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 use commas to avoid using braces:
73
74.. code-block:: c
75
76	if (condition)
77		do_this(), do_that();
78
79Always uses braces for multiple statements:
80
81.. code-block:: c
82
83	if (condition) {
84		do_this();
85		do_that();
86	}
87
88Don't put multiple assignments on a single line either.  Kernel coding style
89is super simple.  Avoid tricky expressions.
90
91
92Outside of comments, documentation and except in Kconfig, spaces are never
93used for indentation, and the above example is deliberately broken.
94
95Get a decent editor and don't leave whitespace at the end of lines.
96
97
982) Breaking long lines and strings
99----------------------------------
100
101Coding style is all about readability and maintainability using commonly
102available tools.
103
104The preferred limit on the length of a single line is 80 columns.
105
106Statements longer than 80 columns should be broken into sensible chunks,
107unless exceeding 80 columns significantly increases readability and does
108not hide information.
109
110Descendants are always substantially shorter than the parent and
111are placed substantially to the right.  A very commonly used style
112is to align descendants to a function open parenthesis.
113
114These same rules are applied to function headers with a long argument list.
115
116However, never break user-visible strings such as printk messages because
117that breaks the ability to grep for them.
118
119
1203) Placing Braces and Spaces
121----------------------------
122
123The other issue that always comes up in C styling is the placement of
124braces.  Unlike the indent size, there are few technical reasons to
125choose one placement strategy over the other, but the preferred way, as
126shown to us by the prophets Kernighan and Ritchie, is to put the opening
127brace last on the line, and put the closing brace first, thusly:
128
129.. code-block:: c
130
131	if (x is true) {
132		we do y
133	}
134
135This applies to all non-function statement blocks (if, switch, for,
136while, do).  E.g.:
137
138.. code-block:: c
139
140	switch (action) {
141	case KOBJ_ADD:
142		return "add";
143	case KOBJ_REMOVE:
144		return "remove";
145	case KOBJ_CHANGE:
146		return "change";
147	default:
148		return NULL;
149	}
150
151However, there is one special case, namely functions: they have the
152opening brace at the beginning of the next line, thus:
153
154.. code-block:: c
155
156	int function(int x)
157	{
158		body of function
159	}
160
161Heretic people all over the world have claimed that this inconsistency
162is ...  well ...  inconsistent, but all right-thinking people know that
163(a) K&R are **right** and (b) K&R are right.  Besides, functions are
164special anyway (you can't nest them in C).
165
166Note that the closing brace is empty on a line of its own, **except** in
167the cases where it is followed by a continuation of the same statement,
168ie a ``while`` in a do-statement or an ``else`` in an if-statement, like
169this:
170
171.. code-block:: c
172
173	do {
174		body of do-loop
175	} while (condition);
176
177and
178
179.. code-block:: c
180
181	if (x == y) {
182		..
183	} else if (x > y) {
184		...
185	} else {
186		....
187	}
188
189Rationale: K&R.
190
191Also, note that this brace-placement also minimizes the number of empty
192(or almost empty) lines, without any loss of readability.  Thus, as the
193supply of new-lines on your screen is not a renewable resource (think
19425-line terminal screens here), you have more empty lines to put
195comments on.
196
197Do not unnecessarily use braces where a single statement will do.
198
199.. code-block:: c
200
201	if (condition)
202		action();
203
204and
205
206.. code-block:: none
207
208	if (condition)
209		do_this();
210	else
211		do_that();
212
213This does not apply if only one branch of a conditional statement is a single
214statement; in the latter case use braces in both branches:
215
216.. code-block:: c
217
218	if (condition) {
219		do_this();
220		do_that();
221	} else {
222		otherwise();
223	}
224
225Also, use braces when a loop contains more than a single simple statement:
226
227.. code-block:: c
228
229	while (condition) {
230		if (test)
231			do_something();
232	}
233
2343.1) Spaces
235***********
236
237Linux kernel style for use of spaces depends (mostly) on
238function-versus-keyword usage.  Use a space after (most) keywords.  The
239notable exceptions are sizeof, typeof, alignof, and __attribute__, which look
240somewhat like functions (and are usually used with parentheses in Linux,
241although they are not required in the language, as in: ``sizeof info`` after
242``struct fileinfo info;`` is declared).
243
244So use a space after these keywords::
245
246	if, switch, case, for, do, while
247
248but not with sizeof, typeof, alignof, or __attribute__.  E.g.,
249
250.. code-block:: c
251
252
253	s = sizeof(struct file);
254
255Do not add spaces around (inside) parenthesized expressions.  This example is
256**bad**:
257
258.. code-block:: c
259
260
261	s = sizeof( struct file );
262
263When declaring pointer data or a function that returns a pointer type, the
264preferred use of ``*`` is adjacent to the data name or function name and not
265adjacent to the type name.  Examples:
266
267.. code-block:: c
268
269
270	char *linux_banner;
271	unsigned long long memparse(char *ptr, char **retptr);
272	char *match_strdup(substring_t *s);
273
274Use one space around (on each side of) most binary and ternary operators,
275such as any of these::
276
277	=  +  -  <  >  *  /  %  |  &  ^  <=  >=  ==  !=  ?  :
278
279but no space after unary operators::
280
281	&  *  +  -  ~  !  sizeof  typeof  alignof  __attribute__  defined
282
283no space before the postfix increment & decrement unary operators::
284
285	++  --
286
287no space after the prefix increment & decrement unary operators::
288
289	++  --
290
291and no space around the ``.`` and ``->`` structure member operators.
