xref: /linux/tools/perf/Documentation/perf-config.txt (revision a1c3be890440a1769ed6f822376a3e3ab0d42994)
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 $HOME/.perfconfig like this:
71
72#
73# This is the config file, and
74# a '#' and ';' character indicates a comment
75#
76
77	[colors]
78		# Color variables
79		top = red, default
80		medium = green, default
81		normal = lightgray, default
82		selected = white, lightgray
83		jump_arrows = blue, default
84		addr = magenta, default
85		root = white, blue
86
87	[tui]
88		# Defaults if linked with libslang
89		report = on
90		annotate = on
91		top = on
92
93	[buildid]
94		# Default, disable using /dev/null
95		dir = ~/.debug
96
97	[annotate]
98		# Defaults
99		hide_src_code = false
100		use_offset = true
101		jump_arrows = true
102		show_nr_jumps = false
103
104	[help]
105		# Format can be man, info, web or html
106		format = man
107		autocorrect = 0
108
109	[ui]
110		show-headers = true
111
112	[call-graph]
113		# fp (framepointer), dwarf
114		record-mode = fp
115		print-type = graph
116		order = caller
117		sort-key = function
118
119	[report]
120		# Defaults
121		sort_order = comm,dso,symbol
122		percent-limit = 0
123		queue-size = 0
124		children = true
125		group = true
126
127	[llvm]
128		dump-obj = true
129		clang-opt = -g
130
131You can hide source code of annotate feature setting the config to false with
132
133	% perf config annotate.hide_src_code=true
134
135If you want to add or modify several config items, you can do like
136
137	% perf config ui.show-headers=false kmem.default=slab
138
139To modify the sort order of report functionality in user config file(i.e. `~/.perfconfig`), do
140
141	% perf config --user report.sort-order=srcline
142
143To change colors of selected line to other foreground and background colors
144in system config file (i.e. `$(sysconf)/perfconfig`), do
145
146	% perf config --system colors.selected=yellow,green
147
148To query the record mode of call graph, do
149
150	% perf config call-graph.record-mode
151
152If you want to know multiple config key/value pairs, you can do like
153
154	% perf config report.queue-size call-graph.order report.children
155
156To query the config value of sort order of call graph in user config file (i.e. `~/.perfconfig`), do
157
158	% perf config --user call-graph.sort-order
159
160To query the config value of buildid directory in system config file (i.e. `$(sysconf)/perfconfig`), do
161
162	% perf config --system buildid.dir
163
164Variables
165~~~~~~~~~
166
167colors.*::
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:
171
172		medium = green, lightgray
173
174	If you want to use the color configured for you terminal, just leave it
175	as 'default', for example:
176
177		medium = default, lightgray
178
179	Available colors:
180	red, yellow, green, cyan, gray, black, blue,
181	white, default, magenta, lightgray
182
183	colors.top::
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'.
188	colors.medium::
189		'medium' means a overhead percentage which has more than 0.5%.
190		Default values are 'green' and 'default'.
191	colors.normal::
192		'normal' means the rest of overhead percentages
193		except 'top', 'medium', 'selected'.
194		Default values are 'lightgray' and 'default'.
195	colors.selected::
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'.
199	colors.jump_arrows::
200		Colors for jump arrows on assembly code listings
201		such as 'jns', 'jmp', 'jane', etc.
202		Default values are 'blue', 'default'.
203	colors.addr::
204		This selects colors for addresses from 'annotate'.
205		Default values are 'magenta', 'default'.
206	colors.root::
207		Colors for headers in the output of a sub-commands (top, report).
208		Default values are 'white', 'blue'.
209
210core.*::
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.
215
216tui.*, gtk.*::
217	Subcommands that can be configured here are 'top', 'report' and 'annotate'.
218	These values are booleans, for example:
219
220	[tui]
221		top = true
222
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.
225
226buildid.*::
227	buildid.dir::
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.
232
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.
236
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
240
241buildid-cache.*::
242	buildid-cache.debuginfod=URLs
243		Specify debuginfod URLs to be used when retrieving perf.data binaries,
244		it follows the same syntax as the DEBUGINFOD_URLS variable, like:
245
246		  buildid-cache.debuginfod=http://192.168.122.174:8002
247
248annotate.*::
249	These are in control of addresses, jump function, source code
250	in lines of assembly code from a specific program.
