2 Linux kernel coding style
4 This is a short document describing the preferred coding style for the
5 linux kernel. Coding style is very personal, and I won't _force_ my
6 views on anybody, but this is what goes for anything that I have to be
7 able to maintain, and I'd prefer it for most other things too. Please
8 at least consider the points made here.
10 First off, I'd suggest printing out a copy of the GNU coding standards,
11 and NOT read it. Burn them, it's a great symbolic gesture.
16 Chapter 1: Indentation
18 Tabs are 8 characters, and thus indentations are also 8 characters.
19 There are heretic movements that try to make indentations 4 (or even 2!)
20 characters deep, and that is akin to trying to define the value of PI to
23 Rationale: The whole idea behind indentation is to clearly define where
24 a block of control starts and ends. Especially when you've been looking
25 at your screen for 20 straight hours, you'll find it a lot easier to see
26 how the indentation works if you have large indentations.
28 Now, some people will claim that having 8-character indentations makes
29 the code move too far to the right, and makes it hard to read on a
30 80-character terminal screen. The answer to that is that if you need
31 more than 3 levels of indentation, you're screwed anyway, and should fix
34 In short, 8-char indents make things easier to read, and have the added
35 benefit of warning you when you're nesting your functions too deep.
38 The preferred way to ease multiple indentation levels in a switch statement is
39 to align the "switch" and its subordinate "case" labels in the same column
40 instead of "double-indenting" the "case" labels. E.g.:
59 Don't put multiple statements on a single line unless you have
62 if (condition) do_this;
63 do_something_everytime;
65 Don't put multiple assignments on a single line either. Kernel coding style
66 is super simple. Avoid tricky expressions.
68 Outside of comments, documentation and except in Kconfig, spaces are never
69 used for indentation, and the above example is deliberately broken.
71 Get a decent editor and don't leave whitespace at the end of lines.
74 Chapter 2: Breaking long lines and strings
76 Coding style is all about readability and maintainability using commonly
79 The limit on the length of lines is 80 columns and this is a strongly
82 Statements longer than 80 columns will be broken into sensible chunks, unless
83 exceeding 80 columns significantly increases readability and does not hide
84 information. Descendants are always substantially shorter than the parent and
85 are placed substantially to the right. The same applies to function headers
86 with a long argument list. However, never break user-visible strings such as
87 printk messages, because that breaks the ability to grep for them.
90 Chapter 3: Placing Braces and Spaces
92 The other issue that always comes up in C styling is the placement of
93 braces. Unlike the indent size, there are few technical reasons to
94 choose one placement strategy over the other, but the preferred way, as
95 shown to us by the prophets Kernighan and Ritchie, is to put the opening
96 brace last on the line, and put the closing brace first, thusly:
102 This applies to all non-function statement blocks (if, switch, for,
116 However, there is one special case, namely functions: they have the
117 opening brace at the beginning of the next line, thus:
124 Heretic people all over the world have claimed that this inconsistency
125 is ... well ... inconsistent, but all right-thinking people know that
126 (a) K&R are _right_ and (b) K&R are right. Besides, functions are
127 special anyway (you can't nest them in C).
129 Note that the closing brace is empty on a line of its own, _except_ in
130 the cases where it is followed by a continuation of the same statement,
131 ie a "while" in a do-statement or an "else" in an if-statement, like
150 Also, note that this brace-placement also minimizes the number of empty
151 (or almost empty) lines, without any loss of readability. Thus, as the
152 supply of new-lines on your screen is not a renewable resource (think
153 25-line terminal screens here), you have more empty lines to put
156 Do not unnecessarily use braces where a single statement will do.
168 This does not apply if only one branch of a conditional statement is a single
169 statement; in the latter case use braces in both branches:
180 Linux kernel style for use of spaces depends (mostly) on
181 function-versus-keyword usage. Use a space after (most) keywords. The
182 notable exceptions are sizeof, typeof, alignof, and __attribute__, which look
183 somewhat like functions (and are usually used with parentheses in Linux,
184 although they are not required in the language, as in: "sizeof info" after
185 "struct fileinfo info;" is declared).
