1perf-config(1) 2============== 3 4NAME 5---- 6perf-config - Get and set variables in a configuration file. 7 8SYNOPSIS 9-------- 10[verse] 11'perf config' [<file-option>] [section.name[=value] ...] 12or 13'perf config' [<file-option>] -l | --list 14 15DESCRIPTION 16----------- 17You can manage variables in a configuration file with this command. 18 19OPTIONS 20------- 21 22-l:: 23--list:: 24 Show current config variables, name and value, for all sections. 25 26--user:: 27 For writing and reading options: write to user 28 '$HOME/.perfconfig' file or read it. 29 30--system:: 31 For writing and reading options: write to system-wide 32 '$(sysconfdir)/perfconfig' or read it. 33 34CONFIGURATION FILE 35------------------ 36 37The perf configuration file contains many variables to change various 38aspects of each of its tools, including output, disk usage, etc. 39The '$HOME/.perfconfig' file is used to store a per-user configuration. 40The file '$(sysconfdir)/perfconfig' can be used to 41store a system-wide default configuration. 42 43One an disable reading config files by setting the PERF_CONFIG environment 44variable to /dev/null, or provide an alternate config file by setting that 45variable. 46 47When reading or writing, the values are read from the system and user 48configuration files by default, and options '--system' and '--user' 49can be used to tell the command to read from or write to only that location. 50 51Syntax 52~~~~~~ 53 54The file consist of sections. A section starts with its name 55surrounded by square brackets and continues till the next section 56begins. Each variable must be in a section, and have the form 57'name = value', for example: 58 59 [section] 60 name1 = value1 61 name2 = value2 62 63Section names are case sensitive and can contain any characters except 64newline (double quote `"` and backslash have to be escaped as `\"` and `\\`, 65respectively). Section headers can't span multiple lines. 66 67Example 68~~~~~~~ 69 70Given a $HOME/.perfconfig like this: 71 72# 73# This is the config file, and 74# a '#' and ';' character indicates a comment 75# 76 77 [colors] 78 # Color variables 79 top = red, default 80 medium = green, default 81 normal = lightgray, default 82 selected = white, lightgray 83 jump_arrows = blue, default 84 addr = magenta, default 85 root = white, blue 86 87 [tui] 88 # Defaults if linked with libslang 89 report = on 90 annotate = on 91 top = on 92 93 [buildid] 94 # Default, disable using /dev/null 95 dir = ~/.debug 96 97 [annotate] 98 # Defaults 99 hide_src_code = false 100 use_offset = true 101 jump_arrows = true 102 show_nr_jumps = false 103 104 [help] 105 # Format can be man, info, web or html 106 format = man 107 autocorrect = 0 108 109 [ui] 110 show-headers = true 111 112 [call-graph] 113 # fp (framepointer), dwarf 114 record-mode = fp 115 print-type = graph 116 order = caller 117 sort-key = function 118 119 [report] 120 # Defaults 121 sort_order = comm,dso,symbol 122 percent-limit = 0 123 queue-size = 0 124 children = true 125 group = true 126 skip-empty = true 127 128 [llvm] 129 dump-obj = true 130 clang-opt = -g 131 132You can hide source code of annotate feature setting the config to false with 133 134 % perf config annotate.hide_src_code=true 135 136If you want to add or modify several config items, you can do like 137 138 % perf config ui.show-headers=false kmem.default=slab 139 140To modify the sort order of report functionality in user config file(i.e. `~/.perfconfig`), do 141 142 % perf config --user report.sort-order=srcline 143 144To change colors of selected line to other foreground and background colors 145in system config file (i.e. `$(sysconf)/perfconfig`), do 146 147 % perf config --system colors.selected=yellow,green 148 149To query the record mode of call graph, do 150 151 % perf config call-graph.record-mode 152 153If you want to know multiple config key/value pairs, you can do like 154 155 % perf config report.queue-size call-graph.order report.children 156 157To query the config value of sort order of call graph in user config file (i.e. `~/.perfconfig`), do 158 159 % perf config --user call-graph.sort-order 160 161To query the config value of buildid directory in system config file (i.e. `$(sysconf)/perfconfig`), do 162 163 % perf config --system buildid.dir 164 165Variables 166~~~~~~~~~ 167 168colors.