1.. _codingstyle:
2
3Linux kernel coding style
4=========================
5
6This is a short document describing the preferred coding style for the
7linux kernel.  Coding style is very personal, and I won't **force** my
8views on anybody, but this is what goes for anything that I have to be
9able to maintain, and I'd prefer it for most other things too.  Please
10at least consider the points made here.
11
12First off, I'd suggest printing out a copy of the GNU coding standards,
13and NOT read it.  Burn them, it's a great symbolic gesture.
14
15Anyway, here goes:
16
17
181) Indentation
19--------------
20
21Tabs are 8 characters, and thus indentations are also 8 characters.
22There are heretic movements that try to make indentations 4 (or even 2!)
23characters deep, and that is akin to trying to define the value of PI to
24be 3.
25
26Rationale: The whole idea behind indentation is to clearly define where
27a block of control starts and ends.  Especially when you've been looking
28at your screen for 20 straight hours, you'll find it a lot easier to see
29how the indentation works if you have large indentations.
30
31Now, some people will claim that having 8-character indentations makes
32the code move too far to the right, and makes it hard to read on a
3380-character terminal screen.  The answer to that is that if you need
34more than 3 levels of indentation, you're screwed anyway, and should fix
35your program.
36
37In short, 8-char indents make things easier to read, and have the added
38benefit of warning you when you're nesting your functions too deep.
39Heed that warning.
40
41The preferred way to ease multiple indentation levels in a switch statement is
42to align the ``switch`` and its subordinate ``case`` labels in the same column
43instead of ``double-indenting`` the ``case`` labels.  E.g.:
44
45.. code-block:: c
46
47	switch (suffix) {
48	case 'G':
49	case 'g':
50		mem <<= 30;
51		break;
52	case 'M':
53	case 'm':
54		mem <<= 20;
55		break;
56	case 'K':
57	case 'k':
58		mem <<= 10;
59		fallthrough;
60	default:
61		break;
62	}
63
64Don't put multiple statements on a single line unless you have
65something to hide:
66
67.. code-block:: c
68
69	if (condition) do_this;
70	  do_something_everytime;
71
72Don't put multiple assignments on a single line either.  Kernel coding style
73is super simple.  Avoid tricky expressions.
74
75Outside of comments, documentation and except in Kconfig, spaces are never
76used for indentation, and the above example is deliberately broken.
77
78Get a decent editor and don't leave whitespace at the end of lines.
79
80
812) Breaking long lines and strings
82----------------------------------
83
84Coding style is all about readability and maintainability using commonly
85available tools.
86
87The preferred limit on the length of a single line is 80 columns.
88
89Statements longer than 80 columns should be broken into sensible chunks,
90unless exceeding 80 columns significantly increases readability and does
91not hide information.
92
93Descendants are always substantially shorter than the parent and
94are placed substantially to the right.  A very commonly used style
95is to align descendants to a function open parenthesis.
96
97These same rules are applied to function headers with a long argument list.
98
99However, never break user-visible strings such as printk messages because
100that breaks the ability to grep for them.
101
102
1033) Placing Braces and Spaces
104----------------------------
105
106The other issue that always comes up in C styling is the placement of
107braces.  Unlike the indent size, there are few technical reasons to
108choose one placement strategy over the other, but the preferred way, as
109shown to us by the prophets Kernighan and Ritchie, is to put the opening
110brace last on the line, and put the closing brace first, thusly:
111
112.. code-block:: c
113
114	if (x is true) {
115		we do y
116	}
117
118This applies to all non-function statement blocks (if, switch, for,
119while, do).  E.g.:
120
121.. code-block:: c
122
123	switch (action) {
124	case KOBJ_ADD:
125		return "add";
126	case KOBJ_REMOVE:
127		return "remove";
128	case KOBJ_CHANGE:
129		return "change";
130	default:
131		return NULL;
132	}
133
134However, there is one special case, namely functions: they have the
135opening brace at the beginning of the next line, thus:
136
137.. code-block:: c
138
139	int function(int x)
140	{
141		body of function
142	}
143
144Heretic people all over the world have claimed that this inconsistency
145is ...  well ...  inconsistent, but all right-thinking people know that
146(a) K&R are **right** and (b) K&R are right.  Besides, functions are
147special anyway (you can't nest them in C).
148
149Note that the closing brace is empty on a line of its own, **except** in
150the cases where it is followed by a continuation of the same statement,
151ie a ``while`` in a do-statement or an ``else`` in an if-statement, like
152this:
153
154.. code-block:: c
155
156	do {
157		body of do-loop
158	} while (condition);
159
160and
161
162.. code-block:: c
163
164	if (x == y) {
165		..
166	} else if (x > y) {
167		...
168	} else {
169		....
170	}
171
172Rationale: K&R.