187 So use a space after these keywords:
189 if, switch, case, for, do, while
191 but not with sizeof, typeof, alignof, or __attribute__. E.g.,
193 s = sizeof(struct file);
195 Do not add spaces around (inside) parenthesized expressions. This example is
198 s = sizeof( struct file );
200 When declaring pointer data or a function that returns a pointer type, the
201 preferred use of '*' is adjacent to the data name or function name and not
202 adjacent to the type name. Examples:
205 unsigned long long memparse(char *ptr, char **retptr);
206 char *match_strdup(substring_t *s);
208 Use one space around (on each side of) most binary and ternary operators,
209 such as any of these:
211 = + - < > * / % | & ^ <= >= == != ? :
213 but no space after unary operators:
215 & * + - ~ ! sizeof typeof alignof __attribute__ defined
217 no space before the postfix increment & decrement unary operators:
221 no space after the prefix increment & decrement unary operators:
225 and no space around the '.' and "->" structure member operators.
227 Do not leave trailing whitespace at the ends of lines. Some editors with
228 "smart" indentation will insert whitespace at the beginning of new lines as
229 appropriate, so you can start typing the next line of code right away.
230 However, some such editors do not remove the whitespace if you end up not
231 putting a line of code there, such as if you leave a blank line. As a result,
232 you end up with lines containing trailing whitespace.
234 Git will warn you about patches that introduce trailing whitespace, and can
235 optionally strip the trailing whitespace for you; however, if applying a series
236 of patches, this may make later patches in the series fail by changing their
242 C is a Spartan language, and so should your naming be. Unlike Modula-2
243 and Pascal programmers, C programmers do not use cute names like
244 ThisVariableIsATemporaryCounter. A C programmer would call that
245 variable "tmp", which is much easier to write, and not the least more
246 difficult to understand.
248 HOWEVER, while mixed-case names are frowned upon, descriptive names for
249 global variables are a must. To call a global function "foo" is a
252 GLOBAL variables (to be used only if you _really_ need them) need to
253 have descriptive names, as do global functions. If you have a function
254 that counts the number of active users, you should call that
255 "count_active_users()" or similar, you should _not_ call it "cntusr()".
257 Encoding the type of a function into the name (so-called Hungarian
258 notation) is brain damaged - the compiler knows the types anyway and can
259 check those, and it only confuses the programmer. No wonder MicroSoft
260 makes buggy programs.
262 LOCAL variable names should be short, and to the point. If you have
263 some random integer loop counter, it should probably be called "i".
264 Calling it "loop_counter" is non-productive, if there is no chance of it
265 being mis-understood. Similarly, "tmp" can be just about any type of
266 variable that is used to hold a temporary value.
268 If you are afraid to mix up your local variable names, you have another
269 problem, which is called the function-growth-hormone-imbalance syndrome.
270 See chapter 6 (Functions).
275 Please don't use things like "vps_t".
276 It's a _mistake_ to use typedef for structures and pointers. When you see a
280 in the source, what does it mean?
281 In contrast, if it says
283 struct virtual_container *a;
285 you can actually tell what "a" is.
287 Lots of people think that typedefs "help readability". Not so. They are
290 (a) totally opaque objects (where the typedef is actively used to _hide_
293 Example: "pte_t" etc. opaque objects that you can only access using
294 the proper accessor functions.
296 NOTE! Opaqueness and "accessor functions" are not good in themselves.
297 The reason we have them for things like pte_t etc. is that there
298 really is absolutely _zero_ portably accessible information there.
300 (b) Clear integer types, where the abstraction _helps_ avoid confusion
301 whether it is "int" or "long".
303 u8/u16/u32 are perfectly fine typedefs, although they fit into
304 category (d) better than here.
306 NOTE! Again - there needs to be a _reason_ for this. If something is
307 "unsigned long", then there's no reason to do
309 typedef unsigned long myflags_t;
311 but if there is a clear reason for why it under certain circumstances
312 might be an "unsigned int" and under other configurations might be
313 "unsigned long", then by all means go ahead and use a typedef.
315 (c) when you use sparse to literally create a _new_ type for
318 (d) New types which are identical to standard C99 types, in certain
319 exceptional circumstances.
321 Although it would only take a short amount of time for the eyes and
322 brain to become accustomed to the standard types like 'uint32_t',
323 some people object to their use anyway.