*:: 169 The variables for customizing the colors used in the output for the 170 'report', 'top' and 'annotate' in the TUI. They should specify the 171 foreground and background colors, separated by a comma, for example: 172 173 medium = green, lightgray 174 175 If you want to use the color configured for you terminal, just leave it 176 as 'default', for example: 177 178 medium = default, lightgray 179 180 Available colors: 181 red, yellow, green, cyan, gray, black, blue, 182 white, default, magenta, lightgray 183 184 colors.top:: 185 'top' means a overhead percentage which is more than 5%. 186 And values of this variable specify percentage colors. 187 Basic key values are foreground-color 'red' and 188 background-color 'default'. 189 colors.medium:: 190 'medium' means a overhead percentage which has more than 0.5%. 191 Default values are 'green' and 'default'. 192 colors.normal:: 193 'normal' means the rest of overhead percentages 194 except 'top', 'medium', 'selected'. 195 Default values are 'lightgray' and 'default'. 196 colors.selected:: 197 This selects the colors for the current entry in a list of entries 198 from sub-commands (top, report, annotate). 199 Default values are 'black' and 'lightgray'. 200 colors.jump_arrows:: 201 Colors for jump arrows on assembly code listings 202 such as 'jns', 'jmp', 'jane', etc. 203 Default values are 'blue', 'default'. 204 colors.addr:: 205 This selects colors for addresses from 'annotate'. 206 Default values are 'magenta', 'default'. 207 colors.root:: 208 Colors for headers in the output of a sub-commands (top, report). 209 Default values are 'white', 'blue'. 210 211core.*:: 212 core.proc-map-timeout:: 213 Sets a timeout (in milliseconds) for parsing /proc/<pid>/maps files. 214 Can be overridden by the --proc-map-timeout option on supported 215 subcommands. The default timeout is 500ms. 216 217tui.*, gtk.*:: 218 Subcommands that can be configured here are 'top', 'report' and 'annotate'. 219 These values are booleans, for example: 220 221 [tui] 222 top = true 223 224 will make the TUI be the default for the 'top' subcommand. Those will be 225 available if the required libs were detected at tool build time. 226 227buildid.*:: 228 buildid.dir:: 229 Each executable and shared library in modern distributions comes with a 230 content based identifier that, if available, will be inserted in a 231 'perf.data' file header to, at analysis time find what is needed to do 232 symbol resolution, code annotation, etc. 233 234 The recording tools also stores a hard link or copy in a per-user 235 directory, $HOME/.debug/, of binaries, shared libraries, /proc/kallsyms 236 and /proc/kcore files to be used at analysis time. 237 238 The buildid.dir variable can be used to either change this directory 239 cache location, or to disable it altogether. If you want to disable it, 240 set buildid.dir to /dev/null. The default is $HOME/.debug 241 242buildid-cache.*:: 243 buildid-cache.debuginfod=URLs 244 Specify debuginfod URLs to be used when retrieving perf.data binaries, 245 it follows the same syntax as the DEBUGINFOD_URLS variable, like: 246 247 buildid-cache.debuginfod=http://192.168.122.174:8002 248 249annotate.*:: 250 These are in control of addresses, jump function, source code 251 in lines of assembly code from a specific program. 252 253 annotate.disassembler_style: 254 Use this to change the default disassembler style to some other value 255 supported by binutils, such as "intel", see the '-M' option help in the 256 'objdump' man page. 257 258 annotate.hide_src_code:: 259 If a program which is analyzed has source code, 260 this option lets 'annotate' print a list of assembly code with the source code. 261 For example, let's see a part of a program. There're four lines. 