1 Linux kernel coding style
2 =========================
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.
19 Tabs are 8 characters, and thus indentations are also 8 characters.
20 There are heretic movements that try to make indentations 4 (or even 2!)
21 characters deep, and that is akin to trying to define the value of PI to
24 Rationale: The whole idea behind indentation is to clearly define where
25 a block of control starts and ends. Especially when you've been looking
26 at your screen for 20 straight hours, you'll find it a lot easier to see
27 how the indentation works if you have large indentations.
29 Now, some people will claim that having 8-character indentations makes
30 the code move too far to the right, and makes it hard to read on a
31 80-character terminal screen. The answer to that is that if you need
32 more than 3 levels of indentation, you're screwed anyway, and should fix
35 In short, 8-char indents make things easier to read, and have the added
36 benefit of warning you when you're nesting your functions too deep.
39 The preferred way to ease multiple indentation levels in a switch statement is
40 to align the ``switch`` and its subordinate ``case`` labels in the same column
41 instead of ``double-indenting`` the ``case`` labels. E.g.:
62 Don't put multiple statements on a single line unless you have
67 if (condition) do_this;
68 do_something_everytime;
70 Don't put multiple assignments on a single line either. Kernel coding style
71 is super simple. Avoid tricky expressions.
73 Outside of comments, documentation and except in Kconfig, spaces are never
74 used for indentation, and the above example is deliberately broken.
76 Get a decent editor and don't leave whitespace at the end of lines.
79 2) Breaking long lines and strings
80 ----------------------------------
82 Coding style is all about readability and maintainability using commonly
85 The limit on the length of lines is 80 columns and this is a strongly
88 Statements longer than 80 columns will be broken into sensible chunks, unless
89 exceeding 80 columns significantly increases readability and does not hide
90 information. Descendants are always substantially shorter than the parent and
91 are placed substantially to the right. The same applies to function headers
92 with a long argument list. However, never break user-visible strings such as
93 printk messages, because that breaks the ability to grep for them.
96 3) Placing Braces and Spaces
97 ----------------------------
99 The other issue that always comes up in C styling is the placement of
100 braces. Unlike the indent size, there are few technical reasons to
101 choose one placement strategy over the other, but the preferred way, as
102 shown to us by the prophets Kernighan and Ritchie, is to put the opening
103 brace last on the line, and put the closing brace first, thusly:
111 This applies to all non-function statement blocks (if, switch, for,
127 However, there is one special case, namely functions: they have the
128 opening brace at the beginning of the next line, thus:
137 Heretic people all over the world have claimed that this inconsistency
138 is ... well ... inconsistent, but all right-thinking people know that
139 (a) K&R are **right** and (b) K&R are right. Besides, functions are
140 special anyway (you can't nest them in C).
142 Note that the closing brace is empty on a line of its own, **except** in
143 the cases where it is followed by a continuation of the same statement,
144 ie a ``while`` in a do-statement or an ``else`` in an if-statement, like
167 Also, note that this brace-placement also minimizes the number of empty
168 (or almost empty) lines, without any loss of readability. Thus, as the
169 supply of new-lines on your screen is not a renewable resource (think
170 25-line terminal screens here), you have more empty lines to put
173 Do not unnecessarily use braces where a single statement will do.
189 This does not apply if only one branch of a conditional statement is a single
190 statement; in the latter case use braces in both branches:
204 Linux kernel style for use of spaces depends (mostly) on
205 function-versus-keyword usage. Use a space after (most) keywords. The
206 notable exceptions are sizeof, typeof, alignof, and __attribute__, which look
207 somewhat like functions (and are usually used with parentheses in Linux,
208 although they are not required in the language, as in: ``sizeof info`` after
209 ``struct fileinfo info;`` is declared).