292
293Do not leave trailing whitespace at the ends of lines.  Some editors with
294``smart`` indentation will insert whitespace at the beginning of new lines as
295appropriate, so you can start typing the next line of code right away.
296However, some such editors do not remove the whitespace if you end up not
297putting a line of code there, such as if you leave a blank line.  As a result,
298you end up with lines containing trailing whitespace.
299
300Git will warn you about patches that introduce trailing whitespace, and can
301optionally strip the trailing whitespace for you; however, if applying a series
302of patches, this may make later patches in the series fail by changing their
303context lines.
304
305
3064) Naming
307---------
308
309C is a Spartan language, and your naming conventions should follow suit.
310Unlike Modula-2 and Pascal programmers, C programmers do not use cute
311names like ThisVariableIsATemporaryCounter. A C programmer would call that
312variable ``tmp``, which is much easier to write, and not the least more
313difficult to understand.
314
315HOWEVER, while mixed-case names are frowned upon, descriptive names for
316global variables are a must.  To call a global function ``foo`` is a
317shooting offense.
318
319GLOBAL variables (to be used only if you **really** need them) need to
320have descriptive names, as do global functions.  If you have a function
321that counts the number of active users, you should call that
322``count_active_users()`` or similar, you should **not** call it ``cntusr()``.
323
324Encoding the type of a function into the name (so-called Hungarian
325notation) is asinine - the compiler knows the types anyway and can check
326those, and it only confuses the programmer.
327
328LOCAL variable names should be short, and to the point.  If you have
329some random integer loop counter, it should probably be called ``i``.
330Calling it ``loop_counter`` is non-productive, if there is no chance of it
331being mis-understood.  Similarly, ``tmp`` can be just about any type of
332variable that is used to hold a temporary value.
333
334If you are afraid to mix up your local variable names, you have another
335problem, which is called the function-growth-hormone-imbalance syndrome.
336See chapter 6 (Functions).
337
338For symbol names and documentation, avoid introducing new usage of
339'master / slave' (or 'slave' independent of 'master') and 'blacklist /
340whitelist'.
341
342Recommended replacements for 'master / slave' are:
343    '{primary,main} / {secondary,replica,subordinate}'
344    '{initiator,requester} / {target,responder}'
345    '{controller,host} / {device,worker,proxy}'
346    'leader / follower'
347    'director / performer'
348
349Recommended replacements for 'blacklist/whitelist' are:
350    'denylist / allowlist'
351    'blocklist / passlist'
352
353Exceptions for introducing new usage is to maintain a userspace ABI/API,
354or when updating code for an existing (as of 2020) hardware or protocol
355specification that mandates those terms. For new specifications
356translate specification usage of the terminology to the kernel coding
357standard where possible.
358
3595) Typedefs
360-----------
361
362Please don't use things like ``vps_t``.
363It's a **mistake** to use typedef for structures and pointers. When you see a
364
365.. code-block:: c
366
367
368	vps_t a;
369
370in the source, what does it mean?
371In contrast, if it says
372
373.. code-block:: c
374
375	struct virtual_container *a;
376
377you can actually tell what ``a`` is.
378
379Lots of people think that typedefs ``help readability``. Not so. They are
380useful only for:
381
382 (a) totally opaque objects (where the typedef is actively used to **hide**
383     what the object is).
384
385     Example: ``pte_t`` etc. opaque objects that you can only access using
386     the proper accessor functions.
387
388     .. note::
389
390       Opaqueness and ``accessor functions`` are not good in themselves.
391       The reason we have them for things like pte_t etc. is that there
392       really is absolutely **zero** portably accessible information there.
393
394 (b) Clear integer types, where the abstraction **helps** avoid confusion
395     whether it is ``int`` or ``long``.
396
397     u8/u16/u32 are perfectly fine typedefs, although they fit into
398     category (d) better than here.
399
400     .. note::
401
402       Again - there needs to be a **reason** for this. If something is
403       ``unsigned long``, then there's no reason to do
404
405	typedef unsigned long myflags_t;
406
407     but if there is a clear reason for why it under certain circumstances
408     might be an ``unsigned int`` and under other configurations might be
409     ``unsigned long``, then by all means go ahead and use a typedef.
410
411 (c) when you use sparse to literally create a **new** type for
412     type-checking.
413
414 (d) New types which are identical to standard C99 types, in certain
415     exceptional circumstances.
416
417     Although it would only take a short amount of time for the eyes and
418     brain to become accustomed to the standard types like ``uint32_t``,
419     some people object to their use anyway.
420
421     Therefore, the Linux-specific ``u8/u16/u32/u64`` types and their
422     signed equivalents which are identical to standard types are
423     permitted -- although they are not mandatory in new code of your
424     own.
425
426     When editing existing code which already uses one or the other set
427     of types, you should conform to the existing choices in that code.