251
252	annotate.disassembler_style:
253		Use this to change the default disassembler style to some other value
254		supported by binutils, such as "intel", see the '-M' option help in the
255		'objdump' man page.
256
257	annotate.hide_src_code::
258		If a program which is analyzed has source code,
259		this option lets 'annotate' print a list of assembly code with the source code.
260		For example, let's see a part of a program. There're four lines.
261		If this option is 'true', they can be printed
262		without source code from a program as below.
263
264		│        push   %rbp
265		│        mov    %rsp,%rbp
266		│        sub    $0x10,%rsp
267		│        mov    (%rdi),%rdx
268
269		But if this option is 'false', source code of the part
270		can be also printed as below. Default is 'false'.
271
272		│      struct rb_node *rb_next(const struct rb_node *node)
273		│      {
274		│        push   %rbp
275		│        mov    %rsp,%rbp
276		│        sub    $0x10,%rsp
277		│              struct rb_node *parent;
278279		│              if (RB_EMPTY_NODE(node))
280		│        mov    (%rdi),%rdx
281		│              return n;
282
283		This option works with tui, stdio2 browsers.
284
285        annotate.use_offset::
286		Basing on a first address of a loaded function, offset can be used.
287		Instead of using original addresses of assembly code,
288		addresses subtracted from a base address can be printed.
289		Let's illustrate an example.
290		If a base address is 0XFFFFFFFF81624d50 as below,
291
292		ffffffff81624d50 <load0>
293
294		an address on assembly code has a specific absolute address as below
295
296		ffffffff816250b8:│  mov    0x8(%r14),%rdi
297
298		but if use_offset is 'true', an address subtracted from a base address is printed.
299		Default is true. This option is only applied to TUI.
300
301		             368:│  mov    0x8(%r14),%rdi
302
303		This option works with tui, stdio2 browsers.
304
305	annotate.jump_arrows::
306		There can be jump instruction among assembly code.
307		Depending on a boolean value of jump_arrows,
308		arrows can be printed or not which represent
309		where do the instruction jump into as below.
310
311		│     ┌──jmp    1333
312		│     │  xchg   %ax,%ax
313		│1330:│  mov    %r15,%r10
314		│1333:└─→cmp    %r15,%r14
315
316		If jump_arrow is 'false', the arrows isn't printed as below.
317		Default is 'false'.
318
319		│      ↓ jmp    1333
320		│        xchg   %ax,%ax
321		│1330:   mov    %r15,%r10
322		│1333:   cmp    %r15,%r14
323
324		This option works with tui browser.
325
326        annotate.show_linenr::
327		When showing source code if this option is 'true',
328		line numbers are printed as below.
329
330		│1628         if (type & PERF_SAMPLE_IDENTIFIER) {
331		│     ↓ jne    508
332		│1628                 data->id = *array;
333		│1629                 array++;
334		│1630         }
335
336		However if this option is 'false', they aren't printed as below.
337		Default is 'false'.
338
339		│             if (type & PERF_SAMPLE_IDENTIFIER) {
340		│     ↓ jne    508
341		│                     data->id = *array;
342		│                     array++;
343		│             }
344
345		This option works with tui, stdio2 browsers.
346
347        annotate.show_nr_jumps::
348		Let's see a part of assembly code.
349
350		│1382:   movb   $0x1,-0x270(%rbp)
351
352		If use this, the number of branches jumping to that address can be printed as below.
353		Default is 'false'.
354
355		│1 1382:   movb   $0x1,-0x270(%rbp)
356
357		This option works with tui, stdio2 browsers.
358
359        annotate.show_total_period::
360		To compare two records on an instruction base, with this option
361		provided, display total number of samples that belong to a line
362		in assembly code. If this option is 'true', total periods are printed
363		instead of percent values as below.