211 So use a space after these keywords::
213 if, switch, case, for, do, while
215 but not with sizeof, typeof, alignof, or __attribute__. E.g.,
220 s = sizeof(struct file);
222 Do not add spaces around (inside) parenthesized expressions. This example is
228 s = sizeof( struct file );
230 When declaring pointer data or a function that returns a pointer type, the
231 preferred use of ``*`` is adjacent to the data name or function name and not
232 adjacent to the type name. Examples:
238 unsigned long long memparse(char *ptr, char **retptr);
239 char *match_strdup(substring_t *s);
241 Use one space around (on each side of) most binary and ternary operators,
242 such as any of these::
244 = + - < > * / % | & ^ <= >= == != ? :
246 but no space after unary operators::
248 & * + - ~ ! sizeof typeof alignof __attribute__ defined
250 no space before the postfix increment & decrement unary operators::
254 no space after the prefix increment & decrement unary operators::
258 and no space around the ``.`` and ``->`` structure member operators.
260 Do not leave trailing whitespace at the ends of lines. Some editors with
261 ``smart`` indentation will insert whitespace at the beginning of new lines as
262 appropriate, so you can start typing the next line of code right away.
263 However, some such editors do not remove the whitespace if you end up not
264 putting a line of code there, such as if you leave a blank line. As a result,
265 you end up with lines containing trailing whitespace.
267 Git will warn you about patches that introduce trailing whitespace, and can
268 optionally strip the trailing whitespace for you; however, if applying a series
269 of patches, this may make later patches in the series fail by changing their
276 C is a Spartan language, and so should your naming be. Unlike Modula-2
277 and Pascal programmers, C programmers do not use cute names like
278 ThisVariableIsATemporaryCounter. A C programmer would call that
279 variable ``tmp``, which is much easier to write, and not the least more
280 difficult to understand.
282 HOWEVER, while mixed-case names are frowned upon, descriptive names for
283 global variables are a must. To call a global function ``foo`` is a
286 GLOBAL variables (to be used only if you **really** need them) need to
287 have descriptive names, as do global functions. If you have a function
288 that counts the number of active users, you should call that
289 ``count_active_users()`` or similar, you should **not** call it ``cntusr()``.
291 Encoding the type of a function into the name (so-called Hungarian
292 notation) is brain damaged - the compiler knows the types anyway and can
293 check those, and it only confuses the programmer. No wonder MicroSoft
294 makes buggy programs.
296 LOCAL variable names should be short, and to the point. If you have
297 some random integer loop counter, it should probably be called ``i``.
298 Calling it ``loop_counter`` is non-productive, if there is no chance of it
299 being mis-understood. Similarly, ``tmp`` can be just about any type of
300 variable that is used to hold a temporary value.
302 If you are afraid to mix up your local variable names, you have another
303 problem, which is called the function-growth-hormone-imbalance syndrome.
304 See chapter 6 (Functions).
310 Please don't use things like ``vps_t``.
311 It's a **mistake** to use typedef for structures and pointers. When you see a
318 in the source, what does it mean?
319 In contrast, if it says
323 struct virtual_container *a;
325 you can actually tell what ``a`` is.
327 Lots of people think that typedefs ``help readability``. Not so. They are
330 (a) totally opaque objects (where the typedef is actively used to **hide**
333 Example: ``pte_t`` etc. opaque objects that you can only access using
334 the proper accessor functions.
338 Opaqueness and ``accessor functions`` are not good in themselves.
339 The reason we have them for things like pte_t etc. is that there
340 really is absolutely **zero** portably accessible information there.
342 (b) Clear integer types, where the abstraction **helps** avoid confusion
343 whether it is ``int`` or ``long``.
345 u8/u16/u32 are perfectly fine typedefs, although they fit into
346 category (d) better than here.
350 Again - there needs to be a **reason** for this. If something is
351 ``unsigned long``, then there's no reason to do
353 typedef unsigned long myflags_t;
355 but if there is a clear reason for why it under certain circumstances
356 might be an ``unsigned int`` and under other configurations might be
357 ``unsigned long``, then by all means go ahead and use a typedef.
359 (c) when you use sparse to literally create a **new** type for
362 (d) New types which are identical to standard C99 types, in certain
363 exceptional circumstances.
365 Although it would only take a short amount of time for the eyes and
366 brain to become accustomed to the standard types like ``uint32_t``,
367 some people object to their use anyway.
369 Therefore, the Linux-specific ``u8/u16/u32/u64`` types and their
370 signed equivalents which are identical to standard types are
371 permitted -- although they are not mandatory in new code of your
374 When editing existing code which already uses one or the other set
375 of types, you should conform to the existing choices in that code.