262 If this option is 'true', they can be printed 263 without source code from a program as below. 264 265 │ push %rbp 266 │ mov %rsp,%rbp 267 │ sub $0x10,%rsp 268 │ mov (%rdi),%rdx 269 270 But if this option is 'false', source code of the part 271 can be also printed as below. Default is 'false'. 272 273 │ struct rb_node *rb_next(const struct rb_node *node) 274 │ { 275 │ push %rbp 276 │ mov %rsp,%rbp 277 │ sub $0x10,%rsp 278 │ struct rb_node *parent; 279 │ 280 │ if (RB_EMPTY_NODE(node)) 281 │ mov (%rdi),%rdx 282 │ return n; 283 284 This option works with tui, stdio2 browsers. 285 286 annotate.use_offset:: 287 Basing on a first address of a loaded function, offset can be used. 288 Instead of using original addresses of assembly code, 289 addresses subtracted from a base address can be printed. 290 Let's illustrate an example. 291 If a base address is 0XFFFFFFFF81624d50 as below, 292 293 ffffffff81624d50 <load0> 294 295 an address on assembly code has a specific absolute address as below 296 297 ffffffff816250b8:│ mov 0x8(%r14),%rdi 298 299 but if use_offset is 'true', an address subtracted from a base address is printed. 300 Default is true. This option is only applied to TUI. 301 302 368:│ mov 0x8(%r14),%rdi 303 304 This option works with tui, stdio2 browsers. 305 306 annotate.jump_arrows:: 307 There can be jump instruction among assembly code. 308 Depending on a boolean value of jump_arrows, 309 arrows can be printed or not which represent 310 where do the instruction jump into as below. 311 312 │ ┌──jmp 1333 313 │ │ xchg %ax,%ax 314 │1330:│ mov %r15,%r10 315 │1333:└─→cmp %r15,%r14 316 317 If jump_arrow is 'false', the arrows isn't printed as below. 318 Default is 'false'. 319 320 │ ↓ jmp 1333 321 │ xchg %ax,%ax 322 │1330: mov %r15,%r10 323 │1333: cmp %r15,%r14 324 325 This option works with tui browser. 326 327 annotate.show_linenr:: 328 When showing source code if this option is 'true', 329 line numbers are printed as below. 330 331 │1628 if (type & PERF_SAMPLE_IDENTIFIER) { 332 │ ↓ jne 508 333 │1628 data->id = *array; 334 │1629 array++; 335 │1630 } 336 337 However if this option is 'false', they aren't printed as below. 338 Default is 'false'. 339 340 │ if (type & PERF_SAMPLE_IDENTIFIER) { 341 │ ↓ jne 508 342 │ data->id = *array; 343 │ array++; 344 │ } 345 346 This option works with tui, stdio2 browsers. 347 348 annotate.show_nr_jumps:: 349 Let's see a part of assembly code. 350 351 │1382: movb $0x1,-0x270(%rbp) 352 353 If use this, the number of branches jumping to that address can be printed as below. 354 Default is 'false'. 355 356 │1 1382: movb $0x1,-0x270(%rbp) 357 358 This option works with tui, stdio2 browsers. 359 360 annotate.show_total_period:: 361 To compare two records on an instruction base, with this option 362 provided, display total number of samples that belong to a line 363 in assembly code. If this option is 'true', total periods are printed 364 instead of percent values as below. 365 366 302 │ mov %eax,%eax 367 368 But if this option is 'false', percent values for overhead are printed i.e. 369 Default is 'false'. 370 371 99.93 │ mov %eax,%eax 372 373 This option works with tui, stdio2, stdio browsers. 374 375 annotate.show_nr_samples:: 376 By default perf annotate shows percentage of samples. This option 377 can be used to print absolute number of samples. Ex, when set as 378 false: 379 380 Percent│ 381 74.03 │ mov %fs:0x28,%rax 382 383 When set as true: 384 385 Samples│ 386 6 │ mov %fs:0x28,%rax 387 388 This option works with tui, stdio2, stdio browsers. 389 390 annotate.offset_level:: 391 Default is '1', meaning just jump targets will have offsets show right beside 392 the instruction. When set to '2' 'call' instructions will also have its offsets 393 shown, 3 or higher will show offsets for all instructions. 394 395 This option works with tui, stdio2 browsers. 396 397 annotate.demangle:: 398 Demangle symbol names to human readable form. Default is 'true'. 399 400 annotate.demangle_kernel:: 401 Demangle kernel symbol names to human readable form. Default is 'true'. 402 403hist.