter the display. |
| 330 | It currently includes: cpu and numa topology of the host system. |
| 331 | It can only be used with the perf script report mode. |
| 332 | |
| 333 | --show-kernel-path:: |
| 334 | Try to resolve the path of [kernel.kallsyms] |
| 335 | |
| 336 | --show-task-events |
| 337 | Display task related events (e.g. FORK, COMM, EXIT). |
| 338 | |
| 339 | --show-mmap-events |
| 340 | Display mmap related events (e.g. MMAP, MMAP2). |
| 341 | |
| 342 | --show-namespace-events |
| 343 | Display namespace events i.e. events of type PERF_RECORD_NAMESPACES. |
| 344 | |
| 345 | --show-switch-events |
| 346 | Display context switch events i.e. events of type PERF_RECORD_SWITCH or |
| 347 | PERF_RECORD_SWITCH_CPU_WIDE. |
| 348 | |
| 349 | --show-lost-events |
| 350 | Display lost events i.e. events of type PERF_RECORD_LOST. |
| 351 | |
| 352 | --show-round-events |
| 353 | Display finished round events i.e. events of type PERF_RECORD_FINISHED_ROUND. |
| 354 | |
| 355 | --show-bpf-events |
| 356 | Display bpf events i.e. events of type PERF_RECORD_KSYMBOL and PERF_RECORD_BPF_EVENT. |
| 357 | |
| 358 | --show-cgroup-events |
| 359 | Display cgroup events i.e. events of type PERF_RECORD_CGROUP. |
| 360 | |
| 361 | --show-text-poke-events |
| 362 | Display text poke events i.e. events of type PERF_RECORD_TEXT_POKE and |
| 363 | PERF_RECORD_KSYMBOL. |
| 364 | |
| 365 | --demangle:: |
| 366 | Demangle symbol names to human readable form. It's enabled by default, |
| 367 | disable with --no-demangle. |
| 368 | |
| 369 | --demangle-kernel:: |
| 370 | Demangle kernel symbol names to human readable form (for C++ kernels). |
| 371 | |
| 372 | --addr2line=<path>:: |
| 373 | Path to addr2line binary. |
| 374 | |
| 375 | --header |
| 376 | Show perf.data header. |
| 377 | |
| 378 | --header-only |
| 379 | Show only perf.data header. |
| 380 | |
| 381 | --itrace:: |
| 382 | Options for decoding instruction tracing data. The options are: |
| 383 | |
| 384 | include::itrace.txt[] |
| 385 | |
| 386 | To disable decoding entirely, use --no-itrace. |
| 387 | |
| 388 | --full-source-path:: |
| 389 | Show the full path for source files for srcline output. |
| 390 | |
| 391 | --max-stack:: |
| 392 | Set the stack depth limit when parsing the callchain, anything |
| 393 | beyond the specified depth will be ignored. This is a trade-off |
| 394 | between information loss and faster processing especially for |
| 395 | workloads that can have a very long callchain stack. |
| 396 | Note that when using the --itrace option the synthesized callchain size |
| 397 | will override this value if the synthesized callchain size is bigger. |
| 398 | |
| 399 | Default: 127 |
| 400 | |
| 401 | --ns:: |
| 402 | Use 9 decimal places when displaying time (i.e. show the nanoseconds) |
| 403 | |
| 404 | -f:: |
| 405 | --force:: |
| 406 | Don't do ownership validation. |
| 407 | |
| 408 | --time:: |
| 409 | Only analyze samples within given time window: <start>,<stop>. Times |
| 410 | have the format seconds.nanoseconds. If start is not given (i.e. time |
| 411 | string is ',x.y') then analysis starts at the beginning of the file. If |
| 412 | stop time is not given (i.e. time string is 'x.y,') then analysis goes |
| 413 | to end of file. Multiple ranges can be separated by spaces, which |
| 414 | requires the argument to be quoted e.g. --time "1234.567,1234.789 1235," |
| 415 | |
| 416 | Also support time percent with multiple time ranges. Time string is |
| 417 | 'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'. |
| 418 | |
| 419 | For example: |
| 420 | Select the second 10% time slice: |
| 421 | perf script --time 10%/2 |
| 422 | |
| 423 | Select from 0% to 10% time slice: |
| 424 | perf script --time 0%-10% |
| 425 | |
| 426 | Select the first and second 10% time slices: |
| 427 | perf script --time 10%/1,10%/2 |
| 428 | |
| 429 | Select from 0% to 10% and 30% to 40% slices: |
| 430 | perf script --time 0%-10%,30%-40% |
| 431 | |
| 432 | --max-blocks:: |