325 Therefore, the Linux-specific 'u8/u16/u32/u64' types and their
326 signed equivalents which are identical to standard types are
327 permitted -- although they are not mandatory in new code of your
330 When editing existing code which already uses one or the other set
331 of types, you should conform to the existing choices in that code.
333 (e) Types safe for use in userspace.
335 In certain structures which are visible to userspace, we cannot
336 require C99 types and cannot use the 'u32' form above. Thus, we
337 use __u32 and similar types in all structures which are shared
340 Maybe there are other cases too, but the rule should basically be to NEVER
341 EVER use a typedef unless you can clearly match one of those rules.
343 In general, a pointer, or a struct that has elements that can reasonably
344 be directly accessed should _never_ be a typedef.
349 Functions should be short and sweet, and do just one thing. They should
350 fit on one or two screenfuls of text (the ISO/ANSI screen size is 80x24,
351 as we all know), and do one thing and do that well.
353 The maximum length of a function is inversely proportional to the
354 complexity and indentation level of that function. So, if you have a
355 conceptually simple function that is just one long (but simple)
356 case-statement, where you have to do lots of small things for a lot of
357 different cases, it's OK to have a longer function.
359 However, if you have a complex function, and you suspect that a
360 less-than-gifted first-year high-school student might not even
361 understand what the function is all about, you should adhere to the
362 maximum limits all the more closely. Use helper functions with
363 descriptive names (you can ask the compiler to in-line them if you think
364 it's performance-critical, and it will probably do a better job of it
365 than you would have done).
367 Another measure of the function is the number of local variables. They
368 shouldn't exceed 5-10, or you're doing something wrong. Re-think the
369 function, and split it into smaller pieces. A human brain can
370 generally easily keep track of about 7 different things, anything more
371 and it gets confused. You know you're brilliant, but maybe you'd like
372 to understand what you did 2 weeks from now.
374 In source files, separate functions with one blank line. If the function is
375 exported, the EXPORT* macro for it should follow immediately after the closing
376 function brace line. E.g.:
378 int system_is_up(void)
380 return system_state == SYSTEM_RUNNING;
382 EXPORT_SYMBOL(system_is_up);
384 In function prototypes, include parameter names with their data types.
385 Although this is not required by the C language, it is preferred in Linux
386 because it is a simple way to add valuable information for the reader.
389 Chapter 7: Centralized exiting of functions
391 Albeit deprecated by some people, the equivalent of the goto statement is
392 used frequently by compilers in form of the unconditional jump instruction.
394 The goto statement comes in handy when a function exits from multiple
395 locations and some common work such as cleanup has to be done. If there is no
396 cleanup needed then just return directly.
398 Choose label names which say what the goto does or why the goto exists. An
399 example of a good name could be "out_free_buffer:" if the goto frees "buffer".
400 Avoid using GW-BASIC names like "err1:" and "err2:", as you would have to
401 renumber them if you ever add or remove exit paths, and they make correctness
402 difficult to verify anyway.
404 It is advised to indent labels with a single space (not tab), so that
405 "diff -p" does not confuse labels with functions.
407 The rationale for using gotos is:
409 - unconditional statements are easier to understand and follow
411 - errors by not updating individual exit points when making
412 modifications are prevented
413 - saves the compiler work to optimize redundant code away ;)
420 buffer = kmalloc(SIZE, GFP_KERNEL);
437 A common type of bug to be aware of is "one err bugs" which look like this:
444 The bug in this code is that on some exit paths "foo" is NULL. Normally the
445 fix for this is to split it up into two error labels "err_free_bar:" and
454 Ideally you should simulate errors to test all exit paths.
457 Chapter 8: Commenting
459 Comments are good, but there is also a danger of over-commenting. NEVER
460 try to explain HOW your code works in a comment: it's much better to
461 write the code so that the _working_ is obvious, and it's a waste of
462 time to explain badly written code.