173
174Also, note that this brace-placement also minimizes the number of empty
175(or almost empty) lines, without any loss of readability.  Thus, as the
176supply of new-lines on your screen is not a renewable resource (think
17725-line terminal screens here), you have more empty lines to put
178comments on.
179
180Do not unnecessarily use braces where a single statement will do.
181
182.. code-block:: c
183
184	if (condition)
185		action();
186
187and
188
189.. code-block:: none
190
191	if (condition)
192		do_this();
193	else
194		do_that();
195
196This does not apply if only one branch of a conditional statement is a single
197statement; in the latter case use braces in both branches:
198
199.. code-block:: c
200
201	if (condition) {
202		do_this();
203		do_that();
204	} else {
205		otherwise();
206	}
207
208Also, use braces when a loop contains more than a single simple statement:
209
210.. code-block:: c
211
212	while (condition) {
213		if (test)
214			do_something();
215	}
216
2173.1) Spaces
218***********
219
220Linux kernel style for use of spaces depends (mostly) on
221function-versus-keyword usage.  Use a space after (most) keywords.  The
222notable exceptions are sizeof, typeof, alignof, and __attribute__, which look
223somewhat like functions (and are usually used with parentheses in Linux,
224although they are not required in the language, as in: ``sizeof info`` after
225``struct fileinfo info;`` is declared).
226
227So use a space after these keywords::
228
229	if, switch, case, for, do, while
230
231but not with sizeof, typeof, alignof, or __attribute__.  E.g.,
232
233.. code-block:: c
234
235
236	s = sizeof(struct file);
237
238Do not add spaces around (inside) parenthesized expressions.  This example is
239**bad**:
240
241.. code-block:: c
242
243
244	s = sizeof( struct file );
245
246When declaring pointer data or a function that returns a pointer type, the
247preferred use of ``*`` is adjacent to the data name or function name and not
248adjacent to the type name.  Examples:
249
250.. code-block:: c
251
252
253	char *linux_banner;
254	unsigned long long memparse(char *ptr, char **retptr);
255	char *match_strdup(substring_t *s);
256
257Use one space around (on each side of) most binary and ternary operators,
258such as any of these::
259
260	=  +  -  <  >  *  /  %  |  &  ^  <=  >=  ==  !=  ?  :
261
262but no space after unary operators::
263
264	&  *  +  -  ~  !  sizeof  typeof  alignof  __attribute__  defined
265
266no space before the postfix increment & decrement unary operators::
267
268	++  --
269
270no space after the prefix increment & decrement unary operators::
271
272	++  --
273
274and no space around the ``.`` and ``->`` structure member operators.
275
276Do not leave trailing whitespace at the ends of lines.  Some editors with
277``smart`` indentation will insert whitespace at the beginning of new lines as
278appropriate, so you can start typing the next line of code right away.
279However, some such editors do not remove the whitespace if you end up not
280putting a line of code there, such as if you leave a blank line.  As a result,
281you end up with lines containing trailing whitespace.
282
283Git will warn you about patches that introduce trailing whitespace, and can
284optionally strip the trailing whitespace for you; however, if applying a series
285of patches, this may make later patches in the series fail by changing their
286context lines.
287
288
2894) Naming
290---------
291
292C is a Spartan language, and your naming conventions should follow suit.
293Unlike Modula-2 and Pascal programmers, C programmers do not use cute
294names like ThisVariableIsATemporaryCounter. A C programmer would call that
295variable ``tmp``, which is much easier to write, and not the least more
296difficult to understand.
297
298HOWEVER, while mixed-case names are frowned upon, descriptive names for
299global variables are a must.  To call a global function ``foo`` is a
300shooting offense.
301
302GLOBAL variables (to be used only if you **really** need them) need to
303have descriptive names, as do global functions.  If you have a function
304that counts the number of active users, you should call that
305``count_active_users()`` or similar, you should **not** call it ``cntusr()``.
306
307Encoding the type of a function into the name (so-called Hungarian
308notation) is asinine - the compiler knows the types anyway and can check
309those, and it only confuses the programmer. No wonder Microsoft makes buggy
310programs.
311
312LOCAL variable names should be short, and to the point.  If you have
313some random integer loop counter, it should probably be called ``i``.
314Calling it ``loop_counter`` is non-productive, if there is no chance of it
315being mis-understood.  Similarly, ``tmp`` can be just about any type of
316variable that is used to hold a temporary value.
317
318If you are afraid to mix up your local variable names, you have another
319problem, which is called the function-growth-hormone-imbalance syndrome.
320See chapter 6 (Functions).
321
322For symbol names and documentation, avoid introducing new usage of
323'master / slave' (or 'slave' independent of 'master') and 'blacklist /
324whitelist'.
