blob: b49b92edb396835632781470f18bb31ab0055387 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001
2 Linux kernel coding style
3
4This is a short document describing the preferred coding style for the
5linux kernel. Coding style is very personal, and I won't _force_ my
6views on anybody, but this is what goes for anything that I have to be
7able to maintain, and I'd prefer it for most other things too. Please
8at least consider the points made here.
9
10First off, I'd suggest printing out a copy of the GNU coding standards,
11and NOT read it. Burn them, it's a great symbolic gesture.
12
13Anyway, here goes:
14
15
16 Chapter 1: Indentation
17
18Tabs are 8 characters, and thus indentations are also 8 characters.
19There are heretic movements that try to make indentations 4 (or even 2!)
20characters deep, and that is akin to trying to define the value of PI to
21be 3.
22
23Rationale: The whole idea behind indentation is to clearly define where
24a block of control starts and ends. Especially when you've been looking
25at your screen for 20 straight hours, you'll find it a lot easier to see
26how the indentation works if you have large indentations.
27
28Now, some people will claim that having 8-character indentations makes
29the code move too far to the right, and makes it hard to read on a
3080-character terminal screen. The answer to that is that if you need
31more than 3 levels of indentation, you're screwed anyway, and should fix
32your program.
33
34In short, 8-char indents make things easier to read, and have the added
35benefit of warning you when you're nesting your functions too deep.
36Heed that warning.
37
Randy Dunlapb3fc9942006-12-10 02:18:56 -080038The preferred way to ease multiple indentation levels in a switch statement is
39to align the "switch" and its subordinate "case" labels in the same column
40instead of "double-indenting" the "case" labels. E.g.:
41
42 switch (suffix) {
43 case 'G':
44 case 'g':
45 mem <<= 30;
46 break;
47 case 'M':
48 case 'm':
49 mem <<= 20;
50 break;
51 case 'K':
52 case 'k':
53 mem <<= 10;
54 /* fall through */
55 default:
56 break;
57 }
58
59
Linus Torvalds1da177e2005-04-16 15:20:36 -070060Don't put multiple statements on a single line unless you have
61something to hide:
62
63 if (condition) do_this;
64 do_something_everytime;
65
Randy Dunlapb3fc9942006-12-10 02:18:56 -080066Don't put multiple assignments on a single line either. Kernel coding style
67is super simple. Avoid tricky expressions.
68
Linus Torvalds1da177e2005-04-16 15:20:36 -070069Outside of comments, documentation and except in Kconfig, spaces are never
70used for indentation, and the above example is deliberately broken.
71
72Get a decent editor and don't leave whitespace at the end of lines.
73
74
75 Chapter 2: Breaking long lines and strings
76
77Coding style is all about readability and maintainability using commonly
78available tools.
79
80The limit on the length of lines is 80 columns and this is a hard limit.
81
82Statements longer than 80 columns will be broken into sensible chunks.
83Descendants are always substantially shorter than the parent and are placed
84substantially to the right. The same applies to function headers with a long
85argument list. Long strings are as well broken into shorter strings.
86
87void fun(int a, int b, int c)
88{
89 if (condition)
90 printk(KERN_WARNING "Warning this is a long printk with "
91 "3 parameters a: %u b: %u "
92 "c: %u \n", a, b, c);
93 else
94 next_statement;
95}
96
Randy Dunlapb3fc9942006-12-10 02:18:56 -080097 Chapter 3: Placing Braces and Spaces
Linus Torvalds1da177e2005-04-16 15:20:36 -070098
99The other issue that always comes up in C styling is the placement of
100braces. Unlike the indent size, there are few technical reasons to
101choose one placement strategy over the other, but the preferred way, as
102shown to us by the prophets Kernighan and Ritchie, is to put the opening
103brace last on the line, and put the closing brace first, thusly:
104
105 if (x is true) {
106 we do y
107 }
108
Randy Dunlapb3fc9942006-12-10 02:18:56 -0800109This applies to all non-function statement blocks (if, switch, for,
110while, do). E.g.:
111
112 switch (action) {
113 case KOBJ_ADD:
114 return "add";
115 case KOBJ_REMOVE:
116 return "remove";
117 case KOBJ_CHANGE:
118 return "change";
119 default:
120 return NULL;
121 }
122
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123However, there is one special case, namely functions: they have the
124opening brace at the beginning of the next line, thus:
125
126 int function(int x)
127 {
128 body of function
129 }
130
131Heretic people all over the world have claimed that this inconsistency
132is ... well ... inconsistent, but all right-thinking people know that
133(a) K&R are _right_ and (b) K&R are right. Besides, functions are
134special anyway (you can't nest them in C).
