xref: /linux/Documentation/kbuild/kconfig-language.rst (revision 1fd1dc41724319406b0aff221a352a400b0ddfc5)
1================
2Kconfig Language
3================
4
5Introduction
6------------
7
8The configuration database is a collection of configuration options
9organized in a tree structure::
10
11	+- Code maturity level options
12	|  +- Prompt for development and/or incomplete code/drivers
13	+- General setup
14	|  +- Networking support
15	|  +- System V IPC
16	|  +- BSD Process Accounting
17	|  +- Sysctl support
18	+- Loadable module support
19	|  +- Enable loadable module support
20	|     +- Set version information on all module symbols
21	|     +- Kernel module loader
22	+- ...
23
24Every entry has its own dependencies. These dependencies are used
25to determine the visibility of an entry. Any child entry is only
26visible if its parent entry is also visible.
27
28Menu entries
29------------
30
31Most entries define a config option; all other entries help to organize
32them. A single configuration option is defined like this::
33
34  config MODVERSIONS
35	bool "Set version information on all module symbols"
36	depends on MODULES
37	help
38	  Usually, modules have to be recompiled whenever you switch to a new
39	  kernel.  ...
40
41Every line starts with a key word and can be followed by multiple
42arguments.  "config" starts a new config entry. The following lines
43define attributes for this config option. Attributes can be the type of
44the config option, input prompt, dependencies, help text and default
45values. A config option can be defined multiple times with the same
46name, but every definition can have only a single input prompt and the
47type must not conflict.
48
49Menu attributes
50---------------
51
52A menu entry can have a number of attributes. Not all of them are
53applicable everywhere (see syntax).
54
55- type definition: "bool"/"tristate"/"string"/"hex"/"int"
56
57  Every config option must have a type. There are only two basic types:
58  tristate and string; the other types are based on these two. The type
59  definition optionally accepts an input prompt, so these two examples
60  are equivalent::
61
62	bool "Networking support"
63
64  and::
65
66	bool
67	prompt "Networking support"
68
69- input prompt: "prompt" <prompt> ["if" <expr>]
70
71  Every menu entry can have at most one prompt, which is used to display
72  to the user. Optionally dependencies only for this prompt can be added
73  with "if". If a prompt is not present, the config option is a non-visible
74  symbol, meaning its value cannot be directly changed by the user (such as
75  altering the value in ``.config``) and the option will not appear in any
76  config menus. Its value can only be set via "default" and "select" (see
77  below).
78
79- default value: "default" <expr> ["if" <expr>]
80
81  A config option can have any number of default values. If multiple
82  default values are visible, only the first defined one is active.
83  Default values are not limited to the menu entry where they are
84  defined. This means the default can be defined somewhere else or be
85  overridden by an earlier definition.
86  The default value is only assigned to the config symbol if no other
87  value was set by the user (via the input prompt above). If an input
88  prompt is visible the default value is presented to the user and can
89  be overridden by him.
90  Optionally, dependencies only for this default value can be added with
91  "if".
92
93 The default value deliberately defaults to 'n' in order to avoid bloating the
94 build. With few exceptions, new config options should not change this. The
95 intent is for "make oldconfig" to add as little as possible to the config from
96 release to release.
97
98 Note:
99	Things that merit "default y/m" include:
100
101	a) A new Kconfig option for something that used to always be built
102	   should be "default y".
103
104	b) A new gatekeeping Kconfig option that hides/shows other Kconfig
105	   options (but does not generate any code of its own), should be
106	   "default y" so people will see those other options.
107
108	c) Sub-driver behavior or similar options for a driver that is
109	   "default n". This allows you to provide sane defaults.
110
111	d) Hardware or infrastructure that everybody expects, such as CONFIG_NET
112	   or CONFIG_BLOCK. These are rare exceptions.