*:: 404 hist.percentage:: 405 This option control the way to calculate overhead of filtered entries - 406 that means the value of this option is effective only if there's a 407 filter (by comm, dso or symbol name). Suppose a following example: 408 409 Overhead Symbols 410 ........ ....... 411 33.33% foo 412 33.33% bar 413 33.33% baz 414 415 This is an original overhead and we'll filter out the first 'foo' 416 entry. The value of 'relative' would increase the overhead of 'bar' 417 and 'baz' to 50.00% for each, while 'absolute' would show their 418 current overhead (33.33%). 419 420ui.*:: 421 ui.show-headers:: 422 This option controls display of column headers (like 'Overhead' and 'Symbol') 423 in 'report' and 'top'. If this option is false, they are hidden. 424 This option is only applied to TUI. 425 426call-graph.*:: 427 The following controls the handling of call-graphs (obtained via the 428 -g/--call-graph options). 429 430 call-graph.record-mode:: 431 The mode for user space can be 'fp' (frame pointer), 'dwarf' 432 and 'lbr'. The value 'dwarf' is effective only if libunwind 433 (or a recent version of libdw) is present on the system; 434 the value 'lbr' only works for certain cpus. The method for 435 kernel space is controlled not by this option but by the 436 kernel config (CONFIG_UNWINDER_*). 437 438 call-graph.dump-size:: 439 The size of stack to dump in order to do post-unwinding. Default is 8192 (byte). 440 When using dwarf into record-mode, the default size will be used if omitted. 441 442 call-graph.print-type:: 443 The print-types can be graph (graph absolute), fractal (graph relative), 444 flat and folded. This option controls a way to show overhead for each callchain 445 entry. Suppose a following example. 446 447 Overhead Symbols 448 ........ ....... 449 40.00% foo 450 | 451 ---foo 452 | 453 |--50.00%--bar 454 | main 455 | 456 --50.00%--baz 457 main 458 459 This output is a 'fractal' format. The 'foo' came from 'bar' and 'baz' exactly 460 half and half so 'fractal' shows 50.00% for each 461 (meaning that it assumes 100% total overhead of 'foo'). 462 463 The 'graph' uses absolute overhead value of 'foo' as total so each of 464 'bar' and 'baz' callchain will have 20.00% of overhead. 465 If 'flat' is used, single column and linear exposure of call chains. 466 'folded' mean call chains are displayed in a line, separated by semicolons. 467 468 call-graph.order:: 469 This option controls print order of callchains. The default is 470 'callee' which means callee is printed at top and then followed by its 471 caller and so on. The 'caller' prints it in reverse order. 472 473 If this option is not set and report.children or top.children is 474 set to true (or the equivalent command line option is given), 475 the default value of this option is changed to 'caller' for the 476 execution of 'perf report' or 'perf top'. Other commands will 477 still default to 'callee'. 478 479 call-graph.sort-key:: 480 The callchains are merged if they contain same information. 481 The sort-key option determines a way to compare the callchains. 482 A value of 'sort-key' can be 'function' or 'address'. 483 The default is 'function'. 484 485 call-graph.threshold:: 486 When there're many callchains it'd print tons of lines. So perf omits 487 small callchains under a certain overhead (threshold) and this option 488 control the threshold. Default is 0.5 (%). The overhead is calculated 489 by value depends on call-graph.print-type. 490 491 call-graph.print-limit:: 492 This is a maximum number of lines of callchain printed for a single 493 histogram entry. Default is 0 which means no limitation. 494 495report.*:: 496 report.sort_order:: 497 Allows changing the default sort order from "comm,dso,symbol" to 498 some other default, for instance "sym,dso" may be more fitting for 499 kernel developers. 500 report.percent-limit:: 501 This one is mostly the same as call-graph.threshold but works for 502 histogram entries. Entries having an overhead lower than this 503 percentage will not be printed. Default is '0'. If percent-limit 504 is '10', only entries which have more than 10% of overhead will be 505 printed. 