364
365		  302 │      mov    %eax,%eax
366
367		But if this option is 'false', percent values for overhead are printed i.e.
368		Default is 'false'.
369
370		99.93 │      mov    %eax,%eax
371
372		This option works with tui, stdio2, stdio browsers.
373
374	annotate.show_nr_samples::
375		By default perf annotate shows percentage of samples. This option
376		can be used to print absolute number of samples. Ex, when set as
377		false:
378
379		Percent│
380		 74.03 │      mov    %fs:0x28,%rax
381
382		When set as true:
383
384		Samples│
385		     6 │      mov    %fs:0x28,%rax
386
387		This option works with tui, stdio2, stdio browsers.
388
389	annotate.offset_level::
390		Default is '1', meaning just jump targets will have offsets show right beside
391		the instruction. When set to '2' 'call' instructions will also have its offsets
392		shown, 3 or higher will show offsets for all instructions.
393
394		This option works with tui, stdio2 browsers.
395
396hist.*::
397	hist.percentage::
398		This option control the way to calculate overhead of filtered entries -
399		that means the value of this option is effective only if there's a
400		filter (by comm, dso or symbol name). Suppose a following example:
401
402		       Overhead  Symbols
403		       ........  .......
404		        33.33%     foo
405		        33.33%     bar
406		        33.33%     baz
407
408	       This is an original overhead and we'll filter out the first 'foo'
409	       entry. The value of 'relative' would increase the overhead of 'bar'
410	       and 'baz' to 50.00% for each, while 'absolute' would show their
411	       current overhead (33.33%).
412
413ui.*::
414	ui.show-headers::
415		This option controls display of column headers (like 'Overhead' and 'Symbol')
416		in 'report' and 'top'. If this option is false, they are hidden.
417		This option is only applied to TUI.
418
419call-graph.*::
420	The following controls the handling of call-graphs (obtained via the
421	-g/--call-graph options).
422
423	call-graph.record-mode::
424		The mode for user space can be 'fp' (frame pointer), 'dwarf'
425		and 'lbr'.  The value 'dwarf' is effective only if libunwind
426		(or a recent version of libdw) is present on the system;
427		the value 'lbr' only works for certain cpus. The method for
428		kernel space is controlled not by this option but by the
429		kernel config (CONFIG_UNWINDER_*).
430
431	call-graph.dump-size::
432		The size of stack to dump in order to do post-unwinding. Default is 8192 (byte).
433		When using dwarf into record-mode, the default size will be used if omitted.
434
435	call-graph.print-type::
436		The print-types can be graph (graph absolute), fractal (graph relative),
437		flat and folded. This option controls a way to show overhead for each callchain
438		entry. Suppose a following example.
439
440                Overhead  Symbols
441                ........  .......
442                  40.00%  foo
443                          |
444                          ---foo
445                             |
446                             |--50.00%--bar
447                             |          main
448                             |
449                              --50.00%--baz
450                                        main
451
452		This output is a 'fractal' format. The 'foo' came from 'bar' and 'baz' exactly
453		half and half so 'fractal' shows 50.00% for each
454		(meaning that it assumes 100% total overhead of 'foo').
455
456		The 'graph' uses absolute overhead value of 'foo' as total so each of
457		'bar' and 'baz' callchain will have 20.00% of overhead.
458		If 'flat' is used, single column and linear exposure of call chains.
459		'folded' mean call chains are displayed in a line, separated by semicolons.
460
461	call-graph.order::
462		This option controls print order of callchains. The default is
463		'callee' which means callee is printed at top and then followed by its
464		caller and so on. The 'caller' prints it in reverse order.
465
466		If this option is not set and report.children or top.children is
467		set to true (or the equivalent command line option is given),
468		the default value of this option is changed to 'caller' for the
469		execution of 'perf report' or 'perf top'. Other commands will
470		still default to 'callee'.
471
472	call-graph.sort-key::
473		The callchains are merged if they contain same information.
474		The sort-key option determines a way to compare the callchains.
475		A value of 'sort-key' can be 'function' or 'address'.
476		The default is 'function'.
477
478	call-graph.threshold::
479		When there're many callchains it'd print tons of lines. So perf omits
480		small callchains under a certain overhead (threshold) and this option
481		control the threshold. Default is 0.5 (%). The overhead is calculated
482		by value depends on call-graph.print-type.
483
484	call-graph.print-limit::
485		This is a maximum number of lines of callchain printed for a single
486		histogram entry. Default is 0 which means no limitation.