113
114- type definition + default value::
115
116	"def_bool"/"def_tristate" <expr> ["if" <expr>]
117
118  This is a shorthand notation for a type definition plus a value.
119  Optionally dependencies for this default value can be added with "if".
120
121- dependencies: "depends on" <expr> ["if" <expr>]
122
123  This defines a dependency for this menu entry. If multiple
124  dependencies are defined, they are connected with '&&'. Dependencies
125  are applied to all other options within this menu entry (which also
126  accept an "if" expression), so these two examples are equivalent::
127
128	bool "foo" if BAR
129	default y if BAR
130
131  and::
132
133	depends on BAR
134	bool "foo"
135	default y
136
137  The dependency definition itself may be conditional by appending "if"
138  followed by an expression. For example::
139
140    config FOO
141	tristate
142	depends on BAR if BAZ
143
144  meaning that FOO is constrained by the value of BAR only if BAZ is
145  also set.
146
147- reverse dependencies: "select" <symbol> ["if" <expr>]
148
149  While normal dependencies reduce the upper limit of a symbol (see
150  below), reverse dependencies can be used to force a lower limit of
151  another symbol. The value of the current menu symbol is used as the
152  minimal value <symbol> can be set to. If <symbol> is selected multiple
153  times, the limit is set to the largest selection.
154  Reverse dependencies can only be used with boolean or tristate
155  symbols.
156
157  Note:
158	select should be used with care. select will force
159	a symbol to a value without visiting the dependencies.
160	By abusing select you are able to select a symbol FOO even
161	if FOO depends on BAR that is not set.
162	In general use select only for non-visible symbols
163	(no prompts anywhere) and for symbols with no dependencies.
164	That will limit the usefulness but on the other hand avoid
165	the illegal configurations all over.
166
167	If "select" <symbol> is followed by "if" <expr>, <symbol> will be
168	selected by the logical AND of the value of the current menu symbol
169	and <expr>. This means, the lower limit can be downgraded due to the
170	presence of "if" <expr>. This behavior may seem weird, but we rely on
171	it. (The future of this behavior is undecided.)
172
173- weak reverse dependencies: "imply" <symbol> ["if" <expr>]
174
175  This is similar to "select" as it enforces a lower limit on another
176  symbol except that the "implied" symbol's value may still be set to n
177  from a direct dependency or with a visible prompt.
178
179  Given the following example::
180
181    config FOO
182	tristate "foo"
183	imply BAZ
184
185    config BAZ
186	tristate "baz"
187	depends on BAR
188
189  The following values are possible:
190
191	===		===		=============	==============
192	FOO		BAR		BAZ's default	choice for BAZ
193	===		===		=============	==============
194	n		y		n		N/m/y
195	m		y		m		M/y/n
196	y		y		y		Y/m/n
197	n		m		n		N/m
198	m		m		m		M/n
199	y		m		m		M/n
200	y		n		*		N
201	===		===		=============	==============
202
203  This is useful e.g. with multiple drivers that want to indicate their
204  ability to hook into a secondary subsystem while allowing the user to
205  configure that subsystem out without also having to unset these drivers.
206
207  Note: If the feature provided by BAZ is highly desirable for FOO,
208  FOO should imply not only BAZ, but also its dependency BAR::
209
210    config FOO
211	tristate "foo"
212	imply BAR
213	imply BAZ
214
215  Note: If "imply" <symbol> is followed by "if" <expr>, the default of <symbol>
216  will be the logical AND of the value of the current menu symbol and <expr>.
217  (The future of this behavior is undecided.)
218
219- limiting menu display: "visible if" <expr>
220
221  This attribute is only applicable to menu blocks, if the condition is
222  false, the menu block is not displayed to the user (the symbols
223  contained there can still be selected by other symbols, though). It is
224  similar to a conditional "prompt" attribute for individual menu
225  entries. Default value of "visible" is true.
226
227- numerical ranges: "range" <symbol> <symbol> ["if" <expr>]
228
229  This allows limiting the range of possible input values for int
230  and hex symbols. The user can only input a value which is larger than
231  or equal to the first symbol and smaller than or equal to the second
232  symbol.
233
234- help text: "help"
235
236  This defines a help text. The end of the help text is determined by
237  the indentation level, this means it ends at the first line which has
238  a smaller indentation than the first line of the help text.
239
240- module attribute: "modules"
241  This declares the symbol to be used as the MODULES symbol, which
242  enables the third modular state for all config symbols.
243  At most one symbol may have the "modules" option set.
244
245- transitional attribute: "transitional"
246  This declares the symbol as transitional, meaning it should be processed
247  during configuration but omitted from newly written .config files.