428
429 (e) Types safe for use in userspace.
430
431     In certain structures which are visible to userspace, we cannot
432     require C99 types and cannot use the ``u32`` form above. Thus, we
433     use __u32 and similar types in all structures which are shared
434     with userspace.
435
436Maybe there are other cases too, but the rule should basically be to NEVER
437EVER use a typedef unless you can clearly match one of those rules.
438
439In general, a pointer, or a struct that has elements that can reasonably
440be directly accessed should **never** be a typedef.
441
442
4436) Functions
444------------
445
446Functions should be short and sweet, and do just one thing.  They should
447fit on one or two screenfuls of text (the ISO/ANSI screen size is 80x24,
448as we all know), and do one thing and do that well.
449
450The maximum length of a function is inversely proportional to the
451complexity and indentation level of that function.  So, if you have a
452conceptually simple function that is just one long (but simple)
453case-statement, where you have to do lots of small things for a lot of
454different cases, it's OK to have a longer function.
455
456However, if you have a complex function, and you suspect that a
457less-than-gifted first-year high-school student might not even
458understand what the function is all about, you should adhere to the
459maximum limits all the more closely.  Use helper functions with
460descriptive names (you can ask the compiler to in-line them if you think
461it's performance-critical, and it will probably do a better job of it
462than you would have done).
463
464Another measure of the function is the number of local variables.  They
465shouldn't exceed 5-10, or you're doing something wrong.  Re-think the
466function, and split it into smaller pieces.  A human brain can
467generally easily keep track of about 7 different things, anything more
468and it gets confused.  You know you're brilliant, but maybe you'd like
469to understand what you did 2 weeks from now.
470
471In source files, separate functions with one blank line.  If the function is
472exported, the **EXPORT** macro for it should follow immediately after the
473closing function brace line.  E.g.:
474
475.. code-block:: c
476
477	int system_is_up(void)
478	{
479		return system_state == SYSTEM_RUNNING;
480	}
481	EXPORT_SYMBOL(system_is_up);
482
4836.1) Function prototypes
484************************
485
486In function prototypes, include parameter names with their data types.
487Although this is not required by the C language, it is preferred in Linux
488because it is a simple way to add valuable information for the reader.
489
490Do not use the ``extern`` keyword with function declarations as this makes
491lines longer and isn't strictly necessary.
492
493When writing function prototypes, please keep the `order of elements regular
494<https://lore.kernel.org/mm-commits/CAHk-=wiOCLRny5aifWNhr621kYrJwhfURsa0vFPeUEm8mF0ufg@mail.gmail.com/>`_.
495For example, using this function declaration example::
496
497 __init void * __must_check action(enum magic value, size_t size, u8 count,
498				   char *fmt, ...) __printf(4, 5) __malloc;
499
500The preferred order of elements for a function prototype is:
501
502- storage class (below, ``static __always_inline``, noting that ``__always_inline``
503  is technically an attribute but is treated like ``inline``)
504- storage class attributes (here, ``__init`` -- i.e. section declarations, but also
505  things like ``__cold``)
506- return type (here, ``void *``)
507- return type attributes (here, ``__must_check``)
508- function name (here, ``action``)
509- function parameters (here, ``(enum magic value, size_t size, u8 count, char *fmt, ...)``,
510  noting that parameter names should always be included)
511- function parameter attributes (here, ``__printf(4, 5)``)
512- function behavior attributes (here, ``__malloc``)
513
514Note that for a function **definition** (i.e. the actual function body),
515the compiler does not allow function parameter attributes after the
516function parameters. In these cases, they should go after the storage
517class attributes (e.g. note the changed position of ``__printf(4, 5)``
518below, compared to the **declaration** example above)::
519
520 static __always_inline __init __printf(4, 5) void * __must_check action(enum magic value,
521		size_t size, u8 count, char *fmt, ...) __malloc
522 {
523	...
524 }
525
5267) Centralized exiting of functions
527-----------------------------------
528
529Albeit deprecated by some people, the equivalent of the goto statement is
530used frequently by compilers in form of the unconditional jump instruction.
531
532The goto statement comes in handy when a function exits from multiple
533locations and some common work such as cleanup has to be done.  If there is no
534cleanup needed then just return directly.
535
536Choose label names which say what the goto does or why the goto exists.  An
537example of a good name could be ``out_free_buffer:`` if the goto frees ``buffer``.
538Avoid using GW-BASIC names like ``err1:`` and ``err2:``, as you would have to
539renumber them if you ever add or remove exit paths, and they make correctness
540difficult to verify anyway.
541
542The rationale for using gotos is:
543
544- unconditional statements are easier to understand and follow
545- nesting is reduced
546- errors by not updating individual exit points when making
547  modifications are prevented
548- saves the compiler work to optimize redundant code away ;)
549
550.. code-block:: c
551
552	int fun(int a)
553	{
554		int result = 0;
555		char *buffer;
556
557		buffer = kmalloc(SIZE, GFP_KERNEL);
558		if (!buffer)
559			return -ENOMEM;
560
561		if (condition1) {
562			while (loop1) {
563				...
564			}
565			result = 1;
566			goto out_free_buffer;
567		}
568		...