506 507 report.queue-size:: 508 This option sets up the maximum allocation size of the internal 509 event queue for ordering events. Default is 0, meaning no limit. 510 511 report.children:: 512 'Children' means functions called from another function. 513 If this option is true, 'perf report' cumulates callchains of children 514 and show (accumulated) total overhead as well as 'Self' overhead. 515 Please refer to the 'perf report' manual. The default is 'true'. 516 517 report.group:: 518 This option is to show event group information together. 519 Example output with this turned on, notice that there is one column 520 per event in the group, ref-cycles and cycles: 521 522 # group: {ref-cycles,cycles} 523 # ======== 524 # 525 # Samples: 7K of event 'anon group { ref-cycles, cycles }' 526 # Event count (approx.): 6876107743 527 # 528 # Overhead Command Shared Object Symbol 529 # ................ ....... ................. ................... 530 # 531 99.84% 99.76% noploop noploop [.] main 532 0.07% 0.00% noploop ld-2.15.so [.] strcmp 533 0.03% 0.00% noploop [kernel.kallsyms] [k] timerqueue_del 534 535 report.skip-empty:: 536 This option can change default stat behavior with empty results. 537 If it's set true, 'perf report --stat' will not show 0 stats. 538 539top.*:: 540 top.children:: 541 Same as 'report.children'. So if it is enabled, the output of 'top' 542 command will have 'Children' overhead column as well as 'Self' overhead 543 column by default. 544 The default is 'true'. 545 546 top.call-graph:: 547 This is identical to 'call-graph.record-mode', except it is 548 applicable only for 'top' subcommand. This option ONLY setup 549 the unwind method. To enable 'perf top' to actually use it, 550 the command line option -g must be specified. 551 552man.*:: 553 man.viewer:: 554 This option can assign a tool to view manual pages when 'help' 555 subcommand was invoked. Supported tools are 'man', 'woman' 556 (with emacs client) and 'konqueror'. Default is 'man'. 557 558 New man viewer tool can be also added using 'man.<tool>.cmd' 559 or use different path using 'man.<tool>.path' config option. 560 561pager.*:: 562 pager.<subcommand>:: 563 When the subcommand is run on stdio, determine whether it uses 564 pager or not based on this value. Default is 'unspecified'. 565 566kmem.*:: 567 kmem.default:: 568 This option decides which allocator is to be analyzed if neither 569 '--slab' nor '--page' option is used. Default is 'slab'. 570 571record.*:: 572 record.build-id:: 573 This option can be 'cache', 'no-cache', 'skip' or 'mmap'. 574 'cache' is to post-process data and save/update the binaries into 575 the build-id cache (in ~/.debug). This is the default. 576 But if this option is 'no-cache', it will not update the build-id cache. 577 'skip' skips post-processing and does not update the cache. 578 'mmap' skips post-processing and reads build-ids from MMAP events. 579 580 record.call-graph:: 581 This is identical to 'call-graph.record-mode', except it is 582 applicable only for 'record' subcommand. This option ONLY setup 583 the unwind method. To enable 'perf record' to actually use it, 584 the command line option -g must be specified. 585 586 record.aio:: 587 Use 'n' control blocks in asynchronous (Posix AIO) trace writing 588 mode ('n' default: 1, max: 4). 589 590diff.*:: 591 diff.order:: 592 This option sets the number of columns to sort the result. 593 The default is 0, which means sorting by baseline. 594 Setting it to 1 will sort the result by delta (or other 595 compute method selected). 596 597 diff.compute:: 598 This options sets the method for computing the diff result. 599 Possible values are 'delta', 'delta-abs', 'ratio' and 600 'wdiff'. Default is 'delta'. 601 602trace.*:: 603 trace.add_events:: 604 Allows adding a set of events to add to the ones specified 605 by the user, or use as a default one if none was specified. 606 The initial use case is to add augmented_raw_syscalls.o to 607 activate the 'perf trace' logic that looks for syscall 608 pointer contents after the normal tracepoint payload. 