464 Generally, you want your comments to tell WHAT your code does, not HOW.
465 Also, try to avoid putting comments inside a function body: if the
466 function is so complex that you need to separately comment parts of it,
467 you should probably go back to chapter 6 for a while. You can make
468 small comments to note or warn about something particularly clever (or
469 ugly), but try to avoid excess. Instead, put the comments at the head
470 of the function, telling people what it does, and possibly WHY it does
473 When commenting the kernel API functions, please use the kernel-doc format.
474 See the files Documentation/kernel-documentation.rst and scripts/kernel-doc
477 Linux style for comments is the C89 "/* ... */" style.
478 Don't use C99-style "// ..." comments.
480 The preferred style for long (multi-line) comments is:
483 * This is the preferred style for multi-line
484 * comments in the Linux kernel source code.
485 * Please use it consistently.
487 * Description: A column of asterisks on the left side,
488 * with beginning and ending almost-blank lines.
491 For files in net/ and drivers/net/ the preferred style for long (multi-line)
492 comments is a little different.
494 /* The preferred comment style for files in net/ and drivers/net
497 * It is nearly the same as the generally preferred comment style,
498 * but there is no initial almost-blank line.
501 It's also important to comment data, whether they are basic types or derived
502 types. To this end, use just one data declaration per line (no commas for
503 multiple data declarations). This leaves you room for a small comment on each
504 item, explaining its use.
507 Chapter 9: You've made a mess of it
509 That's OK, we all do. You've probably been told by your long-time Unix
510 user helper that "GNU emacs" automatically formats the C sources for
511 you, and you've noticed that yes, it does do that, but the defaults it
512 uses are less than desirable (in fact, they are worse than random
513 typing - an infinite number of monkeys typing into GNU emacs would never
514 make a good program).
516 So, you can either get rid of GNU emacs, or change it to use saner
517 values. To do the latter, you can stick the following in your .emacs file:
519 (defun c-lineup-arglist-tabs-only (ignored)
520 "Line up argument lists by tabs, not spaces"
521 (let* ((anchor (c-langelem-pos c-syntactic-element))
522 (column (c-langelem-2nd-pos c-syntactic-element))
523 (offset (- (1+ column) anchor))
524 (steps (floor offset c-basic-offset)))
528 (add-hook 'c-mode-common-hook
533 '("linux" (c-offsets-alist
534 (arglist-cont-nonempty
536 c-lineup-arglist-tabs-only))))))
538 (add-hook 'c-mode-hook
540 (let ((filename (buffer-file-name)))
541 ;; Enable kernel mode for the appropriate files
543 (string-match (expand-file-name "~/src/linux-trees")
545 (setq indent-tabs-mode t)
546 (setq show-trailing-whitespace t)
547 (c-set-style "linux-tabs-only")))))
549 This will make emacs go better with the kernel coding style for C
550 files below ~/src/linux-trees.
552 But even if you fail in getting emacs to do sane formatting, not
553 everything is lost: use "indent".
555 Now, again, GNU indent has the same brain-dead settings that GNU emacs
556 has, which is why you need to give it a few command line options.
557 However, that's not too bad, because even the makers of GNU indent
558 recognize the authority of K&R (the GNU people aren't evil, they are
559 just severely misguided in this matter), so you just give indent the
560 options "-kr -i8" (stands for "K&R, 8 character indents"), or use
561 "scripts/Lindent", which indents in the latest style.
563 "indent" has a lot of options, and especially when it comes to comment
564 re-formatting you may want to take a look at the man page. But
565 remember: "indent" is not a fix for bad programming.
568 Chapter 10: Kconfig configuration files
570 For all of the Kconfig* configuration files throughout the source tree,
571 the indentation is somewhat different. Lines under a "config" definition
572 are indented with one tab, while help text is indented an additional two
576 bool "Auditing support"
579 Enable auditing infrastructure that can be used with another
580 kernel subsystem, such as SELinux (which requires this for
581 logging of avc messages output). Does not do system-call
582 auditing without CONFIG_AUDITSYSCALL.
584 Seriously dangerous features (such as write support for certain
585 filesystems) should advertise this prominently in their prompt string:
588 bool "ADFS write support (DANGEROUS)"
592 For full documentation on the configuration files, see the file
593 Documentation/kbuild/kconfig-language.txt.
596 Chapter 11: Data structures
598 Data structures that have visibility outside the single-threaded
599 environment they are created and destroyed in should always have
600 reference counts. In the kernel, garbage collection doesn't exist (and
601 outside the kernel garbage collection is slow and inefficient), which
602 means that you absolutely _have_ to reference count all your uses.