325
326Recommended replacements for 'master / slave' are:
327    '{primary,main} / {secondary,replica,subordinate}'
328    '{initiator,requester} / {target,responder}'
329    '{controller,host} / {device,worker,proxy}'
330    'leader / follower'
331    'director / performer'
332
333Recommended replacements for 'blacklist/whitelist' are:
334    'denylist / allowlist'
335    'blocklist / passlist'
336
337Exceptions for introducing new usage is to maintain a userspace ABI/API,
338or when updating code for an existing (as of 2020) hardware or protocol
339specification that mandates those terms. For new specifications
340translate specification usage of the terminology to the kernel coding
341standard where possible.
342
3435) Typedefs
344-----------
345
346Please don't use things like ``vps_t``.
347It's a **mistake** to use typedef for structures and pointers. When you see a
348
349.. code-block:: c
350
351
352	vps_t a;
353
354in the source, what does it mean?
355In contrast, if it says
356
357.. code-block:: c
358
359	struct virtual_container *a;
360
361you can actually tell what ``a`` is.
362
363Lots of people think that typedefs ``help readability``. Not so. They are
364useful only for:
365
366 (a) totally opaque objects (where the typedef is actively used to **hide**
367     what the object is).
368
369     Example: ``pte_t`` etc. opaque objects that you can only access using
370     the proper accessor functions.
371
372     .. note::
373
374       Opaqueness and ``accessor functions`` are not good in themselves.
375       The reason we have them for things like pte_t etc. is that there
376       really is absolutely **zero** portably accessible information there.
377
378 (b) Clear integer types, where the abstraction **helps** avoid confusion
379     whether it is ``int`` or ``long``.
380
381     u8/u16/u32 are perfectly fine typedefs, although they fit into
382     category (d) better than here.
383
384     .. note::
385
386       Again - there needs to be a **reason** for this. If something is
387       ``unsigned long``, then there's no reason to do
388
389	typedef unsigned long myflags_t;
390
391     but if there is a clear reason for why it under certain circumstances
392     might be an ``unsigned int`` and under other configurations might be
393     ``unsigned long``, then by all means go ahead and use a typedef.
394
395 (c) when you use sparse to literally create a **new** type for
396     type-checking.
397
398 (d) New types which are identical to standard C99 types, in certain
399     exceptional circumstances.
400
401     Although it would only take a short amount of time for the eyes and
402     brain to become accustomed to the standard types like ``uint32_t``,
403     some people object to their use anyway.
404
405     Therefore, the Linux-specific ``u8/u16/u32/u64`` types and their
406     signed equivalents which are identical to standard types are
407     permitted -- although they are not mandatory in new code of your
408     own.
409
410     When editing existing code which already uses one or the other set
411     of types, you should conform to the existing choices in that code.
412
413 (e) Types safe for use in userspace.
414
415     In certain structures which are visible to userspace, we cannot
416     require C99 types and cannot use the ``u32`` form above. Thus, we
417     use __u32 and similar types in all structures which are shared
418     with userspace.
419
420Maybe there are other cases too, but the rule should basically be to NEVER
421EVER use a typedef unless you can clearly match one of those rules.
422
423In general, a pointer, or a struct that has elements that can reasonably
424be directly accessed should **never** be a typedef.
425
426
4276) Functions
428------------
429
430Functions should be short and sweet, and do just one thing.  They should
431fit on one or two screenfuls of text (the ISO/ANSI screen size is 80x24,
432as we all know), and do one thing and do that well.
433
434The maximum length of a function is inversely proportional to the
435complexity and indentation level of that function.  So, if you have a
436conceptually simple function that is just one long (but simple)
437case-statement, where you have to do lots of small things for a lot of
438different cases, it's OK to have a longer function.
439
440However, if you have a complex function, and you suspect that a
441less-than-gifted first-year high-school student might not even
442understand what the function is all about, you should adhere to the
443maximum limits all the more closely.  Use helper functions with
444descriptive names (you can ask the compiler to in-line them if you think
445it's performance-critical, and it will probably do a better job of it
446than you would have done).
447
448Another measure of the function is the number of local variables.  They
449shouldn't exceed 5-10, or you're doing something wrong.  Re-think the
450function, and split it into smaller pieces.  A human brain can
451generally easily keep track of about 7 different things, anything more
452and it gets confused.  You know you're brilliant, but maybe you'd like
453to understand what you did 2 weeks from now.
454
455In source files, separate functions with one blank line.  If the function is
456exported, the **EXPORT** macro for it should follow immediately after the
457closing function brace line.  E.g.:
458
459.. code-block:: c
460
461	int system_is_up(void)
462	{
463		return system_state == SYSTEM_RUNNING;
464	}
465	EXPORT_SYMBOL(system_is_up);
466
467In function prototypes, include parameter names with their data types.
468Although this is not required by the C language, it is preferred in Linux
469because it is a simple way to add valuable information for the reader.
470
471Do not use the ``extern`` keyword with function prototypes as this makes
472lines longer and isn't strictly necessary.
473
474
4757) Centralized exiting of functions
476-----------------------------------
477
478Albeit deprecated by some people, the equivalent of the goto statement is
479used frequently by compilers in form of the unconditional jump instruction.