377 (e) Types safe for use in userspace.
379 In certain structures which are visible to userspace, we cannot
380 require C99 types and cannot use the ``u32`` form above. Thus, we
381 use __u32 and similar types in all structures which are shared
384 Maybe there are other cases too, but the rule should basically be to NEVER
385 EVER use a typedef unless you can clearly match one of those rules.
387 In general, a pointer, or a struct that has elements that can reasonably
388 be directly accessed should **never** be a typedef.
394 Functions should be short and sweet, and do just one thing. They should
395 fit on one or two screenfuls of text (the ISO/ANSI screen size is 80x24,
396 as we all know), and do one thing and do that well.
398 The maximum length of a function is inversely proportional to the
399 complexity and indentation level of that function. So, if you have a
400 conceptually simple function that is just one long (but simple)
401 case-statement, where you have to do lots of small things for a lot of
402 different cases, it's OK to have a longer function.
404 However, if you have a complex function, and you suspect that a
405 less-than-gifted first-year high-school student might not even
406 understand what the function is all about, you should adhere to the
407 maximum limits all the more closely. Use helper functions with
408 descriptive names (you can ask the compiler to in-line them if you think
409 it's performance-critical, and it will probably do a better job of it
410 than you would have done).
412 Another measure of the function is the number of local variables. They
413 shouldn't exceed 5-10, or you're doing something wrong. Re-think the
414 function, and split it into smaller pieces. A human brain can
415 generally easily keep track of about 7 different things, anything more
416 and it gets confused. You know you're brilliant, but maybe you'd like
417 to understand what you did 2 weeks from now.
419 In source files, separate functions with one blank line. If the function is
420 exported, the **EXPORT** macro for it should follow immediately after the
421 closing function brace line. E.g.:
425 int system_is_up(void)
427 return system_state == SYSTEM_RUNNING;
429 EXPORT_SYMBOL(system_is_up);
431 In function prototypes, include parameter names with their data types.
432 Although this is not required by the C language, it is preferred in Linux
433 because it is a simple way to add valuable information for the reader.
436 7) Centralized exiting of functions
437 -----------------------------------
439 Albeit deprecated by some people, the equivalent of the goto statement is
440 used frequently by compilers in form of the unconditional jump instruction.
442 The goto statement comes in handy when a function exits from multiple
443 locations and some common work such as cleanup has to be done. If there is no
444 cleanup needed then just return directly.
446 Choose label names which say what the goto does or why the goto exists. An
447 example of a good name could be ``out_free_buffer:`` if the goto frees ``buffer``.
448 Avoid using GW-BASIC names like ``err1:`` and ``err2:``, as you would have to
449 renumber them if you ever add or remove exit paths, and they make correctness
450 difficult to verify anyway.
452 It is advised to indent labels with a single space (not tab), so that
453 ``diff -p`` does not confuse labels with functions.
455 The rationale for using gotos is:
457 - unconditional statements are easier to understand and follow
459 - errors by not updating individual exit points when making
460 modifications are prevented
461 - saves the compiler work to optimize redundant code away ;)
470 buffer = kmalloc(SIZE, GFP_KERNEL);
487 A common type of bug to be aware of is ``one err bugs`` which look like this:
496 The bug in this code is that on some exit paths ``foo`` is NULL. Normally the
497 fix for this is to split it up into two error labels ``err_free_bar:`` and
508 Ideally you should simulate errors to test all exit paths.
514 Comments are good, but there is also a danger of over-commenting. NEVER
515 try to explain HOW your code works in a comment: it's much better to
516 write the code so that the **working** is obvious, and it's a waste of
517 time to explain badly written code.
519 Generally, you want your comments to tell WHAT your code does, not HOW.
520 Also, try to avoid putting comments inside a function body: if the
521 function is so complex that you need to separately comment parts of it,
522 you should probably go back to chapter 6 for a while. You can make
523 small comments to note or warn about something particularly clever (or
524 ugly), but try to avoid excess. Instead, put the comments at the head
525 of the function, telling people what it does, and possibly WHY it does
528 When commenting the kernel API functions, please use the kernel-doc format.
529 See the files Documentation/kernel-documentation.rst and scripts/kernel-doc
532 The preferred style for long (multi-line) comments is:
537 * This is the preferred style for multi-line
538 * comments in the Linux kernel source code.