569	out_free_buffer:
570		kfree(buffer);
571		return result;
572	}
573
574A common type of bug to be aware of is ``one err bugs`` which look like this:
575
576.. code-block:: c
577
578	err:
579		kfree(foo->bar);
580		kfree(foo);
581		return ret;
582
583The bug in this code is that on some exit paths ``foo`` is NULL.  Normally the
584fix for this is to split it up into two error labels ``err_free_bar:`` and
585``err_free_foo:``:
586
587.. code-block:: c
588
589	 err_free_bar:
590		kfree(foo->bar);
591	 err_free_foo:
592		kfree(foo);
593		return ret;
594
595Ideally you should simulate errors to test all exit paths.
596
597
5988) Commenting
599-------------
600
601Comments are good, but there is also a danger of over-commenting.  NEVER
602try to explain HOW your code works in a comment: it's much better to
603write the code so that the **working** is obvious, and it's a waste of
604time to explain badly written code.
605
606Generally, you want your comments to tell WHAT your code does, not HOW.
607Also, try to avoid putting comments inside a function body: if the
608function is so complex that you need to separately comment parts of it,
609you should probably go back to chapter 6 for a while.  You can make
610small comments to note or warn about something particularly clever (or
611ugly), but try to avoid excess.  Instead, put the comments at the head
612of the function, telling people what it does, and possibly WHY it does
613it.
614
615When commenting the kernel API functions, please use the kernel-doc format.
616See the files at :ref:`Documentation/doc-guide/ <doc_guide>` and
617``scripts/kernel-doc`` for details.
618
619The preferred style for long (multi-line) comments is:
620
621.. code-block:: c
622
623	/*
624	 * This is the preferred style for multi-line
625	 * comments in the Linux kernel source code.
626	 * Please use it consistently.
627	 *
628	 * Description:  A column of asterisks on the left side,
629	 * with beginning and ending almost-blank lines.
630	 */
631
632For files in net/ and drivers/net/ the preferred style for long (multi-line)
633comments is a little different.
634
635.. code-block:: c
636
637	/* The preferred comment style for files in net/ and drivers/net
638	 * looks like this.
639	 *
640	 * It is nearly the same as the generally preferred comment style,
641	 * but there is no initial almost-blank line.
642	 */
643
644It's also important to comment data, whether they are basic types or derived
645types.  To this end, use just one data declaration per line (no commas for
646multiple data declarations).  This leaves you room for a small comment on each
647item, explaining its use.
648
649
6509) You've made a mess of it
651---------------------------
652
653That's OK, we all do.  You've probably been told by your long-time Unix
654user helper that ``GNU emacs`` automatically formats the C sources for
655you, and you've noticed that yes, it does do that, but the defaults it
656uses are less than desirable (in fact, they are worse than random
657typing - an infinite number of monkeys typing into GNU emacs would never
658make a good program).
659
660So, you can either get rid of GNU emacs, or change it to use saner
661values.  To do the latter, you can stick the following in your .emacs file:
662
663.. code-block:: none
664
665  (defun c-lineup-arglist-tabs-only (ignored)
666    "Line up argument lists by tabs, not spaces"
667    (let* ((anchor (c-langelem-pos c-syntactic-element))
668           (column (c-langelem-2nd-pos c-syntactic-element))
669           (offset (- (1+ column) anchor))
670           (steps (floor offset c-basic-offset)))
671      (* (max steps 1)
672         c-basic-offset)))
673
674  (dir-locals-set-class-variables
675   'linux-kernel
676   '((c-mode . (
677          (c-basic-offset . 8)
678          (c-label-minimum-indentation . 0)
679          (c-offsets-alist . (
680                  (arglist-close         . c-lineup-arglist-tabs-only)
681                  (arglist-cont-nonempty .
682		      (c-lineup-gcc-asm-reg c-lineup-arglist-tabs-only))
683                  (arglist-intro         . +)
684                  (brace-list-intro      . +)
685                  (c                     . c-lineup-C-comments)
686                  (case-label            . 0)
687                  (comment-intro         . c-lineup-comment)
688                  (cpp-define-intro      . +)
689                  (cpp-macro             . -1000)
690                  (cpp-macro-cont        . +)
691                  (defun-block-intro     . +)
692                  (else-clause           . 0)
693                  (func-decl-cont        . +)
694                  (inclass               . +)
695                  (inher-cont            . c-lineup-multi-inher)
696                  (knr-argdecl-intro     . 0)
697                  (label                 . -1000)
698                  (statement             . 0)
699                  (statement-block-intro . +)
700                  (statement-case-intro  . +)
701                  (statement-cont        . +)
702                  (substatement          . +)
703                  ))
704          (indent-tabs-mode . t)
705          (show-trailing-whitespace . t)
706          ))))
707
708  (dir-locals-set-directory-class
709   (expand-file-name "~/src/linux-trees")
710   'linux-kernel)
711
712This will make emacs go better with the kernel coding style for C
713files below ``~/src/linux-trees``.
714
715But even if you fail in getting emacs to do sane formatting, not
716everything is lost: use ``indent``.
717
718Now, again, GNU indent has the same brain-dead settings that GNU emacs
719has, which is why you need to give it a few command line options.