487
488report.*::
489	report.sort_order::
490		Allows changing the default sort order from "comm,dso,symbol" to
491		some other default, for instance "sym,dso" may be more fitting for
492		kernel developers.
493	report.percent-limit::
494		This one is mostly the same as call-graph.threshold but works for
495		histogram entries. Entries having an overhead lower than this
496		percentage will not be printed. Default is '0'. If percent-limit
497		is '10', only entries which have more than 10% of overhead will be
498		printed.
499
500	report.queue-size::
501		This option sets up the maximum allocation size of the internal
502		event queue for ordering events. Default is 0, meaning no limit.
503
504	report.children::
505		'Children' means functions called from another function.
506		If this option is true, 'perf report' cumulates callchains of children
507		and show (accumulated) total overhead as well as 'Self' overhead.
508		Please refer to the 'perf report' manual. The default is 'true'.
509
510	report.group::
511		This option is to show event group information together.
512		Example output with this turned on, notice that there is one column
513		per event in the group, ref-cycles and cycles:
514
515		# group: {ref-cycles,cycles}
516		# ========
517		#
518		# Samples: 7K of event 'anon group { ref-cycles, cycles }'
519		# Event count (approx.): 6876107743
520		#
521		#         Overhead  Command      Shared Object               Symbol
522		# ................  .......  .................  ...................
523		#
524		    99.84%  99.76%  noploop  noploop            [.] main
525		     0.07%   0.00%  noploop  ld-2.15.so         [.] strcmp
526		     0.03%   0.00%  noploop  [kernel.kallsyms]  [k] timerqueue_del
527
528top.*::
529	top.children::
530		Same as 'report.children'. So if it is enabled, the output of 'top'
531		command will have 'Children' overhead column as well as 'Self' overhead
532		column by default.
533		The default is 'true'.
534
535	top.call-graph::
536		This is identical to 'call-graph.record-mode', except it is
537		applicable only for 'top' subcommand. This option ONLY setup
538		the unwind method. To enable 'perf top' to actually use it,
539		the command line option -g must be specified.
540
541man.*::
542	man.viewer::
543		This option can assign a tool to view manual pages when 'help'
544		subcommand was invoked. Supported tools are 'man', 'woman'
545		(with emacs client) and 'konqueror'. Default is 'man'.
546
547		New man viewer tool can be also added using 'man.<tool>.cmd'
548		or use different path using 'man.<tool>.path' config option.
549
550pager.*::
551	pager.<subcommand>::
552		When the subcommand is run on stdio, determine whether it uses
553		pager or not based on this value. Default is 'unspecified'.
554
555kmem.*::
556	kmem.default::
557		This option decides which allocator is to be analyzed if neither
558		'--slab' nor '--page' option is used. Default is 'slab'.
559
560record.*::
561	record.build-id::
562		This option can be 'cache', 'no-cache', 'skip' or 'mmap'.
563		'cache' is to post-process data and save/update the binaries into
564		the build-id cache (in ~/.debug). This is the default.
565		But if this option is 'no-cache', it will not update the build-id cache.
566		'skip' skips post-processing and does not update the cache.
567		'mmap' skips post-processing and reads build-ids from MMAP events.
568
569	record.call-graph::
570		This is identical to 'call-graph.record-mode', except it is
571		applicable only for 'record' subcommand. This option ONLY setup
572		the unwind method. To enable 'perf record' to actually use it,
573		the command line option -g must be specified.
574
575	record.aio::
576		Use 'n' control blocks in asynchronous (Posix AIO) trace writing
577		mode ('n' default: 1, max: 4).
578
579diff.*::
580	diff.order::
581		This option sets the number of columns to sort the result.
582		The default is 0, which means sorting by baseline.
583		Setting it to 1 will sort the result by delta (or other
584		compute method selected).
585
586	diff.compute::
587		This options sets the method for computing the diff result.
588		Possible values are 'delta', 'delta-abs', 'ratio' and
589		'wdiff'.  Default is 'delta'.
590
591trace.*::
592	trace.add_events::
593		Allows adding a set of events to add to the ones specified
594		by the user, or use as a default one if none was specified.
595		The initial use case is to add augmented_raw_syscalls.o to
596		activate the 'perf trace' logic that looks for syscall
597		pointer contents after the normal tracepoint payload.