248  Transitional symbols are useful for backward compatibility during config
249  option migrations - they allow olddefconfig to process existing .config
250  files while ensuring the old option doesn't appear in new configurations.
251
252  A transitional symbol:
253  - Has no prompt (is not visible to users in menus)
254  - Is processed normally during configuration (values are read and used)
255  - Can be referenced in default expressions of other symbols
256  - Is not written to new .config files
257  - Cannot have any other properties (it is a pass-through option)
258
259  Example migration from OLD_NAME to NEW_NAME::
260
261    config NEW_NAME
262	bool "New option name"
263	default OLD_NAME
264	help
265	  This replaces the old CONFIG_OLD_NAME option.
266
267    config OLD_NAME
268	bool
269	transitional
270	help
271	  Transitional config for OLD_NAME to NEW_NAME migration.
272
273  With this setup, existing .config files with "CONFIG_OLD_NAME=y" will
274  result in "CONFIG_NEW_NAME=y" being set, while CONFIG_OLD_NAME will be
275  omitted from newly written .config files.
276
277Menu dependencies
278-----------------
279
280Dependencies define the visibility of a menu entry and can also reduce
281the input range of tristate symbols. The tristate logic used in the
282expressions uses one more state than normal boolean logic to express the
283module state. Dependency expressions have the following syntax::
284
285  <expr> ::= <symbol>                           (1)
286           <symbol> '=' <symbol>                (2)
287           <symbol> '!=' <symbol>               (3)
288           <symbol1> '<' <symbol2>              (4)
289           <symbol1> '>' <symbol2>              (4)
290           <symbol1> '<=' <symbol2>             (4)
291           <symbol1> '>=' <symbol2>             (4)
292           '(' <expr> ')'                       (5)
293           '!' <expr>                           (6)
294           <expr> '&&' <expr>                   (7)
295           <expr> '||' <expr>                   (8)
296
297Expressions are listed in decreasing order of precedence.
298
299(1) Convert the symbol into an expression. Boolean and tristate symbols
300    are simply converted into the respective expression values. All
301    other symbol types result in 'n'.
302(2) If the values of both symbols are equal, it returns 'y',
303    otherwise 'n'.
304(3) If the values of both symbols are equal, it returns 'n',
305    otherwise 'y'.
306(4) If value of <symbol1> is respectively lower, greater, lower-or-equal,
307    or greater-or-equal than value of <symbol2>, it returns 'y',
308    otherwise 'n'.
309(5) Returns the value of the expression. Used to override precedence.
310(6) Returns the result of (2-/expr/).
311(7) Returns the result of min(/expr/, /expr/).
312(8) Returns the result of max(/expr/, /expr/).
313
314An expression can have a value of 'n', 'm' or 'y' (or 0, 1, 2
315respectively for calculations). A menu entry becomes visible when its
316expression evaluates to 'm' or 'y'.
317
318There are two types of symbols: constant and non-constant symbols.
319Non-constant symbols are the most common ones and are defined with the
320'config' statement. Non-constant symbols consist entirely of alphanumeric
321characters or underscores.
322Constant symbols are only part of expressions. Constant symbols are
323always surrounded by single or double quotes. Within the quote, any
324other character is allowed and the quotes can be escaped using '\'.
325
326Menu structure
327--------------
328
329The position of a menu entry in the tree is determined in two ways. First
330it can be specified explicitly::
331
332  menu "Network device support"
333	depends on NET
334
335  config NETDEVICES
336	...
337
338  endmenu
339
340All entries within the "menu" ... "endmenu" block become a submenu of
341"Network device support". All subentries inherit the dependencies from
342the menu entry, e.g. this means the dependency "NET" is added to the
343dependency list of the config option NETDEVICES.
344
345The other way to generate the menu structure is done by analyzing the
346dependencies. If a menu entry somehow depends on the previous entry, it
347can be made a submenu of it. First, the previous (parent) symbol must
348be part of the dependency list and then one of these two conditions
349must be true:
350
351- the child entry must become invisible, if the parent is set to 'n'
352- the child entry must only be visible, if the parent is visible::
353
354    config MODULES
355	bool "Enable loadable module support"
356
357    config MODVERSIONS
358	bool "Set version information on all module symbols"
359	depends on MODULES
360
361    comment "module support disabled"
362	depends on !MODULES
363
364MODVERSIONS directly depends on MODULES, this means it's only visible if
365MODULES is different from 'n'. The comment on the other hand is only
366visible when MODULES is set to 'n'.