| 433 | Set the maximum number of program blocks to print with brstackinsn for |
| 434 | each sample. |
| 435 | |
| 436 | --reltime:: |
| 437 | Print time stamps relative to trace start. |
| 438 | |
| 439 | --deltatime:: |
| 440 | Print time stamps relative to previous event. |
| 441 | |
| 442 | --per-event-dump:: |
| 443 | Create per event files with a "perf.data.EVENT.dump" name instead of |
| 444 | printing to stdout, useful, for instance, for generating flamegraphs. |
| 445 | |
| 446 | --inline:: |
| 447 | If a callgraph address belongs to an inlined function, the inline stack |
| 448 | will be printed. Each entry has function name and file/line. Enabled by |
| 449 | default, disable with --no-inline. |
| 450 | |
| 451 | --insn-trace[=<raw|disasm>]:: |
| 452 | Show instruction stream in bytes (raw) or disassembled (disasm) |
| 453 | for intel_pt traces. The default is 'raw'. To use xed, combine |
| 454 | 'raw' with --xed to show disassembly done by xed. |
| 455 | |
| 456 | --xed:: |
| 457 | Run xed disassembler on output. Requires installing the xed disassembler. |
| 458 | |
| 459 | -S:: |
| 460 | --symbols=symbol[,symbol...]:: |
| 461 | Only consider the listed symbols. Symbols are typically a name |
| 462 | but they may also be hexadecimal address. |
| 463 | |
| 464 | The hexadecimal address may be the start address of a symbol or |
| 465 | any other address to filter the trace records |
| 466 | |
| 467 | For example, to select the symbol noploop or the address 0x4007a0: |
| 468 | perf script --symbols=noploop,0x4007a0 |
| 469 | |
| 470 | Support filtering trace records by symbol name, start address of |
| 471 | symbol, any hexadecimal address and address range. |
| 472 | |
| 473 | The comparison order is: |
| 474 | |
| 475 | 1. symbol name comparison |
| 476 | 2. symbol start address comparison. |
| 477 | 3. any hexadecimal address comparison. |
| 478 | 4. address range comparison (see --addr-range). |
| 479 | |
| 480 | --addr-range:: |
| 481 | Use with -S or --symbols to list traced records within address range. |
| 482 | |
| 483 | For example, to list the traced records within the address range |
| 484 | [0x4007a0, 0x0x4007a9]: |
| 485 | perf script -S 0x4007a0 --addr-range 10 |
| 486 | |
| 487 | --dsos=:: |
| 488 | Only consider symbols in these DSOs. |
| 489 | |
| 490 | --call-trace:: |
| 491 | Show call stream for intel_pt traces. The CPUs are interleaved, but |
| 492 | can be filtered with -C. |
| 493 | |
| 494 | --call-ret-trace:: |
| 495 | Show call and return stream for intel_pt traces. |
| 496 | |
| 497 | --graph-function:: |
| 498 | For itrace only show specified functions and their callees for |
| 499 | itrace. Multiple functions can be separated by comma. |
| 500 | |
| 501 | --switch-on EVENT_NAME:: |
| 502 | Only consider events after this event is found. |
| 503 | |
| 504 | --switch-off EVENT_NAME:: |
| 505 | Stop considering events after this event is found. |
| 506 | |
| 507 | --show-on-off-events:: |
| 508 | Show the --switch-on/off events too. |
| 509 | |
| 510 | --stitch-lbr:: |
| 511 | Show callgraph with stitched LBRs, which may have more complete |
| 512 | callgraph. The perf.data file must have been obtained using |
| 513 | perf record --call-graph lbr. |
| 514 | Disabled by default. In common cases with call stack overflows, |
| 515 | it can recreate better call stacks than the default lbr call stack |
| 516 | output. But this approach is not foolproof. There can be cases |
| 517 | where it creates incorrect call stacks from incorrect matches. |
| 518 | The known limitations include exception handing such as |
| 519 | setjmp/longjmp will have calls/returns not match. |
| 520 | |
| 521 | :GMEXAMPLECMD: script |
| 522 | :GMEXAMPLESUBCMD: |
| 523 | include::guest-files.txt[] |
| 524 | |
| 525 | SEE ALSO |
| 526 | -------- |
| 527 | linkperf:perf-record[1], linkperf:perf-script-perl[1], |
| 528 | linkperf:perf-script-python[1], linkperf:perf-intel-pt[1], |
| 529 | linkperf:perf-dlfilter[1] |