480
481The goto statement comes in handy when a function exits from multiple
482locations and some common work such as cleanup has to be done.  If there is no
483cleanup needed then just return directly.
484
485Choose label names which say what the goto does or why the goto exists.  An
486example of a good name could be ``out_free_buffer:`` if the goto frees ``buffer``.
487Avoid using GW-BASIC names like ``err1:`` and ``err2:``, as you would have to
488renumber them if you ever add or remove exit paths, and they make correctness
489difficult to verify anyway.
490
491The rationale for using gotos is:
492
493- unconditional statements are easier to understand and follow
494- nesting is reduced
495- errors by not updating individual exit points when making
496  modifications are prevented
497- saves the compiler work to optimize redundant code away ;)
498
499.. code-block:: c
500
501	int fun(int a)
502	{
503		int result = 0;
504		char *buffer;
505
506		buffer = kmalloc(SIZE, GFP_KERNEL);
507		if (!buffer)
508			return -ENOMEM;
509
510		if (condition1) {
511			while (loop1) {
512				...
513			}
514			result = 1;
515			goto out_free_buffer;
516		}
517		...
518	out_free_buffer:
519		kfree(buffer);
520		return result;
521	}
522
523A common type of bug to be aware of is ``one err bugs`` which look like this:
524
525.. code-block:: c
526
527	err:
528		kfree(foo->bar);
529		kfree(foo);
530		return ret;
531
532The bug in this code is that on some exit paths ``foo`` is NULL.  Normally the
533fix for this is to split it up into two error labels ``err_free_bar:`` and
534``err_free_foo:``:
535
536.. code-block:: c
537
538	 err_free_bar:
539		kfree(foo->bar);
540	 err_free_foo:
541		kfree(foo);
542		return ret;
543
544Ideally you should simulate errors to test all exit paths.
545
546
5478) Commenting
548-------------
549
550Comments are good, but there is also a danger of over-commenting.  NEVER
551try to explain HOW your code works in a comment: it's much better to
552write the code so that the **working** is obvious, and it's a waste of
553time to explain badly written code.
554
555Generally, you want your comments to tell WHAT your code does, not HOW.
556Also, try to avoid putting comments inside a function body: if the
557function is so complex that you need to separately comment parts of it,
558you should probably go back to chapter 6 for a while.  You can make
559small comments to note or warn about something particularly clever (or
560ugly), but try to avoid excess.  Instead, put the comments at the head
561of the function, telling people what it does, and possibly WHY it does
562it.
563
564When commenting the kernel API functions, please use the kernel-doc format.
565See the files at :ref:`Documentation/doc-guide/ <doc_guide>` and
566``scripts/kernel-doc`` for details.
567
568The preferred style for long (multi-line) comments is:
569
570.. code-block:: c
571
572	/*
573	 * This is the preferred style for multi-line
574	 * comments in the Linux kernel source code.
575	 * Please use it consistently.
576	 *
577	 * Description:  A column of asterisks on the left side,
578	 * with beginning and ending almost-blank lines.
579	 */
580
581For files in net/ and drivers/net/ the preferred style for long (multi-line)
582comments is a little different.
583
584.. code-block:: c
585
586	/* The preferred comment style for files in net/ and drivers/net
587	 * looks like this.
588	 *
589	 * It is nearly the same as the generally preferred comment style,
590	 * but there is no initial almost-blank line.
591	 */
592
593It's also important to comment data, whether they are basic types or derived
594types.  To this end, use just one data declaration per line (no commas for
595multiple data declarations).  This leaves you room for a small comment on each
596item, explaining its use.
597
598
5999) You've made a mess of it
600---------------------------
601
602That's OK, we all do.  You've probably been told by your long-time Unix
603user helper that ``GNU emacs`` automatically formats the C sources for
604you, and you've noticed that yes, it does do that, but the defaults it
605uses are less than desirable (in fact, they are worse than random
606typing - an infinite number of monkeys typing into GNU emacs would never
607make a good program).
608
609So, you can either get rid of GNU emacs, or change it to use saner
610values.  To do the latter, you can stick the following in your .emacs file:
611
612.. code-block:: none
613
614  (defun c-lineup-arglist-tabs-only (ignored)
615    "Line up argument lists by tabs, not spaces"
616    (let* ((anchor (c-langelem-pos c-syntactic-element))
617           (column (c-langelem-2nd-pos c-syntactic-element))
618           (offset (- (1+ column) anchor))
619           (steps (floor offset c-basic-offset)))
620      (* (max steps 1)
621         c-basic-offset)))
622
623  (dir-locals-set-class-variables
624   'linux-kernel
625   '((c-mode . (
626          (c-basic-offset . 8)
627          (c-label-minimum-indentation . 0)
628          (c-offsets-alist . (
629                  (arglist-close         . c-lineup-arglist-tabs-only)
630                  (arglist-cont-nonempty .