135
136Note that the closing brace is empty on a line of its own, _except_ in
137the cases where it is followed by a continuation of the same statement,
138ie a "while" in a do-statement or an "else" in an if-statement, like
139this:
140
141 do {
142 body of do-loop
143 } while (condition);
144
145and
146
147 if (x == y) {
148 ..
149 } else if (x > y) {
150 ...
151 } else {
152 ....
153 }
154
155Rationale: K&R.
156
157Also, note that this brace-placement also minimizes the number of empty
158(or almost empty) lines, without any loss of readability. Thus, as the
159supply of new-lines on your screen is not a renewable resource (think
16025-line terminal screens here), you have more empty lines to put
161comments on.
162
Oliver Neukume659ba42007-05-08 00:30:34 -0700163Do not unnecessarily use braces where a single statement will do.
164
165if (condition)
166 action();
167
168This does not apply if one branch of a conditional statement is a single
169statement. Use braces in both branches.
170
171if (condition) {
172 do_this();
173 do_that();
174} else {
175 otherwise();
176}
177
Randy Dunlapb3fc9942006-12-10 02:18:56 -0800178 3.1: Spaces
179
180Linux kernel style for use of spaces depends (mostly) on
181function-versus-keyword usage. Use a space after (most) keywords. The
182notable exceptions are sizeof, typeof, alignof, and __attribute__, which look
183somewhat like functions (and are usually used with parentheses in Linux,
184although they are not required in the language, as in: "sizeof info" after
185"struct fileinfo info;" is declared).
186
187So use a space after these keywords:
188 if, switch, case, for, do, while
189but not with sizeof, typeof, alignof, or __attribute__. E.g.,
190 s = sizeof(struct file);
191
192Do not add spaces around (inside) parenthesized expressions. This example is
193*bad*:
194
195 s = sizeof( struct file );
196
197When declaring pointer data or a function that returns a pointer type, the
198preferred use of '*' is adjacent to the data name or function name and not
199adjacent to the type name. Examples:
200
201 char *linux_banner;
202 unsigned long long memparse(char *ptr, char **retptr);
203 char *match_strdup(substring_t *s);
204
205Use one space around (on each side of) most binary and ternary operators,
206such as any of these:
207
208 = + - < > * / % | & ^ <= >= == != ? :
209
210but no space after unary operators:
211 & * + - ~ ! sizeof typeof alignof __attribute__ defined
212
213no space before the postfix increment & decrement unary operators:
214 ++ --
215
216no space after the prefix increment & decrement unary operators:
217 ++ --
218
219and no space around the '.' and "->" structure member operators.
220
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221
222 Chapter 4: Naming
223
224C is a Spartan language, and so should your naming be. Unlike Modula-2
225and Pascal programmers, C programmers do not use cute names like
226ThisVariableIsATemporaryCounter. A C programmer would call that
227variable "tmp", which is much easier to write, and not the least more
228difficult to understand.
229
230HOWEVER, while mixed-case names are frowned upon, descriptive names for
231global variables are a must. To call a global function "foo" is a
232shooting offense.
233
234GLOBAL variables (to be used only if you _really_ need them) need to
235have descriptive names, as do global functions. If you have a function
236that counts the number of active users, you should call that
237"count_active_users()" or similar, you should _not_ call it "cntusr()".