367
368
369Kconfig syntax
370--------------
371
372The configuration file describes a series of menu entries, where every
373line starts with a keyword (except help texts). The following keywords
374end a menu entry:
375
376- config
377- menuconfig
378- choice/endchoice
379- comment
380- menu/endmenu
381- if/endif
382- source
383
384The first five also start the definition of a menu entry.
385
386config::
387
388	"config" <symbol>
389	<config options>
390
391This defines a config symbol <symbol> and accepts any of above
392attributes as options.
393
394menuconfig::
395
396	"menuconfig" <symbol>
397	<config options>
398
399This is similar to the simple config entry above, but it also gives a
400hint to front ends, that all suboptions should be displayed as a
401separate list of options. To make sure all the suboptions will really
402show up under the menuconfig entry and not outside of it, every item
403from the <config options> list must depend on the menuconfig symbol.
404In practice, this is achieved by using one of the next two constructs::
405
406  (1):
407  menuconfig M
408  if M
409      config C1
410      config C2
411  endif
412
413  (2):
414  menuconfig M
415  config C1
416      depends on M
417  config C2
418      depends on M
419
420In the following examples (3) and (4), C1 and C2 still have the M
421dependency, but will not appear under menuconfig M anymore, because
422of C0, which doesn't depend on M::
423
424  (3):
425  menuconfig M
426      config C0
427  if M
428      config C1
429      config C2
430  endif
431
432  (4):
433  menuconfig M
434  config C0
435  config C1
436      depends on M
437  config C2
438      depends on M
439
440choices::
441
442	"choice"
443	<choice options>
444	<choice block>
445	"endchoice"
446
447This defines a choice group and accepts "prompt", "default", "depends on", and
448"help" attributes as options.
449
450A choice only allows a single config entry to be selected.
451
452comment::
453
454	"comment" <prompt>
455	<comment options>
456
457This defines a comment which is displayed to the user during the
458configuration process and is also echoed to the output files. The only
459possible options are dependencies.
460
461menu::
462
463	"menu" <prompt>
464	<menu options>
465	<menu block>
466	"endmenu"
467
468This defines a menu block, see "Menu structure" above for more
469information. The only possible options are dependencies and "visible"
470attributes.
471
472if::
473
474	"if" <expr>
475	<if block>
476	"endif"
477
478This defines an if block. The dependency expression <expr> is appended
479to all enclosed menu entries.
480
481source::
482
483	"source" <prompt>
484
485This reads the specified configuration file. This file is always parsed.
486
487mainmenu::
488
489	"mainmenu" <prompt>
490
491This sets the config program's title bar if the config program chooses
492to use it. It should be placed at the top of the configuration, before any
493other statement.
494
495'#' Kconfig source file comment:
496
497An unquoted '#' character anywhere in a source file line indicates
498the beginning of a source file comment.  The remainder of that line
499is a comment.
500
501
502Kconfig hints
503-------------
504This is a collection of Kconfig tips, most of which aren't obvious at
505first glance and most of which have become idioms in several Kconfig
506files.
507
508Adding common features and make the usage configurable
509~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
510It is a common idiom to implement a feature/functionality that are
511relevant for some architectures but not all.
512The recommended way to do so is to use a config variable named HAVE_*
513that is defined in a common Kconfig file and selected by the relevant
514architectures.
515An example is the generic IOMAP functionality.
516
517We would in lib/Kconfig see::
518
519  # Generic IOMAP is used to ...
520  config HAVE_GENERIC_IOMAP
521
522  config GENERIC_IOMAP
523	depends on HAVE_GENERIC_IOMAP && FOO
524
525And in lib/Makefile we would see::
526
527	obj-$(CONFIG_GENERIC_IOMAP) += iomap.o
528
529For each architecture using the generic IOMAP functionality we would see::
530
531  config X86
532	select ...
533	select HAVE_GENERIC_IOMAP
534	select ...