631		      (c-lineup-gcc-asm-reg c-lineup-arglist-tabs-only))
632                  (arglist-intro         . +)
633                  (brace-list-intro      . +)
634                  (c                     . c-lineup-C-comments)
635                  (case-label            . 0)
636                  (comment-intro         . c-lineup-comment)
637                  (cpp-define-intro      . +)
638                  (cpp-macro             . -1000)
639                  (cpp-macro-cont        . +)
640                  (defun-block-intro     . +)
641                  (else-clause           . 0)
642                  (func-decl-cont        . +)
643                  (inclass               . +)
644                  (inher-cont            . c-lineup-multi-inher)
645                  (knr-argdecl-intro     . 0)
646                  (label                 . -1000)
647                  (statement             . 0)
648                  (statement-block-intro . +)
649                  (statement-case-intro  . +)
650                  (statement-cont        . +)
651                  (substatement          . +)
652                  ))
653          (indent-tabs-mode . t)
654          (show-trailing-whitespace . t)
655          ))))
656
657  (dir-locals-set-directory-class
658   (expand-file-name "~/src/linux-trees")
659   'linux-kernel)
660
661This will make emacs go better with the kernel coding style for C
662files below ``~/src/linux-trees``.
663
664But even if you fail in getting emacs to do sane formatting, not
665everything is lost: use ``indent``.
666
667Now, again, GNU indent has the same brain-dead settings that GNU emacs
668has, which is why you need to give it a few command line options.
669However, that's not too bad, because even the makers of GNU indent
670recognize the authority of K&R (the GNU people aren't evil, they are
671just severely misguided in this matter), so you just give indent the
672options ``-kr -i8`` (stands for ``K&R, 8 character indents``), or use
673``scripts/Lindent``, which indents in the latest style.
674
675``indent`` has a lot of options, and especially when it comes to comment
676re-formatting you may want to take a look at the man page.  But
677remember: ``indent`` is not a fix for bad programming.
678
679Note that you can also use the ``clang-format`` tool to help you with
680these rules, to quickly re-format parts of your code automatically,
681and to review full files in order to spot coding style mistakes,
682typos and possible improvements. It is also handy for sorting ``#includes``,
683for aligning variables/macros, for reflowing text and other similar tasks.
684See the file :ref:`Documentation/process/clang-format.rst <clangformat>`
685for more details.
686
687
68810) Kconfig configuration files
689-------------------------------
690
691For all of the Kconfig* configuration files throughout the source tree,
692the indentation is somewhat different.  Lines under a ``config`` definition
693are indented with one tab, while help text is indented an additional two
694spaces.  Example::
695
696  config AUDIT
697	bool "Auditing support"
698	depends on NET
699	help
700	  Enable auditing infrastructure that can be used with another
701	  kernel subsystem, such as SELinux (which requires this for
702	  logging of avc messages output).  Does not do system-call
703	  auditing without CONFIG_AUDITSYSCALL.
704
705Seriously dangerous features (such as write support for certain
706filesystems) should advertise this prominently in their prompt string::
707
708  config ADFS_FS_RW
709	bool "ADFS write support (DANGEROUS)"
710	depends on ADFS_FS
711	...
712
713For full documentation on the configuration files, see the file
714Documentation/kbuild/kconfig-language.rst.
715
716
71711) Data structures
718-------------------
719
720Data structures that have visibility outside the single-threaded
721environment they are created and destroyed in should always have
722reference counts.  In the kernel, garbage collection doesn't exist (and
723outside the kernel garbage collection is slow and inefficient), which
724means that you absolutely **have** to reference count all your uses.
725
726Reference counting means that you can avoid locking, and allows multiple
727users to have access to the data structure in parallel - and not having
728to worry about the structure suddenly going away from under them just
729because they slept or did something else for a while.
730
731Note that locking is **not** a replacement for reference counting.
732Locking is used to keep data structures coherent, while reference
733counting is a memory management technique.  Usually both are needed, and
734they are not to be confused with each other.
735
736Many data structures can indeed have two levels of reference counting,
737when there are users of different ``classes``.  The subclass count counts
738the number of subclass users, and decrements the global count just once
739when the subclass count goes to zero.
740
741Examples of this kind of ``multi-level-reference-counting`` can be found in
742memory management (``struct mm_struct``: mm_users and mm_count), and in
743filesystem code (``struct super_block``: s_count and s_active).
744
745Remember: if another thread can find your data structure, and you don't
746have a reference count on it, you almost certainly have a bug.
747
748
74912) Macros, Enums and RTL
750-------------------------
751
752Names of macros defining constants and labels in enums are capitalized.
753
754.. code-block:: c
755
756	#define CONSTANT 0x12345
757
758Enums are preferred when defining several related constants.
759
760CAPITALIZED macro names are appreciated but macros resembling functions
761may be named in lower case.
762
763Generally, inline functions are preferable to macros resembling functions.