238
239Encoding the type of a function into the name (so-called Hungarian
240notation) is brain damaged - the compiler knows the types anyway and can
241check those, and it only confuses the programmer. No wonder MicroSoft
242makes buggy programs.
243
244LOCAL variable names should be short, and to the point. If you have
245some random integer loop counter, it should probably be called "i".
246Calling it "loop_counter" is non-productive, if there is no chance of it
247being mis-understood. Similarly, "tmp" can be just about any type of
248variable that is used to hold a temporary value.
249
250If you are afraid to mix up your local variable names, you have another
251problem, which is called the function-growth-hormone-imbalance syndrome.
Randy Dunlapb3fc9942006-12-10 02:18:56 -0800252See chapter 6 (Functions).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253
254
Randy Dunlap226a6b82006-06-23 02:05:58 -0700255 Chapter 5: Typedefs
256
257Please don't use things like "vps_t".
258
259It's a _mistake_ to use typedef for structures and pointers. When you see a
260
261 vps_t a;
262
263in the source, what does it mean?
264
265In contrast, if it says
266
267 struct virtual_container *a;
268
269you can actually tell what "a" is.
270
271Lots of people think that typedefs "help readability". Not so. They are
272useful only for:
273
274 (a) totally opaque objects (where the typedef is actively used to _hide_
275 what the object is).
276
277 Example: "pte_t" etc. opaque objects that you can only access using
278 the proper accessor functions.
279
280 NOTE! Opaqueness and "accessor functions" are not good in themselves.
281 The reason we have them for things like pte_t etc. is that there
282 really is absolutely _zero_ portably accessible information there.
283
284 (b) Clear integer types, where the abstraction _helps_ avoid confusion
285 whether it is "int" or "long".
286
287 u8/u16/u32 are perfectly fine typedefs, although they fit into
288 category (d) better than here.
289
290 NOTE! Again - there needs to be a _reason_ for this. If something is
291 "unsigned long", then there's no reason to do
292
293 typedef unsigned long myflags_t;
294
295 but if there is a clear reason for why it under certain circumstances
296 might be an "unsigned int" and under other configurations might be
297 "unsigned long", then by all means go ahead and use a typedef.
298
299 (c) when you use sparse to literally create a _new_ type for
300 type-checking.
301
302 (d) New types which are identical to standard C99 types, in certain
303 exceptional circumstances.
304
305 Although it would only take a short amount of time for the eyes and
306 brain to become accustomed to the standard types like 'uint32_t',
307 some people object to their use anyway.
308
309 Therefore, the Linux-specific 'u8/u16/u32/u64' types and their
310 signed equivalents which are identical to standard types are
311 permitted -- although they are not mandatory in new code of your
312 own.
313
314 When editing existing code which already uses one or the other set
315 of types, you should conform to the existing choices in that code.
316
317 (e) Types safe for use in userspace.
318
319 In certain structures which are visible to userspace, we cannot
320 require C99 types and cannot use the 'u32' form above. Thus, we
321 use __u32 and similar types in all structures which are shared
322 with userspace.
323
324Maybe there are other cases too, but the rule should basically be to NEVER
325EVER use a typedef unless you can clearly match one of those rules.
326
327In general, a pointer, or a struct that has elements that can reasonably
328be directly accessed should _never_ be a typedef.
329
330
331 Chapter 6: Functions
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332
333Functions should be short and sweet, and do just one thing. They should
334fit on one or two screenfuls of text (the ISO/ANSI screen size is 80x24,
335as we all know), and do one thing and do that well.
336
337The maximum length of a function is inversely proportional to the
338complexity and indentation level of that function. So, if you have a
339conceptually simple function that is just one long (but simple)
340case-statement, where you have to do lots of small things for a lot of
341different cases, it's OK to have a longer function.
342
343However, if you have a complex function, and you suspect that a
344less-than-gifted first-year high-school student might not even
345understand what the function is all about, you should adhere to the
346maximum limits all the more closely. Use helper functions with
347descriptive names (you can ask the compiler to in-line them if you think
348it's performance-critical, and it will probably do a better job of it
349than you would have done).