604 Reference counting means that you can avoid locking, and allows multiple
605 users to have access to the data structure in parallel - and not having
606 to worry about the structure suddenly going away from under them just
607 because they slept or did something else for a while.
609 Note that locking is _not_ a replacement for reference counting.
610 Locking is used to keep data structures coherent, while reference
611 counting is a memory management technique. Usually both are needed, and
612 they are not to be confused with each other.
614 Many data structures can indeed have two levels of reference counting,
615 when there are users of different "classes". The subclass count counts
616 the number of subclass users, and decrements the global count just once
617 when the subclass count goes to zero.
619 Examples of this kind of "multi-level-reference-counting" can be found in
620 memory management ("struct mm_struct": mm_users and mm_count), and in
621 filesystem code ("struct super_block": s_count and s_active).
623 Remember: if another thread can find your data structure, and you don't
624 have a reference count on it, you almost certainly have a bug.
627 Chapter 12: Macros, Enums and RTL
629 Names of macros defining constants and labels in enums are capitalized.
631 #define CONSTANT 0x12345
633 Enums are preferred when defining several related constants.
635 CAPITALIZED macro names are appreciated but macros resembling functions
636 may be named in lower case.
638 Generally, inline functions are preferable to macros resembling functions.
640 Macros with multiple statements should be enclosed in a do - while block:
642 #define macrofun(a, b, c) \
648 Things to avoid when using macros:
650 1) macros that affect control flow:
658 is a _very_ bad idea. It looks like a function call but exits the "calling"
659 function; don't break the internal parsers of those who will read the code.
661 2) macros that depend on having a local variable with a magic name:
663 #define FOO(val) bar(index, val)
665 might look like a good thing, but it's confusing as hell when one reads the
666 code and it's prone to breakage from seemingly innocent changes.
668 3) macros with arguments that are used as l-values: FOO(x) = y; will
669 bite you if somebody e.g. turns FOO into an inline function.
671 4) forgetting about precedence: macros defining constants using expressions
672 must enclose the expression in parentheses. Beware of similar issues with
673 macros using parameters.
675 #define CONSTANT 0x4000
676 #define CONSTEXP (CONSTANT | 3)
678 5) namespace collisions when defining local variables in macros resembling
688 ret is a common name for a local variable - __foo_ret is less likely
689 to collide with an existing variable.
691 The cpp manual deals with macros exhaustively. The gcc internals manual also
692 covers RTL which is used frequently with assembly language in the kernel.
695 Chapter 13: Printing kernel messages
697 Kernel developers like to be seen as literate. Do mind the spelling
698 of kernel messages to make a good impression. Do not use crippled
699 words like "dont"; use "do not" or "don't" instead. Make the messages
700 concise, clear, and unambiguous.
702 Kernel messages do not have to be terminated with a period.
704 Printing numbers in parentheses (%d) adds no value and should be avoided.
706 There are a number of driver model diagnostic macros in <linux/device.h>
707 which you should use to make sure messages are matched to the right device
708 and driver, and are tagged with the right level: dev_err(), dev_warn(),
709 dev_info(), and so forth. For messages that aren't associated with a
710 particular device, <linux/printk.h> defines pr_notice(), pr_info(),
711 pr_warn(), pr_err(), etc.
713 Coming up with good debugging messages can be quite a challenge; and once
714 you have them, they can be a huge help for remote troubleshooting. However
715 debug message printing is handled differently than printing other non-debug
716 messages. While the other pr_XXX() functions print unconditionally,
717 pr_debug() does not; it is compiled out by default, unless either DEBUG is
718 defined or CONFIG_DYNAMIC_DEBUG is set. That is true for dev_dbg() also,
719 and a related convention uses VERBOSE_DEBUG to add dev_vdbg() messages to
720 the ones already enabled by DEBUG.
722 Many subsystems have Kconfig debug options to turn on -DDEBUG in the
723 corresponding Makefile; in other cases specific files #define DEBUG. And
724 when a debug message should be unconditionally printed, such as if it is
725 already inside a debug-related #ifdef section, printk(KERN_DEBUG ...) can be
729 Chapter 14: Allocating memory
731 The kernel provides the following general purpose memory allocators:
732 kmalloc(), kzalloc(), kmalloc_array(), kcalloc(), vmalloc(), and
733 vzalloc(). Please refer to the API documentation for further information
736 The preferred form for passing a size of a struct is the following:
738 p = kmalloc(sizeof(*p), ...);
740 The alternative form where struct name is spelled out hurts readability and
741 introduces an opportunity for a bug when the pointer variable type is changed
742 but the corresponding sizeof that is passed to a memory allocator is not.