764
765Macros with multiple statements should be enclosed in a do - while block:
766
767.. code-block:: c
768
769	#define macrofun(a, b, c)			\
770		do {					\
771			if (a == 5)			\
772				do_this(b, c);		\
773		} while (0)
774
775Things to avoid when using macros:
776
7771) macros that affect control flow:
778
779.. code-block:: c
780
781	#define FOO(x)					\
782		do {					\
783			if (blah(x) < 0)		\
784				return -EBUGGERED;	\
785		} while (0)
786
787is a **very** bad idea.  It looks like a function call but exits the ``calling``
788function; don't break the internal parsers of those who will read the code.
789
7902) macros that depend on having a local variable with a magic name:
791
792.. code-block:: c
793
794	#define FOO(val) bar(index, val)
795
796might look like a good thing, but it's confusing as hell when one reads the
797code and it's prone to breakage from seemingly innocent changes.
798
7993) macros with arguments that are used as l-values: FOO(x) = y; will
800bite you if somebody e.g. turns FOO into an inline function.
801
8024) forgetting about precedence: macros defining constants using expressions
803must enclose the expression in parentheses. Beware of similar issues with
804macros using parameters.
805
806.. code-block:: c
807
808	#define CONSTANT 0x4000
809	#define CONSTEXP (CONSTANT | 3)
810
8115) namespace collisions when defining local variables in macros resembling
812functions:
813
814.. code-block:: c
815
816	#define FOO(x)				\
817	({					\
818		typeof(x) ret;			\
819		ret = calc_ret(x);		\
820		(ret);				\
821	})
822
823ret is a common name for a local variable - __foo_ret is less likely
824to collide with an existing variable.
825
826The cpp manual deals with macros exhaustively. The gcc internals manual also
827covers RTL which is used frequently with assembly language in the kernel.
828
829
83013) Printing kernel messages
831----------------------------
832
833Kernel developers like to be seen as literate. Do mind the spelling
834of kernel messages to make a good impression. Do not use incorrect
835contractions like ``dont``; use ``do not`` or ``don't`` instead. Make the
836messages concise, clear, and unambiguous.
837
838Kernel messages do not have to be terminated with a period.
839
840Printing numbers in parentheses (%d) adds no value and should be avoided.
841
842There are a number of driver model diagnostic macros in <linux/device.h>
843which you should use to make sure messages are matched to the right device
844and driver, and are tagged with the right level:  dev_err(), dev_warn(),
845dev_info(), and so forth.  For messages that aren't associated with a
846particular device, <linux/printk.h> defines pr_notice(), pr_info(),
847pr_warn(), pr_err(), etc.
848
849Coming up with good debugging messages can be quite a challenge; and once
850you have them, they can be a huge help for remote troubleshooting.  However
851debug message printing is handled differently than printing other non-debug
852messages.  While the other pr_XXX() functions print unconditionally,
853pr_debug() does not; it is compiled out by default, unless either DEBUG is
854defined or CONFIG_DYNAMIC_DEBUG is set.  That is true for dev_dbg() also,
855and a related convention uses VERBOSE_DEBUG to add dev_vdbg() messages to
856the ones already enabled by DEBUG.
857
858Many subsystems have Kconfig debug options to turn on -DDEBUG in the
859corresponding Makefile; in other cases specific files #define DEBUG.  And
860when a debug message should be unconditionally printed, such as if it is
861already inside a debug-related #ifdef section, printk(KERN_DEBUG ...) can be
862used.
863
864
86514) Allocating memory
866---------------------
867
868The kernel provides the following general purpose memory allocators:
869kmalloc(), kzalloc(), kmalloc_array(), kcalloc(), vmalloc(), and
870vzalloc().  Please refer to the API documentation for further information
871about them.  :ref:`Documentation/core-api/memory-allocation.rst
872<memory_allocation>`
873
874The preferred form for passing a size of a struct is the following:
875
876.. code-block:: c
877
878	p = kmalloc(sizeof(*p), ...);
879
880The alternative form where struct name is spelled out hurts readability and
881introduces an opportunity for a bug when the pointer variable type is changed
882but the corresponding sizeof that is passed to a memory allocator is not.
883
884Casting the return value which is a void pointer is redundant. The conversion
885from void pointer to any other pointer type is guaranteed by the C programming
886language.
887
888The preferred form for allocating an array is the following:
889
890.. code-block:: c
891
892	p = kmalloc_array(n, sizeof(...), ...);
893
894The preferred form for allocating a zeroed array is the following:
895
896.. code-block:: c
897
898	p = kcalloc(n, sizeof(...), ...);
899
900Both forms check for overflow on the allocation size n * sizeof(...),
901and return NULL if that occurred.
902
903These generic allocation functions all emit a stack dump on failure when used
904without __GFP_NOWARN so there is no use in emitting an additional failure
905message when NULL is returned.