609 610 trace.args_alignment:: 611 Number of columns to align the argument list, default is 70, 612 use 40 for the strace default, zero to no alignment. 613 614 trace.no_inherit:: 615 Do not follow children threads. 616 617 trace.show_arg_names:: 618 Should syscall argument names be printed? If not then trace.show_zeros 619 will be set. 620 621 trace.show_duration:: 622 Show syscall duration. 623 624 trace.show_prefix:: 625 If set to 'yes' will show common string prefixes in tables. The default 626 is to remove the common prefix in things like "MAP_SHARED", showing just "SHARED". 627 628 trace.show_timestamp:: 629 Show syscall start timestamp. 630 631 trace.show_zeros:: 632 Do not suppress syscall arguments that are equal to zero. 633 634 trace.tracepoint_beautifiers:: 635 Use "libtraceevent" to use that library to augment the tracepoint arguments, 636 "libbeauty", the default, to use the same argument beautifiers used in the 637 strace-like sys_enter+sys_exit lines. 638 639ftrace.*:: 640 ftrace.tracer:: 641 Can be used to select the default tracer when neither -G nor 642 -F option is not specified. Possible values are 'function' and 643 'function_graph'. 644 645llvm.*:: 646 llvm.clang-path:: 647 Path to clang. If omit, search it from $PATH. 648 649 llvm.clang-bpf-cmd-template:: 650 Cmdline template. Below lines show its default value. Environment 651 variable is used to pass options. 652 "$CLANG_EXEC -D__KERNEL__ -D__NR_CPUS__=$NR_CPUS "\ 653 "-DLINUX_VERSION_CODE=$LINUX_VERSION_CODE " \ 654 "$CLANG_OPTIONS $PERF_BPF_INC_OPTIONS $KERNEL_INC_OPTIONS " \ 655 "-Wno-unused-value -Wno-pointer-sign " \ 656 "-working-directory $WORKING_DIR " \ 657 "-c \"$CLANG_SOURCE\" -target bpf $CLANG_EMIT_LLVM -O2 -o - $LLVM_OPTIONS_PIPE" 658 659 llvm.clang-opt:: 660 Options passed to clang. 661 662 llvm.kbuild-dir:: 663 kbuild directory. If not set, use /lib/modules/`uname -r`/build. 664 If set to "" deliberately, skip kernel header auto-detector. 665 666 llvm.kbuild-opts:: 667 Options passed to 'make' when detecting kernel header options. 668 669 llvm.dump-obj:: 670 Enable perf dump BPF object files compiled by LLVM. 671 672 llvm.opts:: 673 Options passed to llc. 674 675samples.*:: 676 677 samples.context:: 678 Define how many ns worth of time to show 679 around samples in perf report sample context browser. 680 681scripts.*:: 682 683 Any option defines a script that is added to the scripts menu 684 in the interactive perf browser and whose output is displayed. 685 The name of the option is the name, the value is a script command line. 686 The script gets the same options passed as a full perf script, 687 in particular -i perfdata file, --cpu, --tid 688 689convert.*:: 690 691 convert.queue-size:: 692 Limit the size of ordered_events queue, so we could control 693 allocation size of perf data files without proper finished 694 round events. 695stat.*:: 696 697 stat.big-num:: 698 (boolean) Change the default for "--big-num". To make 699 "--no-big-num" the default, set "stat.big-num=false". 700 701intel-pt.*:: 702 703 intel-pt.cache-divisor:: 704 705 intel-pt.mispred-all:: 706 If set, Intel PT decoder will set the mispred flag on all 707 branches. 708 709 intel-pt.max-loops:: 710 If set and non-zero, the maximum number of unconditional 711 branches decoded without consuming any trace packets. If 712 the maximum is exceeded there will be a "Never-ending loop" 713 error. The default is 100000. 714 715auxtrace.*:: 716 717 auxtrace.dumpdir:: 718 s390 only. The directory to save the auxiliary trace buffer 719 can be changed using this option. Ex, auxtrace.dumpdir=/tmp. 720 If the directory does not exist or has the wrong file type, 721 the current directory is used. 722 723daemon.*:: 724 725 daemon.base:: 726 Base path for daemon data. All sessions data are stored under 727 this path. 728 729session-<NAME>.*:: 730 731 session-<NAME>.run:: 732 733 Defines new record session for daemon. The value is record's 734 command line without the 'record' keyword. 735 736 737SEE ALSO 738-------- 739linkperf:perf[1] 740