720However, that's not too bad, because even the makers of GNU indent
721recognize the authority of K&R (the GNU people aren't evil, they are
722just severely misguided in this matter), so you just give indent the
723options ``-kr -i8`` (stands for ``K&R, 8 character indents``), or use
724``scripts/Lindent``, which indents in the latest style.
725
726``indent`` has a lot of options, and especially when it comes to comment
727re-formatting you may want to take a look at the man page.  But
728remember: ``indent`` is not a fix for bad programming.
729
730Note that you can also use the ``clang-format`` tool to help you with
731these rules, to quickly re-format parts of your code automatically,
732and to review full files in order to spot coding style mistakes,
733typos and possible improvements. It is also handy for sorting ``#includes``,
734for aligning variables/macros, for reflowing text and other similar tasks.
735See the file :ref:`Documentation/process/clang-format.rst <clangformat>`
736for more details.
737
738
73910) Kconfig configuration files
740-------------------------------
741
742For all of the Kconfig* configuration files throughout the source tree,
743the indentation is somewhat different.  Lines under a ``config`` definition
744are indented with one tab, while help text is indented an additional two
745spaces.  Example::
746
747  config AUDIT
748	bool "Auditing support"
749	depends on NET
750	help
751	  Enable auditing infrastructure that can be used with another
752	  kernel subsystem, such as SELinux (which requires this for
753	  logging of avc messages output).  Does not do system-call
754	  auditing without CONFIG_AUDITSYSCALL.
755
756Seriously dangerous features (such as write support for certain
757filesystems) should advertise this prominently in their prompt string::
758
759  config ADFS_FS_RW
760	bool "ADFS write support (DANGEROUS)"
761	depends on ADFS_FS
762	...
763
764For full documentation on the configuration files, see the file
765Documentation/kbuild/kconfig-language.rst.
766
767
76811) Data structures
769-------------------
770
771Data structures that have visibility outside the single-threaded
772environment they are created and destroyed in should always have
773reference counts.  In the kernel, garbage collection doesn't exist (and
774outside the kernel garbage collection is slow and inefficient), which
775means that you absolutely **have** to reference count all your uses.
776
777Reference counting means that you can avoid locking, and allows multiple
778users to have access to the data structure in parallel - and not having
779to worry about the structure suddenly going away from under them just
780because they slept or did something else for a while.
781
782Note that locking is **not** a replacement for reference counting.
783Locking is used to keep data structures coherent, while reference
784counting is a memory management technique.  Usually both are needed, and
785they are not to be confused with each other.
786
787Many data structures can indeed have two levels of reference counting,
788when there are users of different ``classes``.  The subclass count counts
789the number of subclass users, and decrements the global count just once
790when the subclass count goes to zero.
791
792Examples of this kind of ``multi-level-reference-counting`` can be found in
793memory management (``struct mm_struct``: mm_users and mm_count), and in
794filesystem code (``struct super_block``: s_count and s_active).
795
796Remember: if another thread can find your data structure, and you don't
797have a reference count on it, you almost certainly have a bug.
798
799
80012) Macros, Enums and RTL
801-------------------------
802
803Names of macros defining constants and labels in enums are capitalized.
804
805.. code-block:: c
806
807	#define CONSTANT 0x12345
808
809Enums are preferred when defining several related constants.
810
811CAPITALIZED macro names are appreciated but macros resembling functions
812may be named in lower case.
813
814Generally, inline functions are preferable to macros resembling functions.
815
816Macros with multiple statements should be enclosed in a do - while block:
817
818.. code-block:: c
819
820	#define macrofun(a, b, c)			\
821		do {					\
822			if (a == 5)			\
823				do_this(b, c);		\
824		} while (0)
825
826Things to avoid when using macros:
827
8281) macros that affect control flow:
829
830.. code-block:: c
831
832	#define FOO(x)					\
833		do {					\
834			if (blah(x) < 0)		\
835				return -EBUGGERED;	\
836		} while (0)
837
838is a **very** bad idea.  It looks like a function call but exits the ``calling``
839function; don't break the internal parsers of those who will read the code.
840
8412) macros that depend on having a local variable with a magic name:
842
843.. code-block:: c
844
845	#define FOO(val) bar(index, val)
846
847might look like a good thing, but it's confusing as hell when one reads the
848code and it's prone to breakage from seemingly innocent changes.
849
8503) macros with arguments that are used as l-values: FOO(x) = y; will
851bite you if somebody e.g. turns FOO into an inline function.
852
8534) forgetting about precedence: macros defining constants using expressions
854must enclose the expression in parentheses. Beware of similar issues with
855macros using parameters.
856
857.. code-block:: c
858
859	#define CONSTANT 0x4000
860	#define CONSTEXP (CONSTANT | 3)
861
8625) namespace collisions when defining local variables in macros resembling
863functions:
864
865.. code-block:: c
866
867	#define FOO(x)				\
868	({					\
869		typeof(x) ret;			\
870		ret = calc_ret(x);		\
871		(ret);				\
872	})
873
874ret is a common name for a local variable - __foo_ret is less likely
875to collide with an existing variable.
876
877The cpp manual deals with macros exhaustively. The gcc internals manual also
878covers RTL which is used frequently with assembly language in the kernel.