906
90715) The inline disease
908----------------------
909
910There appears to be a common misperception that gcc has a magic "make me
911faster" speedup option called ``inline``. While the use of inlines can be
912appropriate (for example as a means of replacing macros, see Chapter 12), it
913very often is not. Abundant use of the inline keyword leads to a much bigger
914kernel, which in turn slows the system as a whole down, due to a bigger
915icache footprint for the CPU and simply because there is less memory
916available for the pagecache. Just think about it; a pagecache miss causes a
917disk seek, which easily takes 5 milliseconds. There are a LOT of cpu cycles
918that can go into these 5 milliseconds.
919
920A reasonable rule of thumb is to not put inline at functions that have more
921than 3 lines of code in them. An exception to this rule are the cases where
922a parameter is known to be a compiletime constant, and as a result of this
923constantness you *know* the compiler will be able to optimize most of your
924function away at compile time. For a good example of this later case, see
925the kmalloc() inline function.
926
927Often people argue that adding inline to functions that are static and used
928only once is always a win since there is no space tradeoff. While this is
929technically correct, gcc is capable of inlining these automatically without
930help, and the maintenance issue of removing the inline when a second user
931appears outweighs the potential value of the hint that tells gcc to do
932something it would have done anyway.
933
934
93516) Function return values and names
936------------------------------------
937
938Functions can return values of many different kinds, and one of the
939most common is a value indicating whether the function succeeded or
940failed.  Such a value can be represented as an error-code integer
941(-Exxx = failure, 0 = success) or a ``succeeded`` boolean (0 = failure,
942non-zero = success).
943
944Mixing up these two sorts of representations is a fertile source of
945difficult-to-find bugs.  If the C language included a strong distinction
946between integers and booleans then the compiler would find these mistakes
947for us... but it doesn't.  To help prevent such bugs, always follow this
948convention::
949
950	If the name of a function is an action or an imperative command,
951	the function should return an error-code integer.  If the name
952	is a predicate, the function should return a "succeeded" boolean.
953
954For example, ``add work`` is a command, and the add_work() function returns 0
955for success or -EBUSY for failure.  In the same way, ``PCI device present`` is
956a predicate, and the pci_dev_present() function returns 1 if it succeeds in
957finding a matching device or 0 if it doesn't.
958
959All EXPORTed functions must respect this convention, and so should all
960public functions.  Private (static) functions need not, but it is
961recommended that they do.
962
963Functions whose return value is the actual result of a computation, rather
964than an indication of whether the computation succeeded, are not subject to
965this rule.  Generally they indicate failure by returning some out-of-range
966result.  Typical examples would be functions that return pointers; they use
967NULL or the ERR_PTR mechanism to report failure.
968
969
97017) Using bool
971--------------
972
973The Linux kernel bool type is an alias for the C99 _Bool type. bool values can
974only evaluate to 0 or 1, and implicit or explicit conversion to bool
975automatically converts the value to true or false. When using bool types the
976!! construction is not needed, which eliminates a class of bugs.
977
978When working with bool values the true and false definitions should be used
979instead of 1 and 0.
980
981bool function return types and stack variables are always fine to use whenever
982appropriate. Use of bool is encouraged to improve readability and is often a
983better option than 'int' for storing boolean values.
984
985Do not use bool if cache line layout or size of the value matters, as its size
986and alignment varies based on the compiled architecture. Structures that are
987optimized for alignment and size should not use bool.
988
989If a structure has many true/false values, consider consolidating them into a
990bitfield with 1 bit members, or using an appropriate fixed width type, such as
991u8.
992
993Similarly for function arguments, many true/false values can be consolidated
994into a single bitwise 'flags' argument and 'flags' can often be a more
995readable alternative if the call-sites have naked true/false constants.
996
997Otherwise limited use of bool in structures and arguments can improve
998readability.
999
100018) Don't re-invent the kernel macros
1001-------------------------------------
1002
1003The header file include/linux/kernel.h contains a number of macros that
1004you should use, rather than explicitly coding some variant of them yourself.
1005For example, if you need to calculate the length of an array, take advantage
1006of the macro
1007
1008.. code-block:: c
1009
1010	#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
1011
1012Similarly, if you need to calculate the size of some structure member, use
1013
1014.. code-block:: c
1015
1016	#define sizeof_field(t, f) (sizeof(((t*)0)->f))
1017
1018There are also min() and max() macros that do strict type checking if you
1019need them.  Feel free to peruse that header file to see what else is already
1020defined that you shouldn't reproduce in your code.