350
351Another measure of the function is the number of local variables. They
352shouldn't exceed 5-10, or you're doing something wrong. Re-think the
353function, and split it into smaller pieces. A human brain can
354generally easily keep track of about 7 different things, anything more
355and it gets confused. You know you're brilliant, but maybe you'd like
356to understand what you did 2 weeks from now.
357
Randy Dunlapb3fc9942006-12-10 02:18:56 -0800358In source files, separate functions with one blank line. If the function is
359exported, the EXPORT* macro for it should follow immediately after the closing
360function brace line. E.g.:
361
362int system_is_up(void)
363{
364 return system_state == SYSTEM_RUNNING;
365}
366EXPORT_SYMBOL(system_is_up);
367
368In function prototypes, include parameter names with their data types.
369Although this is not required by the C language, it is preferred in Linux
370because it is a simple way to add valuable information for the reader.
371
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372
Randy Dunlap226a6b82006-06-23 02:05:58 -0700373 Chapter 7: Centralized exiting of functions
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374
375Albeit deprecated by some people, the equivalent of the goto statement is
376used frequently by compilers in form of the unconditional jump instruction.
377
378The goto statement comes in handy when a function exits from multiple
379locations and some common work such as cleanup has to be done.
380
381The rationale is:
382
383- unconditional statements are easier to understand and follow
384- nesting is reduced
385- errors by not updating individual exit points when making
386 modifications are prevented
387- saves the compiler work to optimize redundant code away ;)
388
Jesper Juhldc3d28d2006-01-09 20:53:51 -0800389int fun(int a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390{
391 int result = 0;
392 char *buffer = kmalloc(SIZE);
393
394 if (buffer == NULL)
395 return -ENOMEM;
396
397 if (condition1) {
398 while (loop1) {
399 ...
400 }
401 result = 1;
402 goto out;
403 }
404 ...
405out:
406 kfree(buffer);
407 return result;
408}
409
Randy Dunlap226a6b82006-06-23 02:05:58 -0700410 Chapter 8: Commenting
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411
412Comments are good, but there is also a danger of over-commenting. NEVER
413try to explain HOW your code works in a comment: it's much better to
414write the code so that the _working_ is obvious, and it's a waste of
415time to explain badly written code.
416
417Generally, you want your comments to tell WHAT your code does, not HOW.
418Also, try to avoid putting comments inside a function body: if the
419function is so complex that you need to separately comment parts of it,
Randy Dunlapb3fc9942006-12-10 02:18:56 -0800420you should probably go back to chapter 6 for a while. You can make
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421small comments to note or warn about something particularly clever (or
422ugly), but try to avoid excess. Instead, put the comments at the head
423of the function, telling people what it does, and possibly WHY it does
424it.
425
Randy Dunlapb3fc9942006-12-10 02:18:56 -0800426When commenting the kernel API functions, please use the kernel-doc format.
Pekka J Enberge776eba2005-09-10 00:26:44 -0700427See the files Documentation/kernel-doc-nano-HOWTO.txt and scripts/kernel-doc
428for details.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429
Randy Dunlapb3fc9942006-12-10 02:18:56 -0800430Linux style for comments is the C89 "/* ... */" style.
431Don't use C99-style "// ..." comments.
432
433The preferred style for long (multi-line) comments is:
434
435 /*
436 * This is the preferred style for multi-line
437 * comments in the Linux kernel source code.
438 * Please use it consistently.
439 *
440 * Description: A column of asterisks on the left side,
441 * with beginning and ending almost-blank lines.
442 */
443
444It's also important to comment data, whether they are basic types or derived
445types. To this end, use just one data declaration per line (no commas for
446multiple data declarations). This leaves you room for a small comment on each
447item, explaining its use.