879
880
88113) Printing kernel messages
882----------------------------
883
884Kernel developers like to be seen as literate. Do mind the spelling
885of kernel messages to make a good impression. Do not use incorrect
886contractions like ``dont``; use ``do not`` or ``don't`` instead. Make the
887messages concise, clear, and unambiguous.
888
889Kernel messages do not have to be terminated with a period.
890
891Printing numbers in parentheses (%d) adds no value and should be avoided.
892
893There are a number of driver model diagnostic macros in <linux/dev_printk.h>
894which you should use to make sure messages are matched to the right device
895and driver, and are tagged with the right level:  dev_err(), dev_warn(),
896dev_info(), and so forth.  For messages that aren't associated with a
897particular device, <linux/printk.h> defines pr_notice(), pr_info(),
898pr_warn(), pr_err(), etc.
899
900Coming up with good debugging messages can be quite a challenge; and once
901you have them, they can be a huge help for remote troubleshooting.  However
902debug message printing is handled differently than printing other non-debug
903messages.  While the other pr_XXX() functions print unconditionally,
904pr_debug() does not; it is compiled out by default, unless either DEBUG is
905defined or CONFIG_DYNAMIC_DEBUG is set.  That is true for dev_dbg() also,
906and a related convention uses VERBOSE_DEBUG to add dev_vdbg() messages to
907the ones already enabled by DEBUG.
908
909Many subsystems have Kconfig debug options to turn on -DDEBUG in the
910corresponding Makefile; in other cases specific files #define DEBUG.  And
911when a debug message should be unconditionally printed, such as if it is
912already inside a debug-related #ifdef section, printk(KERN_DEBUG ...) can be
913used.
914
915
91614) Allocating memory
917---------------------
918
919The kernel provides the following general purpose memory allocators:
920kmalloc(), kzalloc(), kmalloc_array(), kcalloc(), vmalloc(), and
921vzalloc().  Please refer to the API documentation for further information
922about them.  :ref:`Documentation/core-api/memory-allocation.rst
923<memory_allocation>`
924
925The preferred form for passing a size of a struct is the following:
926
927.. code-block:: c
928
929	p = kmalloc(sizeof(*p), ...);
930
931The alternative form where struct name is spelled out hurts readability and
932introduces an opportunity for a bug when the pointer variable type is changed
933but the corresponding sizeof that is passed to a memory allocator is not.
934
935Casting the return value which is a void pointer is redundant. The conversion
936from void pointer to any other pointer type is guaranteed by the C programming
937language.
938
939The preferred form for allocating an array is the following:
940
941.. code-block:: c
942
943	p = kmalloc_array(n, sizeof(...), ...);
944
945The preferred form for allocating a zeroed array is the following:
946
947.. code-block:: c
948
949	p = kcalloc(n, sizeof(...), ...);
950
951Both forms check for overflow on the allocation size n * sizeof(...),
952and return NULL if that occurred.
953
954These generic allocation functions all emit a stack dump on failure when used
955without __GFP_NOWARN so there is no use in emitting an additional failure
956message when NULL is returned.
957
95815) The inline disease
959----------------------
960
961There appears to be a common misperception that gcc has a magic "make me
962faster" speedup option called ``inline``. While the use of inlines can be
963appropriate (for example as a means of replacing macros, see Chapter 12), it
964very often is not. Abundant use of the inline keyword leads to a much bigger
965kernel, which in turn slows the system as a whole down, due to a bigger
966icache footprint for the CPU and simply because there is less memory
967available for the pagecache. Just think about it; a pagecache miss causes a
968disk seek, which easily takes 5 milliseconds. There are a LOT of cpu cycles
969that can go into these 5 milliseconds.
970
971A reasonable rule of thumb is to not put inline at functions that have more
972than 3 lines of code in them. An exception to this rule are the cases where
973a parameter is known to be a compiletime constant, and as a result of this
974constantness you *know* the compiler will be able to optimize most of your
975function away at compile time. For a good example of this later case, see
976the kmalloc() inline function.
977
978Often people argue that adding inline to functions that are static and used
979only once is always a win since there is no space tradeoff. While this is
980technically correct, gcc is capable of inlining these automatically without
981help, and the maintenance issue of removing the inline when a second user
982appears outweighs the potential value of the hint that tells gcc to do
983something it would have done anyway.
984
985
98616) Function return values and names
987------------------------------------
988
989Functions can return values of many different kinds, and one of the
990most common is a value indicating whether the function succeeded or
991failed.  Such a value can be represented as an error-code integer
992(-Exxx = failure, 0 = success) or a ``succeeded`` boolean (0 = failure,
993non-zero = success).
994
995Mixing up these two sorts of representations is a fertile source of
996difficult-to-find bugs.  If the C language included a strong distinction
997between integers and booleans then the compiler would find these mistakes
998for us... but it doesn't.  To help prevent such bugs, always follow this
999convention::
1000
1001	If the name of a function is an action or an imperative command,
1002	the function should return an error-code integer.  If the name
1003	is a predicate, the function should return a "succeeded" boolean.