744 Casting the return value which is a void pointer is redundant. The conversion
745 from void pointer to any other pointer type is guaranteed by the C programming
748 The preferred form for allocating an array is the following:
750 p = kmalloc_array(n, sizeof(...), ...);
752 The preferred form for allocating a zeroed array is the following:
754 p = kcalloc(n, sizeof(...), ...);
756 Both forms check for overflow on the allocation size n * sizeof(...),
757 and return NULL if that occurred.
760 Chapter 15: The inline disease
762 There appears to be a common misperception that gcc has a magic "make me
763 faster" speedup option called "inline". While the use of inlines can be
764 appropriate (for example as a means of replacing macros, see Chapter 12), it
765 very often is not. Abundant use of the inline keyword leads to a much bigger
766 kernel, which in turn slows the system as a whole down, due to a bigger
767 icache footprint for the CPU and simply because there is less memory
768 available for the pagecache. Just think about it; a pagecache miss causes a
769 disk seek, which easily takes 5 milliseconds. There are a LOT of cpu cycles
770 that can go into these 5 milliseconds.
772 A reasonable rule of thumb is to not put inline at functions that have more
773 than 3 lines of code in them. An exception to this rule are the cases where
774 a parameter is known to be a compiletime constant, and as a result of this
775 constantness you *know* the compiler will be able to optimize most of your
776 function away at compile time. For a good example of this later case, see
777 the kmalloc() inline function.
779 Often people argue that adding inline to functions that are static and used
780 only once is always a win since there is no space tradeoff. While this is
781 technically correct, gcc is capable of inlining these automatically without
782 help, and the maintenance issue of removing the inline when a second user
783 appears outweighs the potential value of the hint that tells gcc to do
784 something it would have done anyway.
787 Chapter 16: Function return values and names
789 Functions can return values of many different kinds, and one of the
790 most common is a value indicating whether the function succeeded or
791 failed. Such a value can be represented as an error-code integer
792 (-Exxx = failure, 0 = success) or a "succeeded" boolean (0 = failure,
795 Mixing up these two sorts of representations is a fertile source of
796 difficult-to-find bugs. If the C language included a strong distinction
797 between integers and booleans then the compiler would find these mistakes
798 for us... but it doesn't. To help prevent such bugs, always follow this
801 If the name of a function is an action or an imperative command,
802 the function should return an error-code integer. If the name
803 is a predicate, the function should return a "succeeded" boolean.
805 For example, "add work" is a command, and the add_work() function returns 0
806 for success or -EBUSY for failure. In the same way, "PCI device present" is
807 a predicate, and the pci_dev_present() function returns 1 if it succeeds in
808 finding a matching device or 0 if it doesn't.
810 All EXPORTed functions must respect this convention, and so should all
811 public functions. Private (static) functions need not, but it is
812 recommended that they do.
814 Functions whose return value is the actual result of a computation, rather
815 than an indication of whether the computation succeeded, are not subject to
816 this rule. Generally they indicate failure by returning some out-of-range
817 result. Typical examples would be functions that return pointers; they use
818 NULL or the ERR_PTR mechanism to report failure.
821 Chapter 17: Don't re-invent the kernel macros
823 The header file include/linux/kernel.h contains a number of macros that
824 you should use, rather than explicitly coding some variant of them yourself.
825 For example, if you need to calculate the length of an array, take advantage
828 #define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
830 Similarly, if you need to calculate the size of some structure member, use
832 #define FIELD_SIZEOF(t, f) (sizeof(((t*)0)->f))
834 There are also min() and max() macros that do strict type checking if you
835 need them. Feel free to peruse that header file to see what else is already
836 defined that you shouldn't reproduce in your code.
