6 perf-config - Get and set variables in a configuration file.
11 'perf config' [<file-option>] [section.name[=value] ...]
13 'perf config' [<file-option>] -l | --list
17 You can manage variables in a configuration file with this command.
24 Show current config variables, name and value, for all sections.
27 For writing and reading options: write to user
28 '$HOME/.perfconfig' file or read it.
31 For writing and reading options: write to system-wide
32 '$(sysconfdir)/perfconfig' or read it.
37 The perf configuration file contains many variables to change various
38 aspects of each of its tools, including output, disk usage, etc.
39 The '$HOME/.perfconfig' file is used to store a per-user configuration.
40 The file '$(sysconfdir)/perfconfig' can be used to
41 store a system-wide default configuration.
43 One an disable reading config files by setting the PERF_CONFIG environment
44 variable to /dev/null, or provide an alternate config file by setting that
47 When reading or writing, the values are read from the system and user
48 configuration files by default, and options '--system' and '--user'
49 can be used to tell the command to read from or write to only that location.
54 The file consist of sections. A section starts with its name
55 surrounded by square brackets and continues till the next section
56 begins. Each variable must be in a section, and have the form
57 'name = value', for example:
63 Section names are case sensitive and can contain any characters except
64 newline (double quote `"` and backslash have to be escaped as `\"` and `\\`,
65 respectively). Section headers can't span multiple lines.
70 Given a $HOME/.perfconfig like this:
73 # This is the config file, and
74 # a '#' and ';' character indicates a comment
80 medium = green, default
81 normal = lightgray, default
82 selected = white, lightgray
83 jump_arrows = blue, default
84 addr = magenta, default
88 # Defaults if linked with libslang
94 # Default, disable using /dev/null
102 show_nr_jumps = false
105 # Format can be man, info, web or html
113 # fp (framepointer), dwarf
121 sort_order = comm,dso,symbol
131 You can hide source code of annotate feature setting the config to false with
133 % perf config annotate.hide_src_code=true
135 If you want to add or modify several config items, you can do like
137 % perf config ui.show-headers=false kmem.default=slab
139 To modify the sort order of report functionality in user config file(i.e. `~/.perfconfig`), do
141 % perf config --user report sort-order=srcline
143 To change colors of selected line to other foreground and background colors
144 in system config file (i.e. `$(sysconf)/perfconfig`), do
146 % perf config --system colors.selected=yellow,green
148 To query the record mode of call graph, do
150 % perf config call-graph.record-mode
152 If you want to know multiple config key/value pairs, you can do like
154 % perf config report.queue-size call-graph.order report.children
156 To query the config value of sort order of call graph in user config file (i.e. `~/.perfconfig`), do
158 % perf config --user call-graph.sort-order
160 To query the config value of buildid directory in system config file (i.e. `$(sysconf)/perfconfig`), do
162 % perf config --system buildid.dir
168 The variables for customizing the colors used in the output for the
169 'report', 'top' and 'annotate' in the TUI. They should specify the
170 foreground and background colors, separated by a comma, for example:
172 medium = green, lightgray
174 If you want to use the color configured for you terminal, just leave it
175 as 'default', for example:
177 medium = default, lightgray
180 red, yellow, green, cyan, gray, black, blue,
181 white, default, magenta, lightgray
184 'top' means a overhead percentage which is more than 5%.
185 And values of this variable specify percentage colors.
186 Basic key values are foreground-color 'red' and
187 background-color 'default'.
189 'medium' means a overhead percentage which has more than 0.5%.
190 Default values are 'green' and 'default'.
192 'normal' means the rest of overhead percentages
193 except 'top', 'medium', 'selected'.
194 Default values are 'lightgray' and 'default'.
196 This selects the colors for the current entry in a list of entries
197 from sub-commands (top, report, annotate).
198 Default values are 'black' and 'lightgray'.
200 Colors for jump arrows on assembly code listings
201 such as 'jns', 'jmp', 'jane', etc.
202 Default values are 'blue', 'default'.
204 This selects colors for addresses from 'annotate'.
205 Default values are 'magenta', 'default'.
207 Colors for headers in the output of a sub-commands (top, report).
208 Default values are 'white', 'blue'.
211 core.proc-map-timeout::
212 Sets a timeout (in milliseconds) for parsing /proc/<pid>/maps files.
213 Can be overridden by the --proc-map-timeout option on supported
214 subcommands. The default timeout is 500ms.