598
599	trace.args_alignment::
600		Number of columns to align the argument list, default is 70,
601		use 40 for the strace default, zero to no alignment.
602
603	trace.no_inherit::
604		Do not follow children threads.
605
606	trace.show_arg_names::
607		Should syscall argument names be printed? If not then trace.show_zeros
608		will be set.
609
610	trace.show_duration::
611		Show syscall duration.
612
613	trace.show_prefix::
614		If set to 'yes' will show common string prefixes in tables. The default
615		is to remove the common prefix in things like "MAP_SHARED", showing just "SHARED".
616
617	trace.show_timestamp::
618		Show syscall start timestamp.
619
620	trace.show_zeros::
621		Do not suppress syscall arguments that are equal to zero.
622
623	trace.tracepoint_beautifiers::
624		Use "libtraceevent" to use that library to augment the tracepoint arguments,
625		"libbeauty", the default, to use the same argument beautifiers used in the
626		strace-like sys_enter+sys_exit lines.
627
628ftrace.*::
629	ftrace.tracer::
630		Can be used to select the default tracer when neither -G nor
631		-F option is not specified. Possible values are 'function' and
632		'function_graph'.
633
634llvm.*::
635	llvm.clang-path::
636		Path to clang. If omit, search it from $PATH.
637
638	llvm.clang-bpf-cmd-template::
639		Cmdline template. Below lines show its default value. Environment
640		variable is used to pass options.
641		"$CLANG_EXEC -D__KERNEL__ -D__NR_CPUS__=$NR_CPUS "\
642		"-DLINUX_VERSION_CODE=$LINUX_VERSION_CODE "	\
643		"$CLANG_OPTIONS $PERF_BPF_INC_OPTIONS $KERNEL_INC_OPTIONS " \
644		"-Wno-unused-value -Wno-pointer-sign "		\
645		"-working-directory $WORKING_DIR "		\
646		"-c \"$CLANG_SOURCE\" -target bpf $CLANG_EMIT_LLVM -O2 -o - $LLVM_OPTIONS_PIPE"
647
648	llvm.clang-opt::
649		Options passed to clang.
650
651	llvm.kbuild-dir::
652		kbuild directory. If not set, use /lib/modules/`uname -r`/build.
653		If set to "" deliberately, skip kernel header auto-detector.
654
655	llvm.kbuild-opts::
656		Options passed to 'make' when detecting kernel header options.
657
658	llvm.dump-obj::
659		Enable perf dump BPF object files compiled by LLVM.
660
661	llvm.opts::
662		Options passed to llc.
663
664samples.*::
665
666	samples.context::
667		Define how many ns worth of time to show
668		around samples in perf report sample context browser.
669
670scripts.*::
671
672	Any option defines a script that is added to the scripts menu
673	in the interactive perf browser and whose output is displayed.
674	The name of the option is the name, the value is a script command line.
675	The script gets the same options passed as a full perf script,
676	in particular -i perfdata file, --cpu, --tid
677
678convert.*::
679
680	convert.queue-size::
681		Limit the size of ordered_events queue, so we could control
682		allocation size of perf data files without proper finished
683		round events.
684stat.*::
685
686	stat.big-num::
687		(boolean) Change the default for "--big-num". To make
688		"--no-big-num" the default, set "stat.big-num=false".
689
690intel-pt.*::
691
692	intel-pt.cache-divisor::
693
694	intel-pt.mispred-all::
695		If set, Intel PT decoder will set the mispred flag on all
696		branches.
697
698auxtrace.*::
699
700	auxtrace.dumpdir::
701		s390 only. The directory to save the auxiliary trace buffer
702		can be changed using this option. Ex, auxtrace.dumpdir=/tmp.
703		If the directory does not exist or has the wrong file type,
704		the current directory is used.
705
706daemon.*::
707
708	daemon.base::
709		Base path for daemon data. All sessions data are stored under
710		this path.
711
712session-<NAME>.*::
713
714	session-<NAME>.run::
715
716		Defines new record session for daemon. The value is record's
717		command line without the 'record' keyword.
718
719
720SEE ALSO
721--------
722linkperf:perf[1]
723