535
536Note: we use the existing config option and avoid creating a new
537config variable to select HAVE_GENERIC_IOMAP.
538
539Note: the use of the internal config variable HAVE_GENERIC_IOMAP, it is
540introduced to overcome the limitation of select which will force a
541config option to 'y' no matter the dependencies.
542The dependencies are moved to the symbol GENERIC_IOMAP and we avoid the
543situation where select forces a symbol equals to 'y'.
544
545Adding features that need compiler support
546~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
547
548There are several features that need compiler support. The recommended way
549to describe the dependency on the compiler feature is to use "depends on"
550followed by a test macro::
551
552  config STACKPROTECTOR
553	bool "Stack Protector buffer overflow detection"
554	depends on $(cc-option,-fstack-protector)
555	...
556
557If you need to expose a compiler capability to makefiles and/or C source files,
558`CC_HAS_` is the recommended prefix for the config option::
559
560  config CC_HAS_FOO
561	def_bool $(success,$(srctree)/scripts/cc-check-foo.sh $(CC))
562
563Build as module only
564~~~~~~~~~~~~~~~~~~~~
565To restrict a component build to module-only, qualify its config symbol
566with "depends on m".  E.g.::
567
568  config FOO
569	depends on BAR && m
570
571limits FOO to module (=m) or disabled (=n).
572
573Compile-testing
574~~~~~~~~~~~~~~~
575If a config symbol has a dependency, but the code controlled by the config
576symbol can still be compiled if the dependency is not met, it is encouraged to
577increase build coverage by adding an "|| COMPILE_TEST" clause to the
578dependency. This is especially useful for drivers for more exotic hardware, as
579it allows continuous-integration systems to compile-test the code on a more
580common system, and detect bugs that way.
581Note that compile-tested code should avoid crashing when run on a system where
582the dependency is not met.
583
584Architecture and platform dependencies
585~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
586Due to the presence of stubs, most drivers can now be compiled on most
587architectures. However, this does not mean it makes sense to have all drivers
588available everywhere, as the actual hardware may only exist on specific
589architectures and platforms. This is especially true for on-SoC IP cores,
590which may be limited to a specific vendor or SoC family.
591
592To prevent asking the user about drivers that cannot be used on the system(s)
593the user is compiling a kernel for, and if it makes sense, config symbols
594controlling the compilation of a driver should contain proper dependencies,
595limiting the visibility of the symbol to (a superset of) the platform(s) the
596driver can be used on. The dependency can be an architecture (e.g. ARM) or
597platform (e.g. ARCH_OMAP4) dependency. This makes life simpler not only for
598distro config owners, but also for every single developer or user who
599configures a kernel.
600
601Such a dependency can be relaxed by combining it with the compile-testing rule
602above, leading to:
603
604  config FOO
605	bool "Support for foo hardware"
606	depends on ARCH_FOO_VENDOR || COMPILE_TEST
607
608Optional dependencies
609~~~~~~~~~~~~~~~~~~~~~
610
611Some drivers are able to optionally use a feature from another module
612or build cleanly with that module disabled, but cause a link failure
613when trying to use that loadable module from a built-in driver.
614
615The recommended way to express this optional dependency in Kconfig logic
616uses the conditional form::
617
618  config FOO
619	tristate "Support for foo hardware"
620	depends on BAR if BAR
621
622This slightly counterintuitive style is also widely used::
623
624  config FOO
625	tristate "Support for foo hardware"
626	depends on BAR || !BAR
627
628This means that there is either a dependency on BAR that disallows
629the combination of FOO=y with BAR=m, or BAR is completely disabled. The BAR
630module must provide all the stubs for !BAR case.
631
632For a more formalized approach if there are multiple drivers that have
633the same dependency, a helper symbol can be used, like::
634
635  config FOO
636	tristate "Support for foo hardware"
637	depends on BAR_OPTIONAL
638
639  config BAR_OPTIONAL
640	def_tristate BAR || !BAR
641
642Much less favorable way to express optional dependency is IS_REACHABLE() within
643the module code, useful for example when the module BAR does not provide
644!BAR stubs::
645
646	foo_init()
647	{
648		if (IS_REACHABLE(CONFIG_BAR))
649			bar_register(&foo);
650		...