539 * Please use it consistently.
541 * Description: A column of asterisks on the left side,
542 * with beginning and ending almost-blank lines.
545 For files in net/ and drivers/net/ the preferred style for long (multi-line)
546 comments is a little different.
550 /* The preferred comment style for files in net/ and drivers/net
553 * It is nearly the same as the generally preferred comment style,
554 * but there is no initial almost-blank line.
557 It's also important to comment data, whether they are basic types or derived
558 types. To this end, use just one data declaration per line (no commas for
559 multiple data declarations). This leaves you room for a small comment on each
560 item, explaining its use.
563 9) You've made a mess of it
564 ---------------------------
566 That's OK, we all do. You've probably been told by your long-time Unix
567 user helper that ``GNU emacs`` automatically formats the C sources for
568 you, and you've noticed that yes, it does do that, but the defaults it
569 uses are less than desirable (in fact, they are worse than random
570 typing - an infinite number of monkeys typing into GNU emacs would never
571 make a good program).
573 So, you can either get rid of GNU emacs, or change it to use saner
574 values. To do the latter, you can stick the following in your .emacs file:
578 (defun c-lineup-arglist-tabs-only (ignored)
579 "Line up argument lists by tabs, not spaces"
580 (let* ((anchor (c-langelem-pos c-syntactic-element))
581 (column (c-langelem-2nd-pos c-syntactic-element))
582 (offset (- (1+ column) anchor))
583 (steps (floor offset c-basic-offset)))
587 (add-hook 'c-mode-common-hook
592 '("linux" (c-offsets-alist
593 (arglist-cont-nonempty
595 c-lineup-arglist-tabs-only))))))
597 (add-hook 'c-mode-hook
599 (let ((filename (buffer-file-name)))
600 ;; Enable kernel mode for the appropriate files
602 (string-match (expand-file-name "~/src/linux-trees")
604 (setq indent-tabs-mode t)
605 (setq show-trailing-whitespace t)
606 (c-set-style "linux-tabs-only")))))
608 This will make emacs go better with the kernel coding style for C
609 files below ``~/src/linux-trees``.
611 But even if you fail in getting emacs to do sane formatting, not
612 everything is lost: use ``indent``.
614 Now, again, GNU indent has the same brain-dead settings that GNU emacs
615 has, which is why you need to give it a few command line options.
616 However, that's not too bad, because even the makers of GNU indent
617 recognize the authority of K&R (the GNU people aren't evil, they are
618 just severely misguided in this matter), so you just give indent the
619 options ``-kr -i8`` (stands for ``K&R, 8 character indents``), or use
620 ``scripts/Lindent``, which indents in the latest style.
622 ``indent`` has a lot of options, and especially when it comes to comment
623 re-formatting you may want to take a look at the man page. But
624 remember: ``indent`` is not a fix for bad programming.
627 10) Kconfig configuration files
628 -------------------------------
630 For all of the Kconfig* configuration files throughout the source tree,
631 the indentation is somewhat different. Lines under a ``config`` definition
632 are indented with one tab, while help text is indented an additional two
636 bool "Auditing support"
639 Enable auditing infrastructure that can be used with another
640 kernel subsystem, such as SELinux (which requires this for
641 logging of avc messages output). Does not do system-call
642 auditing without CONFIG_AUDITSYSCALL.
644 Seriously dangerous features (such as write support for certain
645 filesystems) should advertise this prominently in their prompt string::
648 bool "ADFS write support (DANGEROUS)"
652 For full documentation on the configuration files, see the file
653 Documentation/kbuild/kconfig-language.txt.
659 Data structures that have visibility outside the single-threaded
660 environment they are created and destroyed in should always have
661 reference counts. In the kernel, garbage collection doesn't exist (and
662 outside the kernel garbage collection is slow and inefficient), which
663 means that you absolutely **have** to reference count all your uses.
665 Reference counting means that you can avoid locking, and allows multiple
666 users to have access to the data structure in parallel - and not having
667 to worry about the structure suddenly going away from under them just
668 because they slept or did something else for a while.