448
449
Randy Dunlap226a6b82006-06-23 02:05:58 -0700450 Chapter 9: You've made a mess of it
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451
452That's OK, we all do. You've probably been told by your long-time Unix
453user helper that "GNU emacs" automatically formats the C sources for
454you, and you've noticed that yes, it does do that, but the defaults it
455uses are less than desirable (in fact, they are worse than random
456typing - an infinite number of monkeys typing into GNU emacs would never
457make a good program).
458
459So, you can either get rid of GNU emacs, or change it to use saner
460values. To do the latter, you can stick the following in your .emacs file:
461
462(defun linux-c-mode ()
463 "C mode with adjusted defaults for use with the Linux kernel."
464 (interactive)
465 (c-mode)
466 (c-set-style "K&R")
467 (setq tab-width 8)
468 (setq indent-tabs-mode t)
469 (setq c-basic-offset 8))
470
471This will define the M-x linux-c-mode command. When hacking on a
472module, if you put the string -*- linux-c -*- somewhere on the first
473two lines, this mode will be automatically invoked. Also, you may want
474to add
475
476(setq auto-mode-alist (cons '("/usr/src/linux.*/.*\\.[ch]$" . linux-c-mode)
477 auto-mode-alist))
478
479to your .emacs file if you want to have linux-c-mode switched on
480automagically when you edit source files under /usr/src/linux.
481
482But even if you fail in getting emacs to do sane formatting, not
483everything is lost: use "indent".
484
485Now, again, GNU indent has the same brain-dead settings that GNU emacs
486has, which is why you need to give it a few command line options.
487However, that's not too bad, because even the makers of GNU indent
488recognize the authority of K&R (the GNU people aren't evil, they are
489just severely misguided in this matter), so you just give indent the
490options "-kr -i8" (stands for "K&R, 8 character indents"), or use
491"scripts/Lindent", which indents in the latest style.
492
493"indent" has a lot of options, and especially when it comes to comment
494re-formatting you may want to take a look at the man page. But
495remember: "indent" is not a fix for bad programming.
496
497
Robert P. J. Day6754bb42007-05-23 13:57:42 -0700498 Chapter 10: Kconfig configuration files
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499
Robert P. J. Day6754bb42007-05-23 13:57:42 -0700500For all of the Kconfig* configuration files throughout the source tree,
501the indentation is somewhat different. Lines under a "config" definition
502are indented with one tab, while help text is indented an additional two
503spaces. Example:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
Robert P. J. Day6754bb42007-05-23 13:57:42 -0700505config AUDIT
506 bool "Auditing support"
507 depends on NET
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 help
Robert P. J. Day6754bb42007-05-23 13:57:42 -0700509 Enable auditing infrastructure that can be used with another
510 kernel subsystem, such as SELinux (which requires this for
511 logging of avc messages output). Does not do system-call
512 auditing without CONFIG_AUDITSYSCALL.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513
Robert P. J. Day6754bb42007-05-23 13:57:42 -0700514Features that might still be considered unstable should be defined as
515dependent on "EXPERIMENTAL":
516
517config SLUB
518 depends on EXPERIMENTAL && !ARCH_USES_SLAB_PAGE_STRUCT
519 bool "SLUB (Unqueued Allocator)"
520 ...
521
522while seriously dangerous features (such as write support for certain
523filesystems) should advertise this prominently in their prompt string:
524
525config ADFS_FS_RW
526 bool "ADFS write support (DANGEROUS)"
527 depends on ADFS_FS
528 ...
529
530For full documentation on the configuration files, see the file
531Documentation/kbuild/kconfig-language.txt.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532
533
Randy Dunlap226a6b82006-06-23 02:05:58 -0700534 Chapter 11: Data structures
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535
536Data structures that have visibility outside the single-threaded
537environment they are created and destroyed in should always have
538reference counts. In the kernel, garbage collection doesn't exist (and
539outside the kernel garbage collection is slow and inefficient), which
540means that you absolutely _have_ to reference count all your uses.
541
542Reference counting means that you can avoid locking, and allows multiple
543users to have access to the data structure in parallel - and not having
544to worry about the structure suddenly going away from under them just
545because they slept or did something else for a while.