1021
1022
102319) Editor modelines and other cruft
1024------------------------------------
1025
1026Some editors can interpret configuration information embedded in source files,
1027indicated with special markers.  For example, emacs interprets lines marked
1028like this:
1029
1030.. code-block:: c
1031
1032	-*- mode: c -*-
1033
1034Or like this:
1035
1036.. code-block:: c
1037
1038	/*
1039	Local Variables:
1040	compile-command: "gcc -DMAGIC_DEBUG_FLAG foo.c"
1041	End:
1042	*/
1043
1044Vim interprets markers that look like this:
1045
1046.. code-block:: c
1047
1048	/* vim:set sw=8 noet */
1049
1050Do not include any of these in source files.  People have their own personal
1051editor configurations, and your source files should not override them.  This
1052includes markers for indentation and mode configuration.  People may use their
1053own custom mode, or may have some other magic method for making indentation
1054work correctly.
1055
1056
105720) Inline assembly
1058-------------------
1059
1060In architecture-specific code, you may need to use inline assembly to interface
1061with CPU or platform functionality.  Don't hesitate to do so when necessary.
1062However, don't use inline assembly gratuitously when C can do the job.  You can
1063and should poke hardware from C when possible.
1064
1065Consider writing simple helper functions that wrap common bits of inline
1066assembly, rather than repeatedly writing them with slight variations.  Remember
1067that inline assembly can use C parameters.
1068
1069Large, non-trivial assembly functions should go in .S files, with corresponding
1070C prototypes defined in C header files.  The C prototypes for assembly
1071functions should use ``asmlinkage``.
1072
1073You may need to mark your asm statement as volatile, to prevent GCC from
1074removing it if GCC doesn't notice any side effects.  You don't always need to
1075do so, though, and doing so unnecessarily can limit optimization.
1076
1077When writing a single inline assembly statement containing multiple
1078instructions, put each instruction on a separate line in a separate quoted
1079string, and end each string except the last with ``\n\t`` to properly indent
1080the next instruction in the assembly output:
1081
1082.. code-block:: c
1083
1084	asm ("magic %reg1, #42\n\t"
1085	     "more_magic %reg2, %reg3"
1086	     : /* outputs */ : /* inputs */ : /* clobbers */);
1087
1088
108921) Conditional Compilation
1090---------------------------
1091
1092Wherever possible, don't use preprocessor conditionals (#if, #ifdef) in .c
1093files; doing so makes code harder to read and logic harder to follow.  Instead,
1094use such conditionals in a header file defining functions for use in those .c
1095files, providing no-op stub versions in the #else case, and then call those
1096functions unconditionally from .c files.  The compiler will avoid generating
1097any code for the stub calls, producing identical results, but the logic will
1098remain easy to follow.
1099
1100Prefer to compile out entire functions, rather than portions of functions or
1101portions of expressions.  Rather than putting an ifdef in an expression, factor
1102out part or all of the expression into a separate helper function and apply the
1103conditional to that function.
1104
1105If you have a function or variable which may potentially go unused in a
1106particular configuration, and the compiler would warn about its definition
1107going unused, mark the definition as __maybe_unused rather than wrapping it in
1108a preprocessor conditional.  (However, if a function or variable *always* goes
1109unused, delete it.)
1110
1111Within code, where possible, use the IS_ENABLED macro to convert a Kconfig
1112symbol into a C boolean expression, and use it in a normal C conditional:
1113
1114.. code-block:: c
1115
1116	if (IS_ENABLED(CONFIG_SOMETHING)) {
1117		...
1118	}
1119
1120The compiler will constant-fold the conditional away, and include or exclude
1121the block of code just as with an #ifdef, so this will not add any runtime
1122overhead.  However, this approach still allows the C compiler to see the code
1123inside the block, and check it for correctness (syntax, types, symbol
1124references, etc).  Thus, you still have to use an #ifdef if the code inside the
1125block references symbols that will not exist if the condition is not met.
1126
1127At the end of any non-trivial #if or #ifdef block (more than a few lines),
1128place a comment after the #endif on the same line, noting the conditional
1129expression used.  For instance:
1130
1131.. code-block:: c
1132
1133	#ifdef CONFIG_SOMETHING
1134	...
1135	#endif /* CONFIG_SOMETHING */
1136
1137
1138Appendix I) References
1139----------------------
1140
1141The C Programming Language, Second Edition
1142by Brian W. Kernighan and Dennis M. Ritchie.
1143Prentice Hall, Inc., 1988.
1144ISBN 0-13-110362-8 (paperback), 0-13-110370-9 (hardback).
1145
1146The Practice of Programming
1147by Brian W. Kernighan and Rob Pike.
1148Addison-Wesley, Inc., 1999.
1149ISBN 0-201-61586-X.
1150
1151GNU manuals - where in compliance with K&R and this text - for cpp, gcc,
1152gcc internals and indent, all available from https://www.gnu.org/manual/
1153
1154WG14 is the international standardization working group for the programming
1155language C, URL: http://www.open-std.org/JTC1/SC22/WG14/
1156
1157Kernel :ref:`process/coding-style.rst <codingstyle>`, by greg@kroah.com at OLS 2002:
1158http://www.kroah.com/linux/talks/ols_2002_kernel_codingstyle_talk/html/
1159