651	}
652
653IS_REACHABLE() is generally discouraged, because the code will be silently
654discarded, when CONFIG_BAR=m and this code is built-in. This is not what users
655usually expect when enabling BAR as module.
656
657Kconfig recursive dependency limitations
658~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
659
660If you've hit the Kconfig error: "recursive dependency detected" you've run
661into a recursive dependency issue with Kconfig, a recursive dependency can be
662summarized as a circular dependency. The kconfig tools need to ensure that
663Kconfig files comply with specified configuration requirements. In order to do
664that kconfig must determine the values that are possible for all Kconfig
665symbols, this is currently not possible if there is a circular relation
666between two or more Kconfig symbols. For more details refer to the "Simple
667Kconfig recursive issue" subsection below. Kconfig does not do recursive
668dependency resolution; this has a few implications for Kconfig file writers.
669We'll first explain why this issues exists and then provide an example
670technical limitation which this brings upon Kconfig developers. Eager
671developers wishing to try to address this limitation should read the next
672subsections.
673
674Simple Kconfig recursive issue
675~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
676
677Read: Documentation/kbuild/Kconfig.recursion-issue-01
678
679Test with::
680
681  make KBUILD_KCONFIG=Documentation/kbuild/Kconfig.recursion-issue-01 allnoconfig
682
683Cumulative Kconfig recursive issue
684~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
685
686Read: Documentation/kbuild/Kconfig.recursion-issue-02
687
688Test with::
689
690  make KBUILD_KCONFIG=Documentation/kbuild/Kconfig.recursion-issue-02 allnoconfig
691
692Practical solutions to kconfig recursive issue
693~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
694
695Developers who run into the recursive Kconfig issue have two options
696at their disposal. We document them below and also provide a list of
697historical issues resolved through these different solutions.
698
699  a) Remove any superfluous "select FOO" or "depends on FOO"
700  b) Match dependency semantics:
701
702	b1) Swap all "select FOO" to "depends on FOO" or,
703
704	b2) Swap all "depends on FOO" to "select FOO"
705
706The resolution to a) can be tested with the sample Kconfig file
707Documentation/kbuild/Kconfig.recursion-issue-01 through the removal
708of the "select CORE" from CORE_BELL_A_ADVANCED as that is implicit already
709since CORE_BELL_A depends on CORE. At times it may not be possible to remove
710some dependency criteria, for such cases you can work with solution b).
711
712The two different resolutions for b) can be tested in the sample Kconfig file
713Documentation/kbuild/Kconfig.recursion-issue-02.
714
715Below is a list of examples of prior fixes for these types of recursive issues;
716all errors appear to involve one or more "select" statements and one or more
717"depends on".
718
719============    ===================================
720commit          fix
721============    ===================================
72206b718c01208    select A -> depends on A
723c22eacfe82f9    depends on A -> depends on B
7246a91e854442c    select A -> depends on A
725118c565a8f2e    select A -> select B
726f004e5594705    select A -> depends on A
727c7861f37b4c6    depends on A -> (null)
72880c69915e5fb    select A -> (null)              (1)
729c2218e26c0d0    select A -> depends on A        (1)
730d6ae99d04e1c    select A -> depends on A
73195ca19cf8cbf    select A -> depends on A
7328f057d7bca54    depends on A -> (null)
7338f057d7bca54    depends on A -> select A
734a0701f04846e    select A -> depends on A
7350c8b92f7f259    depends on A -> (null)
736e4e9e0540928    select A -> depends on A        (2)
7377453ea886e87    depends on A > (null)           (1)
7387b1fff7e4fdf    select A -> depends on A
73986c747d2a4f0    select A -> depends on A
740d9f9ab51e55e    select A -> depends on A
7410c51a4d8abd6    depends on A -> select A        (3)
742e98062ed6dc4    select A -> depends on A        (3)
74391e5d284a7f1    select A -> (null)
744============    ===================================
745
746(1) Partial (or no) quote of error.
747(2) That seems to be the gist of that fix.
748(3) Same error.