670 Note that locking is **not** a replacement for reference counting.
671 Locking is used to keep data structures coherent, while reference
672 counting is a memory management technique. Usually both are needed, and
673 they are not to be confused with each other.
675 Many data structures can indeed have two levels of reference counting,
676 when there are users of different ``classes``. The subclass count counts
677 the number of subclass users, and decrements the global count just once
678 when the subclass count goes to zero.
680 Examples of this kind of ``multi-level-reference-counting`` can be found in
681 memory management (``struct mm_struct``: mm_users and mm_count), and in
682 filesystem code (``struct super_block``: s_count and s_active).
684 Remember: if another thread can find your data structure, and you don't
685 have a reference count on it, you almost certainly have a bug.
688 12) Macros, Enums and RTL
689 -------------------------
691 Names of macros defining constants and labels in enums are capitalized.
695 #define CONSTANT 0x12345
697 Enums are preferred when defining several related constants.
699 CAPITALIZED macro names are appreciated but macros resembling functions
700 may be named in lower case.
702 Generally, inline functions are preferable to macros resembling functions.
704 Macros with multiple statements should be enclosed in a do - while block:
708 #define macrofun(a, b, c) \
714 Things to avoid when using macros:
716 1) macros that affect control flow:
726 is a **very** bad idea. It looks like a function call but exits the ``calling``
727 function; don't break the internal parsers of those who will read the code.
729 2) macros that depend on having a local variable with a magic name:
733 #define FOO(val) bar(index, val)
735 might look like a good thing, but it's confusing as hell when one reads the
736 code and it's prone to breakage from seemingly innocent changes.
738 3) macros with arguments that are used as l-values: FOO(x) = y; will
739 bite you if somebody e.g. turns FOO into an inline function.
741 4) forgetting about precedence: macros defining constants using expressions
742 must enclose the expression in parentheses. Beware of similar issues with
743 macros using parameters.
747 #define CONSTANT 0x4000
748 #define CONSTEXP (CONSTANT | 3)
750 5) namespace collisions when defining local variables in macros resembling
762 ret is a common name for a local variable - __foo_ret is less likely
763 to collide with an existing variable.
765 The cpp manual deals with macros exhaustively. The gcc internals manual also
766 covers RTL which is used frequently with assembly language in the kernel.
769 13) Printing kernel messages
770 ----------------------------
772 Kernel developers like to be seen as literate. Do mind the spelling
773 of kernel messages to make a good impression. Do not use crippled
774 words like ``dont``; use ``do not`` or ``don't`` instead. Make the messages
775 concise, clear, and unambiguous.
777 Kernel messages do not have to be terminated with a period.
779 Printing numbers in parentheses (%d) adds no value and should be avoided.
781 There are a number of driver model diagnostic macros in <linux/device.h>
782 which you should use to make sure messages are matched to the right device
783 and driver, and are tagged with the right level: dev_err(), dev_warn(),
784 dev_info(), and so forth. For messages that aren't associated with a
785 particular device, <linux/printk.h> defines pr_notice(), pr_info(),
786 pr_warn(), pr_err(), etc.
788 Coming up with good debugging messages can be quite a challenge; and once
789 you have them, they can be a huge help for remote troubleshooting. However
790 debug message printing is handled differently than printing other non-debug
791 messages. While the other pr_XXX() functions print unconditionally,
792 pr_debug() does not; it is compiled out by default, unless either DEBUG is
793 defined or CONFIG_DYNAMIC_DEBUG is set. That is true for dev_dbg() also,
794 and a related convention uses VERBOSE_DEBUG to add dev_vdbg() messages to
795 the ones already enabled by DEBUG.
797 Many subsystems have Kconfig debug options to turn on -DDEBUG in the
798 corresponding Makefile; in other cases specific files #define DEBUG. And
799 when a debug message should be unconditionally printed, such as if it is
800 already inside a debug-related #ifdef section, printk(KERN_DEBUG ...) can be
804 14) Allocating memory
805 ---------------------
807 The kernel provides the following general purpose memory allocators:
808 kmalloc(), kzalloc(), kmalloc_array(), kcalloc(), vmalloc(), and
809 vzalloc(). Please refer to the API documentation for further information
812 The preferred form for passing a size of a struct is the following:
816 p = kmalloc(sizeof(*p), ...);
818 The alternative form where struct name is spelled out hurts readability and
819 introduces an opportunity for a bug when the pointer variable type is changed
820 but the corresponding sizeof that is passed to a memory allocator is not.