1004
1005For example, ``add work`` is a command, and the add_work() function returns 0
1006for success or -EBUSY for failure.  In the same way, ``PCI device present`` is
1007a predicate, and the pci_dev_present() function returns 1 if it succeeds in
1008finding a matching device or 0 if it doesn't.
1009
1010All EXPORTed functions must respect this convention, and so should all
1011public functions.  Private (static) functions need not, but it is
1012recommended that they do.
1013
1014Functions whose return value is the actual result of a computation, rather
1015than an indication of whether the computation succeeded, are not subject to
1016this rule.  Generally they indicate failure by returning some out-of-range
1017result.  Typical examples would be functions that return pointers; they use
1018NULL or the ERR_PTR mechanism to report failure.
1019
1020
102117) Using bool
1022--------------
1023
1024The Linux kernel bool type is an alias for the C99 _Bool type. bool values can
1025only evaluate to 0 or 1, and implicit or explicit conversion to bool
1026automatically converts the value to true or false. When using bool types the
1027!! construction is not needed, which eliminates a class of bugs.
1028
1029When working with bool values the true and false definitions should be used
1030instead of 1 and 0.
1031
1032bool function return types and stack variables are always fine to use whenever
1033appropriate. Use of bool is encouraged to improve readability and is often a
1034better option than 'int' for storing boolean values.
1035
1036Do not use bool if cache line layout or size of the value matters, as its size
1037and alignment varies based on the compiled architecture. Structures that are
1038optimized for alignment and size should not use bool.
1039
1040If a structure has many true/false values, consider consolidating them into a
1041bitfield with 1 bit members, or using an appropriate fixed width type, such as
1042u8.
1043
1044Similarly for function arguments, many true/false values can be consolidated
1045into a single bitwise 'flags' argument and 'flags' can often be a more
1046readable alternative if the call-sites have naked true/false constants.
1047
1048Otherwise limited use of bool in structures and arguments can improve
1049readability.
1050
105118) Don't re-invent the kernel macros
1052-------------------------------------
1053
1054The header file include/linux/kernel.h contains a number of macros that
1055you should use, rather than explicitly coding some variant of them yourself.
1056For example, if you need to calculate the length of an array, take advantage
1057of the macro
1058
1059.. code-block:: c
1060
1061	#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
1062
1063Similarly, if you need to calculate the size of some structure member, use
1064
1065.. code-block:: c
1066
1067	#define sizeof_field(t, f) (sizeof(((t*)0)->f))
1068
1069There are also min() and max() macros that do strict type checking if you
1070need them.  Feel free to peruse that header file to see what else is already
1071defined that you shouldn't reproduce in your code.
1072
1073
107419) Editor modelines and other cruft
1075------------------------------------
1076
1077Some editors can interpret configuration information embedded in source files,
1078indicated with special markers.  For example, emacs interprets lines marked
1079like this:
1080
1081.. code-block:: c
1082
1083	-*- mode: c -*-
1084
1085Or like this:
1086
1087.. code-block:: c
1088
1089	/*
1090	Local Variables:
1091	compile-command: "gcc -DMAGIC_DEBUG_FLAG foo.c"
1092	End:
1093	*/
1094
1095Vim interprets markers that look like this:
1096
1097.. code-block:: c
1098
1099	/* vim:set sw=8 noet */
1100
1101Do not include any of these in source files.  People have their own personal
1102editor configurations, and your source files should not override them.  This
1103includes markers for indentation and mode configuration.  People may use their
1104own custom mode, or may have some other magic method for making indentation
1105work correctly.
1106
1107
110820) Inline assembly
1109-------------------
1110
1111In architecture-specific code, you may need to use inline assembly to interface
1112with CPU or platform functionality.  Don't hesitate to do so when necessary.
1113However, don't use inline assembly gratuitously when C can do the job.  You can
1114and should poke hardware from C when possible.
1115
1116Consider writing simple helper functions that wrap common bits of inline
1117assembly, rather than repeatedly writing them with slight variations.  Remember
1118that inline assembly can use C parameters.
1119
1120Large, non-trivial assembly functions should go in .S files, with corresponding
1121C prototypes defined in C header files.  The C prototypes for assembly
1122functions should use ``asmlinkage``.
1123
1124You may need to mark your asm statement as volatile, to prevent GCC from
1125removing it if GCC doesn't notice any side effects.  You don't always need to
1126do so, though, and doing so unnecessarily can limit optimization.
1127
1128When writing a single inline assembly statement containing multiple
1129instructions, put each instruction on a separate line in a separate quoted
1130string, and end each string except the last with ``\n\t`` to properly indent
1131the next instruction in the assembly output:
1132
1133.. code-block:: c
1134
1135	asm ("magic %reg1, #42\n\t"
1136	     "more_magic %reg2, %reg3"
1137	     : /* outputs */ : /* inputs */ : /* clobbers */);
1138
1139
114021) Conditional Compilation
1141---------------------------
1142
1143Wherever possible, don't use preprocessor conditionals (#if, #ifdef) in .c
1144files; doing so makes code harder to read and logic harder to follow.  Instead,
1145use such conditionals in a header file defining functions for use in those .c
1146files, providing no-op stub versions in the #else case, and then call those
1147functions unconditionally from .c files.  The compiler will avoid generating
1148any code for the stub calls, producing identical results, but the logic will
1149remain easy to follow.