217 Subcommands that can be configured here are 'top', 'report' and 'annotate'.
218 These values are booleans, for example:
223 will make the TUI be the default for the 'top' subcommand. Those will be
224 available if the required libs were detected at tool build time.
228 Each executable and shared library in modern distributions comes with a
229 content based identifier that, if available, will be inserted in a
230 'perf.data' file header to, at analysis time find what is needed to do
231 symbol resolution, code annotation, etc.
233 The recording tools also stores a hard link or copy in a per-user
234 directory, $HOME/.debug/, of binaries, shared libraries, /proc/kallsyms
235 and /proc/kcore files to be used at analysis time.
237 The buildid.dir variable can be used to either change this directory
238 cache location, or to disable it altogether. If you want to disable it,
239 set buildid.dir to /dev/null. The default is $HOME/.debug
242 These are in control of addresses, jump function, source code
243 in lines of assembly code from a specific program.
245 annotate.hide_src_code::
246 If a program which is analyzed has source code,
247 this option lets 'annotate' print a list of assembly code with the source code.
248 For example, let's see a part of a program. There're four lines.
249 If this option is 'true', they can be printed
250 without source code from a program as below.
257 But if this option is 'false', source code of the part
258 can be also printed as below. Default is 'false'.
260 │ struct rb_node *rb_next(const struct rb_node *node)
265 │ struct rb_node *parent;
267 │ if (RB_EMPTY_NODE(node))
271 This option works with tui, stdio2 browsers.
273 annotate.use_offset::
274 Basing on a first address of a loaded function, offset can be used.
275 Instead of using original addresses of assembly code,
276 addresses subtracted from a base address can be printed.
277 Let's illustrate an example.
278 If a base address is 0XFFFFFFFF81624d50 as below,
280 ffffffff81624d50 <load0>
282 an address on assembly code has a specific absolute address as below
284 ffffffff816250b8:│ mov 0x8(%r14),%rdi
286 but if use_offset is 'true', an address subtracted from a base address is printed.
287 Default is true. This option is only applied to TUI.
289 368:│ mov 0x8(%r14),%rdi
291 This option works with tui, stdio2 browsers.
293 annotate.jump_arrows::
294 There can be jump instruction among assembly code.
295 Depending on a boolean value of jump_arrows,
296 arrows can be printed or not which represent
297 where do the instruction jump into as below.
301 │1330:│ mov %r15,%r10
302 │1333:└─→cmp %r15,%r14
304 If jump_arrow is 'false', the arrows isn't printed as below.
312 This option works with tui browser.
314 annotate.show_linenr::
315 When showing source code if this option is 'true',
316 line numbers are printed as below.
318 │1628 if (type & PERF_SAMPLE_IDENTIFIER) {
320 │1628 data->id = *array;
324 However if this option is 'false', they aren't printed as below.
327 │ if (type & PERF_SAMPLE_IDENTIFIER) {
333 This option works with tui, stdio2 browsers.
335 annotate.show_nr_jumps::
336 Let's see a part of assembly code.
338 │1382: movb $0x1,-0x270(%rbp)
340 If use this, the number of branches jumping to that address can be printed as below.
343 │1 1382: movb $0x1,-0x270(%rbp)
345 This option works with tui, stdio2 browsers.
347 annotate.show_total_period::
348 To compare two records on an instruction base, with this option
349 provided, display total number of samples that belong to a line
350 in assembly code. If this option is 'true', total periods are printed
351 instead of percent values as below.
355 But if this option is 'false', percent values for overhead are printed i.e.
358 99.93 │ mov %eax,%eax
360 This option works with tui, stdio2, stdio browsers.
362 annotate.show_nr_samples::
363 By default perf annotate shows percentage of samples. This option
364 can be used to print absolute number of samples. Ex, when set as
368 74.03 │ mov %fs:0x28,%rax
373 6 │ mov %fs:0x28,%rax
375 This option works with tui, stdio2, stdio browsers.
377 annotate.offset_level::
378 Default is '1', meaning just jump targets will have offsets show right beside
379 the instruction. When set to '2' 'call' instructions will also have its offsets
380 shown, 3 or higher will show offsets for all instructions.
382 This option works with tui, stdio2 browsers.
386 This option control the way to calculate overhead of filtered entries -
387 that means the value of this option is effective only if there's a
388 filter (by comm, dso or symbol name). Suppose a following example:
396 This is an original overhead and we'll filter out the first 'foo'
397 entry. The value of 'relative' would increase the overhead of 'bar'
398 and 'baz' to 50.00% for each, while 'absolute' would show their
399 current overhead (33.33%).