749
750Future kconfig work
751~~~~~~~~~~~~~~~~~~~
752
753Work on kconfig is welcomed on both areas of clarifying semantics and on
754evaluating the use of a full SAT solver for it. A full SAT solver can be
755desirable to enable more complex dependency mappings and / or queries,
756for instance one possible use case for a SAT solver could be that of handling
757the current known recursive dependency issues. It is not known if this would
758address such issues but such evaluation is desirable. If support for a full SAT
759solver proves too complex or that it cannot address recursive dependency issues
760Kconfig should have at least clear and well defined semantics which also
761addresses and documents limitations or requirements such as the ones dealing
762with recursive dependencies.
763
764Further work on both of these areas is welcomed on Kconfig. We elaborate
765on both of these in the next two subsections.
766
767Semantics of Kconfig
768~~~~~~~~~~~~~~~~~~~~
769
770The use of Kconfig is broad, Linux is now only one of Kconfig's users:
771one study has completed a broad analysis of Kconfig use in 12 projects [0]_.
772Despite its widespread use, and although this document does a reasonable job
773in documenting basic Kconfig syntax a more precise definition of Kconfig
774semantics is welcomed. One project deduced Kconfig semantics through
775the use of the xconfig configurator [1]_. Work should be done to confirm if
776the deduced semantics matches our intended Kconfig design goals.
777Another project formalized a denotational semantics of a core subset of
778the Kconfig language [10]_.
779
780Having well defined semantics can be useful for tools for practical
781evaluation of dependencies, for instance one such case was work to
782express in boolean abstraction of the inferred semantics of Kconfig to
783translate Kconfig logic into boolean formulas and run a SAT solver on this to
784find dead code / features (always inactive), 114 dead features were found in
785Linux using this methodology [1]_ (Section 8: Threats to validity).
786The kismet tool, based on the semantics in [10]_, finds abuses of reverse
787dependencies and has led to dozens of committed fixes to Linux Kconfig files [11]_.
788
789Confirming this could prove useful as Kconfig stands as one of the leading
790industrial variability modeling languages [1]_ [2]_. Its study would help
791evaluate practical uses of such languages, their use was only theoretical
792and real world requirements were not well understood. As it stands though
793only reverse engineering techniques have been used to deduce semantics from
794variability modeling languages such as Kconfig [3]_.
795
796.. [0] https://www.eng.uwaterloo.ca/~shshe/kconfig_semantics.pdf
797.. [1] https://gsd.uwaterloo.ca/sites/default/files/vm-2013-berger.pdf
798.. [2] https://gsd.uwaterloo.ca/sites/default/files/ase241-berger_0.pdf
799.. [3] https://gsd.uwaterloo.ca/sites/default/files/icse2011.pdf
800
801Full SAT solver for Kconfig
802~~~~~~~~~~~~~~~~~~~~~~~~~~~
803
804Although SAT solvers [4]_ haven't yet been used by Kconfig directly, as noted
805in the previous subsection, work has been done however to express in boolean
806abstraction the inferred semantics of Kconfig to translate Kconfig logic into
807boolean formulas and run a SAT solver on it [5]_. Another known related project
808is CADOS [6]_ (former VAMOS [7]_) and the tools, mainly undertaker [8]_, which
809has been introduced first with [9]_.  The basic concept of undertaker is to
810extract variability models from Kconfig and put them together with a
811propositional formula extracted from CPP #ifdefs and build-rules into a SAT
812solver in order to find dead code, dead files, and dead symbols. If using a SAT
813solver is desirable on Kconfig one approach would be to evaluate repurposing
814such efforts somehow on Kconfig. There is enough interest from mentors of
815existing projects to not only help advise how to integrate this work upstream
816but also help maintain it long term. Interested developers should visit:
817
818https://kernelnewbies.org/KernelProjects/kconfig-sat
819
820.. [4] https://www.cs.cornell.edu/~sabhar/chapters/SATSolvers-KR-Handbook.pdf
821.. [5] https://gsd.uwaterloo.ca/sites/default/files/vm-2013-berger.pdf
822.. [6] https://cados.cs.fau.de
823.. [7] https://vamos.cs.fau.de
824.. [8] https://undertaker.cs.fau.de
825.. [9] https://www4.cs.fau.de/Publications/2011/tartler_11_eurosys.pdf
826.. [10] https://paulgazzillo.com/papers/esecfse21.pdf
827.. [11] https://github.com/paulgazz/kmax
828