822 Casting the return value which is a void pointer is redundant. The conversion
823 from void pointer to any other pointer type is guaranteed by the C programming
826 The preferred form for allocating an array is the following:
830 p = kmalloc_array(n, sizeof(...), ...);
832 The preferred form for allocating a zeroed array is the following:
836 p = kcalloc(n, sizeof(...), ...);
838 Both forms check for overflow on the allocation size n * sizeof(...),
839 and return NULL if that occurred.
842 15) The inline disease
843 ----------------------
845 There appears to be a common misperception that gcc has a magic "make me
846 faster" speedup option called ``inline``. While the use of inlines can be
847 appropriate (for example as a means of replacing macros, see Chapter 12), it
848 very often is not. Abundant use of the inline keyword leads to a much bigger
849 kernel, which in turn slows the system as a whole down, due to a bigger
850 icache footprint for the CPU and simply because there is less memory
851 available for the pagecache. Just think about it; a pagecache miss causes a
852 disk seek, which easily takes 5 milliseconds. There are a LOT of cpu cycles
853 that can go into these 5 milliseconds.
855 A reasonable rule of thumb is to not put inline at functions that have more
856 than 3 lines of code in them. An exception to this rule are the cases where
857 a parameter is known to be a compiletime constant, and as a result of this
858 constantness you *know* the compiler will be able to optimize most of your
859 function away at compile time. For a good example of this later case, see
860 the kmalloc() inline function.
862 Often people argue that adding inline to functions that are static and used
863 only once is always a win since there is no space tradeoff. While this is
864 technically correct, gcc is capable of inlining these automatically without
865 help, and the maintenance issue of removing the inline when a second user
866 appears outweighs the potential value of the hint that tells gcc to do
867 something it would have done anyway.
870 16) Function return values and names
871 ------------------------------------
873 Functions can return values of many different kinds, and one of the
874 most common is a value indicating whether the function succeeded or
875 failed. Such a value can be represented as an error-code integer
876 (-Exxx = failure, 0 = success) or a ``succeeded`` boolean (0 = failure,
879 Mixing up these two sorts of representations is a fertile source of
880 difficult-to-find bugs. If the C language included a strong distinction
881 between integers and booleans then the compiler would find these mistakes
882 for us... but it doesn't. To help prevent such bugs, always follow this
885 If the name of a function is an action or an imperative command,
886 the function should return an error-code integer. If the name
887 is a predicate, the function should return a "succeeded" boolean.
889 For example, ``add work`` is a command, and the add_work() function returns 0
890 for success or -EBUSY for failure. In the same way, ``PCI device present`` is
891 a predicate, and the pci_dev_present() function returns 1 if it succeeds in
892 finding a matching device or 0 if it doesn't.
894 All EXPORTed functions must respect this convention, and so should all
895 public functions. Private (static) functions need not, but it is
896 recommended that they do.
898 Functions whose return value is the actual result of a computation, rather
899 than an indication of whether the computation succeeded, are not subject to
900 this rule. Generally they indicate failure by returning some out-of-range
901 result. Typical examples would be functions that return pointers; they use
902 NULL or the ERR_PTR mechanism to report failure.
905 17) Don't re-invent the kernel macros
906 -------------------------------------
908 The header file include/linux/kernel.h contains a number of macros that
909 you should use, rather than explicitly coding some variant of them yourself.
910 For example, if you need to calculate the length of an array, take advantage
915 #define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
917 Similarly, if you need to calculate the size of some structure member, use
921 #define FIELD_SIZEOF(t, f) (sizeof(((t*)0)->f))
923 There are also min() and max() macros that do strict type checking if you
924 need them. Feel free to peruse that header file to see what else is already
925 defined that you shouldn't reproduce in your code.