403 This option controls display of column headers (like 'Overhead' and 'Symbol')
404 in 'report' and 'top'. If this option is false, they are hidden.
405 This option is only applied to TUI.
408 When sub-commands 'top' and 'report' work with -g/—-children
409 there're options in control of call-graph.
411 call-graph.record-mode::
412 The record-mode can be 'fp' (frame pointer), 'dwarf' and 'lbr'.
413 The value of 'dwarf' is effective only if perf detect needed library
414 (libunwind or a recent version of libdw).
415 'lbr' only work for cpus that support it.
417 call-graph.dump-size::
418 The size of stack to dump in order to do post-unwinding. Default is 8192 (byte).
419 When using dwarf into record-mode, the default size will be used if omitted.
421 call-graph.print-type::
422 The print-types can be graph (graph absolute), fractal (graph relative),
423 flat and folded. This option controls a way to show overhead for each callchain
424 entry. Suppose a following example.
438 This output is a 'fractal' format. The 'foo' came from 'bar' and 'baz' exactly
439 half and half so 'fractal' shows 50.00% for each
440 (meaning that it assumes 100% total overhead of 'foo').
442 The 'graph' uses absolute overhead value of 'foo' as total so each of
443 'bar' and 'baz' callchain will have 20.00% of overhead.
444 If 'flat' is used, single column and linear exposure of call chains.
445 'folded' mean call chains are displayed in a line, separated by semicolons.
448 This option controls print order of callchains. The default is
449 'callee' which means callee is printed at top and then followed by its
450 caller and so on. The 'caller' prints it in reverse order.
452 If this option is not set and report.children or top.children is
453 set to true (or the equivalent command line option is given),
454 the default value of this option is changed to 'caller' for the
455 execution of 'perf report' or 'perf top'. Other commands will
456 still default to 'callee'.
458 call-graph.sort-key::
459 The callchains are merged if they contain same information.
460 The sort-key option determines a way to compare the callchains.
461 A value of 'sort-key' can be 'function' or 'address'.
462 The default is 'function'.
464 call-graph.threshold::
465 When there're many callchains it'd print tons of lines. So perf omits
466 small callchains under a certain overhead (threshold) and this option
467 control the threshold. Default is 0.5 (%). The overhead is calculated
468 by value depends on call-graph.print-type.
470 call-graph.print-limit::
471 This is a maximum number of lines of callchain printed for a single
472 histogram entry. Default is 0 which means no limitation.
476 Allows changing the default sort order from "comm,dso,symbol" to
477 some other default, for instance "sym,dso" may be more fitting for
479 report.percent-limit::
480 This one is mostly the same as call-graph.threshold but works for
481 histogram entries. Entries having an overhead lower than this
482 percentage will not be printed. Default is '0'. If percent-limit
483 is '10', only entries which have more than 10% of overhead will be
487 This option sets up the maximum allocation size of the internal
488 event queue for ordering events. Default is 0, meaning no limit.
491 'Children' means functions called from another function.
492 If this option is true, 'perf report' cumulates callchains of children
493 and show (accumulated) total overhead as well as 'Self' overhead.
494 Please refer to the 'perf report' manual. The default is 'true'.
497 This option is to show event group information together.
498 Example output with this turned on, notice that there is one column
499 per event in the group, ref-cycles and cycles:
501 # group: {ref-cycles,cycles}
504 # Samples: 7K of event 'anon group { ref-cycles, cycles }'
505 # Event count (approx.): 6876107743
507 # Overhead Command Shared Object Symbol
508 # ................ ....... ................. ...................
510 99.84% 99.76% noploop noploop [.] main
511 0.07% 0.00% noploop ld-2.15.so [.] strcmp
512 0.03% 0.00% noploop [kernel.kallsyms] [k] timerqueue_del
516 Same as 'report.children'. So if it is enabled, the output of 'top'
517 command will have 'Children' overhead column as well as 'Self' overhead
519 The default is 'true'.
522 This is identical to 'call-graph.record-mode', except it is
523 applicable only for 'top' subcommand. This option ONLY setup
524 the unwind method. To enable 'perf top' to actually use it,
525 the command line option -g must be specified.
529 This option can assign a tool to view manual pages when 'help'
530 subcommand was invoked. Supported tools are 'man', 'woman'
531 (with emacs client) and 'konqueror'. Default is 'man'.