839 Chapter 18: Editor modelines and other cruft
841 Some editors can interpret configuration information embedded in source files,
842 indicated with special markers. For example, emacs interprets lines marked
851 compile-command: "gcc -DMAGIC_DEBUG_FLAG foo.c"
855 Vim interprets markers that look like this:
857 /* vim:set sw=8 noet */
859 Do not include any of these in source files. People have their own personal
860 editor configurations, and your source files should not override them. This
861 includes markers for indentation and mode configuration. People may use their
862 own custom mode, or may have some other magic method for making indentation
866 Chapter 19: Inline assembly
868 In architecture-specific code, you may need to use inline assembly to interface
869 with CPU or platform functionality. Don't hesitate to do so when necessary.
870 However, don't use inline assembly gratuitously when C can do the job. You can
871 and should poke hardware from C when possible.
873 Consider writing simple helper functions that wrap common bits of inline
874 assembly, rather than repeatedly writing them with slight variations. Remember
875 that inline assembly can use C parameters.
877 Large, non-trivial assembly functions should go in .S files, with corresponding
878 C prototypes defined in C header files. The C prototypes for assembly
879 functions should use "asmlinkage".
881 You may need to mark your asm statement as volatile, to prevent GCC from
882 removing it if GCC doesn't notice any side effects. You don't always need to
883 do so, though, and doing so unnecessarily can limit optimization.
885 When writing a single inline assembly statement containing multiple
886 instructions, put each instruction on a separate line in a separate quoted
887 string, and end each string except the last with \n\t to properly indent the
888 next instruction in the assembly output:
890 asm ("magic %reg1, #42\n\t"
891 "more_magic %reg2, %reg3"
892 : /* outputs */ : /* inputs */ : /* clobbers */);
895 Chapter 20: Conditional Compilation
897 Wherever possible, don't use preprocessor conditionals (#if, #ifdef) in .c
898 files; doing so makes code harder to read and logic harder to follow. Instead,
899 use such conditionals in a header file defining functions for use in those .c
900 files, providing no-op stub versions in the #else case, and then call those
901 functions unconditionally from .c files. The compiler will avoid generating
902 any code for the stub calls, producing identical results, but the logic will
903 remain easy to follow.
905 Prefer to compile out entire functions, rather than portions of functions or
906 portions of expressions. Rather than putting an ifdef in an expression, factor
907 out part or all of the expression into a separate helper function and apply the
908 conditional to that function.
910 If you have a function or variable which may potentially go unused in a
911 particular configuration, and the compiler would warn about its definition
912 going unused, mark the definition as __maybe_unused rather than wrapping it in
913 a preprocessor conditional. (However, if a function or variable *always* goes
916 Within code, where possible, use the IS_ENABLED macro to convert a Kconfig
917 symbol into a C boolean expression, and use it in a normal C conditional:
919 if (IS_ENABLED(CONFIG_SOMETHING)) {
923 The compiler will constant-fold the conditional away, and include or exclude
924 the block of code just as with an #ifdef, so this will not add any runtime
925 overhead. However, this approach still allows the C compiler to see the code
926 inside the block, and check it for correctness (syntax, types, symbol
927 references, etc). Thus, you still have to use an #ifdef if the code inside the
928 block references symbols that will not exist if the condition is not met.
930 At the end of any non-trivial #if or #ifdef block (more than a few lines),
931 place a comment after the #endif on the same line, noting the conditional
932 expression used. For instance:
934 #ifdef CONFIG_SOMETHING
936 #endif /* CONFIG_SOMETHING */
939 Appendix I: References
941 The C Programming Language, Second Edition
942 by Brian W. Kernighan and Dennis M. Ritchie.
943 Prentice Hall, Inc., 1988.
944 ISBN 0-13-110362-8 (paperback), 0-13-110370-9 (hardback).
946 The Practice of Programming
947 by Brian W. Kernighan and Rob Pike.
948 Addison-Wesley, Inc., 1999.
951 GNU manuals - where in compliance with K&R and this text - for cpp, gcc,
952 gcc internals and indent, all available from http://www.gnu.org/manual/
954 WG14 is the international standardization working group for the programming
955 language C, URL: http://www.open-std.org/JTC1/SC22/WG14/
957 Kernel CodingStyle, by greg@kroah.com at OLS 2002:
958 http://www.kroah.com/linux/talks/ols_2002_kernel_codingstyle_talk/html/