1150
1151Prefer to compile out entire functions, rather than portions of functions or
1152portions of expressions.  Rather than putting an ifdef in an expression, factor
1153out part or all of the expression into a separate helper function and apply the
1154conditional to that function.
1155
1156If you have a function or variable which may potentially go unused in a
1157particular configuration, and the compiler would warn about its definition
1158going unused, mark the definition as __maybe_unused rather than wrapping it in
1159a preprocessor conditional.  (However, if a function or variable *always* goes
1160unused, delete it.)
1161
1162Within code, where possible, use the IS_ENABLED macro to convert a Kconfig
1163symbol into a C boolean expression, and use it in a normal C conditional:
1164
1165.. code-block:: c
1166
1167	if (IS_ENABLED(CONFIG_SOMETHING)) {
1168		...
1169	}
1170
1171The compiler will constant-fold the conditional away, and include or exclude
1172the block of code just as with an #ifdef, so this will not add any runtime
1173overhead.  However, this approach still allows the C compiler to see the code
1174inside the block, and check it for correctness (syntax, types, symbol
1175references, etc).  Thus, you still have to use an #ifdef if the code inside the
1176block references symbols that will not exist if the condition is not met.
1177
1178At the end of any non-trivial #if or #ifdef block (more than a few lines),
1179place a comment after the #endif on the same line, noting the conditional
1180expression used.  For instance:
1181
1182.. code-block:: c
1183
1184	#ifdef CONFIG_SOMETHING
1185	...
1186	#endif /* CONFIG_SOMETHING */
1187
1188
118922) Do not crash the kernel
1190---------------------------
1191
1192In general, the decision to crash the kernel belongs to the user, rather
1193than to the kernel developer.
1194
1195Avoid panic()
1196*************
1197
1198panic() should be used with care and primarily only during system boot.
1199panic() is, for example, acceptable when running out of memory during boot and
1200not being able to continue.
1201
1202Use WARN() rather than BUG()
1203****************************
1204
1205Do not add new code that uses any of the BUG() variants, such as BUG(),
1206BUG_ON(), or VM_BUG_ON(). Instead, use a WARN*() variant, preferably
1207WARN_ON_ONCE(), and possibly with recovery code. Recovery code is not
1208required if there is no reasonable way to at least partially recover.
1209
1210"I'm too lazy to do error handling" is not an excuse for using BUG(). Major
1211internal corruptions with no way of continuing may still use BUG(), but need
1212good justification.
1213
1214Use WARN_ON_ONCE() rather than WARN() or WARN_ON()
1215**************************************************
1216
1217WARN_ON_ONCE() is generally preferred over WARN() or WARN_ON(), because it
1218is common for a given warning condition, if it occurs at all, to occur
1219multiple times. This can fill up and wrap the kernel log, and can even slow
1220the system enough that the excessive logging turns into its own, additional
1221problem.
1222
1223Do not WARN lightly
1224*******************
1225
1226WARN*() is intended for unexpected, this-should-never-happen situations.
1227WARN*() macros are not to be used for anything that is expected to happen
1228during normal operation. These are not pre- or post-condition asserts, for
1229example. Again: WARN*() must not be used for a condition that is expected
1230to trigger easily, for example, by user space actions. pr_warn_once() is a
1231possible alternative, if you need to notify the user of a problem.
1232
1233Do not worry about panic_on_warn users
1234**************************************
1235
1236A few more words about panic_on_warn: Remember that ``panic_on_warn`` is an
1237available kernel option, and that many users set this option. This is why
1238there is a "Do not WARN lightly" writeup, above. However, the existence of
1239panic_on_warn users is not a valid reason to avoid the judicious use
1240WARN*(). That is because, whoever enables panic_on_warn has explicitly
1241asked the kernel to crash if a WARN*() fires, and such users must be
1242prepared to deal with the consequences of a system that is somewhat more
1243likely to crash.
1244
1245Use BUILD_BUG_ON() for compile-time assertions
1246**********************************************
1247
1248The use of BUILD_BUG_ON() is acceptable and encouraged, because it is a
1249compile-time assertion that has no effect at runtime.
1250
1251Appendix I) References
1252----------------------
1253
1254The C Programming Language, Second Edition
1255by Brian W. Kernighan and Dennis M. Ritchie.
1256Prentice Hall, Inc., 1988.
1257ISBN 0-13-110362-8 (paperback), 0-13-110370-9 (hardback).
1258
1259The Practice of Programming
1260by Brian W. Kernighan and Rob Pike.
1261Addison-Wesley, Inc., 1999.
1262ISBN 0-201-61586-X.
1263
1264GNU manuals - where in compliance with K&R and this text - for cpp, gcc,
1265gcc internals and indent, all available from https://www.gnu.org/manual/
1266
1267WG14 is the international standardization working group for the programming
1268language C, URL: http://www.open-std.org/JTC1/SC22/WG14/
1269
1270Kernel CodingStyle, by greg@kroah.com at OLS 2002:
1271http://www.kroah.com/linux/talks/ols_2002_kernel_codingstyle_talk/html/
1272