546
547Note that locking is _not_ a replacement for reference counting.
548Locking is used to keep data structures coherent, while reference
549counting is a memory management technique. Usually both are needed, and
550they are not to be confused with each other.
551
552Many data structures can indeed have two levels of reference counting,
553when there are users of different "classes". The subclass count counts
554the number of subclass users, and decrements the global count just once
555when the subclass count goes to zero.
556
557Examples of this kind of "multi-level-reference-counting" can be found in
558memory management ("struct mm_struct": mm_users and mm_count), and in
559filesystem code ("struct super_block": s_count and s_active).
560
561Remember: if another thread can find your data structure, and you don't
562have a reference count on it, you almost certainly have a bug.
563
564
Randy Dunlap226a6b82006-06-23 02:05:58 -0700565 Chapter 12: Macros, Enums and RTL
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566
567Names of macros defining constants and labels in enums are capitalized.
568
569#define CONSTANT 0x12345
570
571Enums are preferred when defining several related constants.
572
573CAPITALIZED macro names are appreciated but macros resembling functions
574may be named in lower case.
575
576Generally, inline functions are preferable to macros resembling functions.
577
578Macros with multiple statements should be enclosed in a do - while block:
579
580#define macrofun(a, b, c) \
581 do { \
582 if (a == 5) \
583 do_this(b, c); \
584 } while (0)
585
586Things to avoid when using macros:
587
5881) macros that affect control flow:
589
590#define FOO(x) \
591 do { \
592 if (blah(x) < 0) \
593 return -EBUGGERED; \
594 } while(0)
595
596is a _very_ bad idea. It looks like a function call but exits the "calling"
597function; don't break the internal parsers of those who will read the code.
598
5992) macros that depend on having a local variable with a magic name:
600
601#define FOO(val) bar(index, val)
602
603might look like a good thing, but it's confusing as hell when one reads the
604code and it's prone to breakage from seemingly innocent changes.
605
6063) macros with arguments that are used as l-values: FOO(x) = y; will
607bite you if somebody e.g. turns FOO into an inline function.
608
6094) forgetting about precedence: macros defining constants using expressions
610must enclose the expression in parentheses. Beware of similar issues with
611macros using parameters.
612
613#define CONSTANT 0x4000
614#define CONSTEXP (CONSTANT | 3)
615
616The cpp manual deals with macros exhaustively. The gcc internals manual also
617covers RTL which is used frequently with assembly language in the kernel.
618
619
Randy Dunlap226a6b82006-06-23 02:05:58 -0700620 Chapter 13: Printing kernel messages
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621
622Kernel developers like to be seen as literate. Do mind the spelling
623of kernel messages to make a good impression. Do not use crippled
624words like "dont" and use "do not" or "don't" instead.
625
626Kernel messages do not have to be terminated with a period.
627
628Printing numbers in parentheses (%d) adds no value and should be avoided.
629
630
Randy Dunlap226a6b82006-06-23 02:05:58 -0700631 Chapter 14: Allocating memory
Pekka J Enbergaf4e5a22005-09-16 19:28:11 -0700632
633The kernel provides the following general purpose memory allocators:
634kmalloc(), kzalloc(), kcalloc(), and vmalloc(). Please refer to the API
635documentation for further information about them.
636
637The preferred form for passing a size of a struct is the following:
638
639 p = kmalloc(sizeof(*p), ...);
640
641The alternative form where struct name is spelled out hurts readability and
642introduces an opportunity for a bug when the pointer variable type is changed
643but the corresponding sizeof that is passed to a memory allocator is not.
644
645Casting the return value which is a void pointer is redundant. The conversion
646from void pointer to any other pointer type is guaranteed by the C programming
647language.