928 18) Editor modelines and other cruft
929 ------------------------------------
931 Some editors can interpret configuration information embedded in source files,
932 indicated with special markers. For example, emacs interprets lines marked
945 compile-command: "gcc -DMAGIC_DEBUG_FLAG foo.c"
949 Vim interprets markers that look like this:
953 /* vim:set sw=8 noet */
955 Do not include any of these in source files. People have their own personal
956 editor configurations, and your source files should not override them. This
957 includes markers for indentation and mode configuration. People may use their
958 own custom mode, or may have some other magic method for making indentation
965 In architecture-specific code, you may need to use inline assembly to interface
966 with CPU or platform functionality. Don't hesitate to do so when necessary.
967 However, don't use inline assembly gratuitously when C can do the job. You can
968 and should poke hardware from C when possible.
970 Consider writing simple helper functions that wrap common bits of inline
971 assembly, rather than repeatedly writing them with slight variations. Remember
972 that inline assembly can use C parameters.
974 Large, non-trivial assembly functions should go in .S files, with corresponding
975 C prototypes defined in C header files. The C prototypes for assembly
976 functions should use ``asmlinkage``.
978 You may need to mark your asm statement as volatile, to prevent GCC from
979 removing it if GCC doesn't notice any side effects. You don't always need to
980 do so, though, and doing so unnecessarily can limit optimization.
982 When writing a single inline assembly statement containing multiple
983 instructions, put each instruction on a separate line in a separate quoted
984 string, and end each string except the last with \n\t to properly indent the
985 next instruction in the assembly output:
989 asm ("magic %reg1, #42\n\t"
990 "more_magic %reg2, %reg3"
991 : /* outputs */ : /* inputs */ : /* clobbers */);
994 20) Conditional Compilation
995 ---------------------------
997 Wherever possible, don't use preprocessor conditionals (#if, #ifdef) in .c
998 files; doing so makes code harder to read and logic harder to follow. Instead,
999 use such conditionals in a header file defining functions for use in those .c
1000 files, providing no-op stub versions in the #else case, and then call those
1001 functions unconditionally from .c files. The compiler will avoid generating
1002 any code for the stub calls, producing identical results, but the logic will
1003 remain easy to follow.
1005 Prefer to compile out entire functions, rather than portions of functions or
1006 portions of expressions. Rather than putting an ifdef in an expression, factor
1007 out part or all of the expression into a separate helper function and apply the
1008 conditional to that function.
1010 If you have a function or variable which may potentially go unused in a
1011 particular configuration, and the compiler would warn about its definition
1012 going unused, mark the definition as __maybe_unused rather than wrapping it in
1013 a preprocessor conditional. (However, if a function or variable *always* goes
1016 Within code, where possible, use the IS_ENABLED macro to convert a Kconfig
1017 symbol into a C boolean expression, and use it in a normal C conditional:
1021 if (IS_ENABLED(CONFIG_SOMETHING)) {
1025 The compiler will constant-fold the conditional away, and include or exclude
1026 the block of code just as with an #ifdef, so this will not add any runtime
1027 overhead. However, this approach still allows the C compiler to see the code
1028 inside the block, and check it for correctness (syntax, types, symbol
1029 references, etc). Thus, you still have to use an #ifdef if the code inside the
1030 block references symbols that will not exist if the condition is not met.
1032 At the end of any non-trivial #if or #ifdef block (more than a few lines),
1033 place a comment after the #endif on the same line, noting the conditional
1034 expression used. For instance:
1038 #ifdef CONFIG_SOMETHING
1040 #endif /* CONFIG_SOMETHING */
1043 Appendix I) References
1044 ----------------------
1046 The C Programming Language, Second Edition
1047 by Brian W. Kernighan and Dennis M. Ritchie.
1048 Prentice Hall, Inc., 1988.
1049 ISBN 0-13-110362-8 (paperback), 0-13-110370-9 (hardback).
1051 The Practice of Programming
1052 by Brian W. Kernighan and Rob Pike.
1053 Addison-Wesley, Inc., 1999.
1056 GNU manuals - where in compliance with K&R and this text - for cpp, gcc,
1057 gcc internals and indent, all available from http://www.gnu.org/manual/
1059 WG14 is the international standardization working group for the programming
1060 language C, URL: http://www.open-std.org/JTC1/SC22/WG14/
1062 Kernel CodingStyle, by greg@kroah.com at OLS 2002:
1063 http://www.kroah.com/linux/talks/ols_2002_kernel_codingstyle_talk/html/