533 New man viewer tool can be also added using 'man.<tool>.cmd'
534 or use different path using 'man.<tool>.path' config option.
538 When the subcommand is run on stdio, determine whether it uses
539 pager or not based on this value. Default is 'unspecified'.
543 This option decides which allocator is to be analyzed if neither
544 '--slab' nor '--page' option is used. Default is 'slab'.
548 This option can be 'cache', 'no-cache' or 'skip'.
549 'cache' is to post-process data and save/update the binaries into
550 the build-id cache (in ~/.debug). This is the default.
551 But if this option is 'no-cache', it will not update the build-id cache.
552 'skip' skips post-processing and does not update the cache.
555 This is identical to 'call-graph.record-mode', except it is
556 applicable only for 'record' subcommand. This option ONLY setup
557 the unwind method. To enable 'perf record' to actually use it,
558 the command line option -g must be specified.
561 Use 'n' control blocks in asynchronous (Posix AIO) trace writing
562 mode ('n' default: 1, max: 4).
566 This option sets the number of columns to sort the result.
567 The default is 0, which means sorting by baseline.
568 Setting it to 1 will sort the result by delta (or other
569 compute method selected).
572 This options sets the method for computing the diff result.
573 Possible values are 'delta', 'delta-abs', 'ratio' and
574 'wdiff'. Default is 'delta'.
578 Allows adding a set of events to add to the ones specified
579 by the user, or use as a default one if none was specified.
580 The initial use case is to add augmented_raw_syscalls.o to
581 activate the 'perf trace' logic that looks for syscall
582 pointer contents after the normal tracepoint payload.
584 trace.args_alignment::
585 Number of columns to align the argument list, default is 70,
586 use 40 for the strace default, zero to no alignment.
589 Do not follow children threads.
591 trace.show_arg_names::
592 Should syscall argument names be printed? If not then trace.show_zeros
595 trace.show_duration::
596 Show syscall duration.
599 If set to 'yes' will show common string prefixes in tables. The default
600 is to remove the common prefix in things like "MAP_SHARED", showing just "SHARED".
602 trace.show_timestamp::
603 Show syscall start timestamp.
606 Do not suppress syscall arguments that are equal to zero.
608 trace.tracepoint_beautifiers::
609 Use "libtraceevent" to use that library to augment the tracepoint arguments,
610 "libbeauty", the default, to use the same argument beautifiers used in the
611 strace-like sys_enter+sys_exit lines.
615 Can be used to select the default tracer. Possible values are
616 'function' and 'function_graph'.
620 Path to clang. If omit, search it from $PATH.
622 llvm.clang-bpf-cmd-template::
623 Cmdline template. Below lines show its default value. Environment
624 variable is used to pass options.
625 "$CLANG_EXEC -D__KERNEL__ -D__NR_CPUS__=$NR_CPUS "\
626 "-DLINUX_VERSION_CODE=$LINUX_VERSION_CODE " \
627 "$CLANG_OPTIONS $PERF_BPF_INC_OPTIONS $KERNEL_INC_OPTIONS " \
628 "-Wno-unused-value -Wno-pointer-sign " \
629 "-working-directory $WORKING_DIR " \
630 "-c \"$CLANG_SOURCE\" -target bpf $CLANG_EMIT_LLVM -O2 -o - $LLVM_OPTIONS_PIPE"
633 Options passed to clang.
636 kbuild directory. If not set, use /lib/modules/`uname -r`/build.
637 If set to "" deliberately, skip kernel header auto-detector.
640 Options passed to 'make' when detecting kernel header options.
643 Enable perf dump BPF object files compiled by LLVM.
646 Options passed to llc.
651 Define how many ns worth of time to show
652 around samples in perf report sample context browser.
656 Any option defines a script that is added to the scripts menu
657 in the interactive perf browser and whose output is displayed.
658 The name of the option is the name, the value is a script command line.
659 The script gets the same options passed as a full perf script,
660 in particular -i perfdata file, --cpu, --tid
665 Limit the size of ordered_events queue, so we could control
666 allocation size of perf data files without proper finished
671 intel-pt.cache-divisor::
673 intel-pt.mispred-all::
674 If set, Intel PT decoder will set the mispred flag on all
680 s390 only. The directory to save the auxiliary trace buffer
681 can be changed using this option. Ex, auxtrace.dumpdir=/tmp.
682 If the directory does not exist or has the wrong file type,
683 the current directory is used.