648
649
Randy Dunlap226a6b82006-06-23 02:05:58 -0700650 Chapter 15: The inline disease
Arjan van de Vena771f2b2006-01-08 01:05:04 -0800651
652There appears to be a common misperception that gcc has a magic "make me
653faster" speedup option called "inline". While the use of inlines can be
Jesper Juhl53ab97a2007-05-08 00:31:06 -0700654appropriate (for example as a means of replacing macros, see Chapter 12), it
Arjan van de Vena771f2b2006-01-08 01:05:04 -0800655very often is not. Abundant use of the inline keyword leads to a much bigger
656kernel, which in turn slows the system as a whole down, due to a bigger
657icache footprint for the CPU and simply because there is less memory
658available for the pagecache. Just think about it; a pagecache miss causes a
659disk seek, which easily takes 5 miliseconds. There are a LOT of cpu cycles
660that can go into these 5 miliseconds.
661
662A reasonable rule of thumb is to not put inline at functions that have more
663than 3 lines of code in them. An exception to this rule are the cases where
664a parameter is known to be a compiletime constant, and as a result of this
665constantness you *know* the compiler will be able to optimize most of your
666function away at compile time. For a good example of this later case, see
667the kmalloc() inline function.
668
669Often people argue that adding inline to functions that are static and used
670only once is always a win since there is no space tradeoff. While this is
671technically correct, gcc is capable of inlining these automatically without
672help, and the maintenance issue of removing the inline when a second user
673appears outweighs the potential value of the hint that tells gcc to do
674something it would have done anyway.
675
676
Alan Sternc16a02d2006-09-29 02:01:21 -0700677 Chapter 16: Function return values and names
678
679Functions can return values of many different kinds, and one of the
680most common is a value indicating whether the function succeeded or
681failed. Such a value can be represented as an error-code integer
682(-Exxx = failure, 0 = success) or a "succeeded" boolean (0 = failure,
683non-zero = success).
684
685Mixing up these two sorts of representations is a fertile source of
686difficult-to-find bugs. If the C language included a strong distinction
687between integers and booleans then the compiler would find these mistakes
688for us... but it doesn't. To help prevent such bugs, always follow this
689convention:
690
691 If the name of a function is an action or an imperative command,
692 the function should return an error-code integer. If the name
693 is a predicate, the function should return a "succeeded" boolean.
694
695For example, "add work" is a command, and the add_work() function returns 0
696for success or -EBUSY for failure. In the same way, "PCI device present" is
697a predicate, and the pci_dev_present() function returns 1 if it succeeds in
698finding a matching device or 0 if it doesn't.
699
700All EXPORTed functions must respect this convention, and so should all
701public functions. Private (static) functions need not, but it is
702recommended that they do.
703
704Functions whose return value is the actual result of a computation, rather
705than an indication of whether the computation succeeded, are not subject to
706this rule. Generally they indicate failure by returning some out-of-range
707result. Typical examples would be functions that return pointers; they use
708NULL or the ERR_PTR mechanism to report failure.
709
710
Robert P. J. Day58637ec2006-12-22 01:09:11 -0800711 Chapter 17: Don't re-invent the kernel macros
712
713The header file include/linux/kernel.h contains a number of macros that
714you should use, rather than explicitly coding some variant of them yourself.
715For example, if you need to calculate the length of an array, take advantage
716of the macro
717
718 #define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
719
720Similarly, if you need to calculate the size of some structure member, use
721
722 #define FIELD_SIZEOF(t, f) (sizeof(((t*)0)->f))
723
724There are also min() and max() macros that do strict type checking if you
725need them. Feel free to peruse that header file to see what else is already
726defined that you shouldn't reproduce in your code.
727
728
Arjan van de Vena771f2b2006-01-08 01:05:04 -0800729
Randy Dunlap226a6b82006-06-23 02:05:58 -0700730 Appendix I: References
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731
732The C Programming Language, Second Edition
733by Brian W. Kernighan and Dennis M. Ritchie.
734Prentice Hall, Inc., 1988.
735ISBN 0-13-110362-8 (paperback), 0-13-110370-9 (hardback).
736URL: http://cm.bell-labs.com/cm/cs/cbook/
737
738The Practice of Programming
739by Brian W. Kernighan and Rob Pike.
740Addison-Wesley, Inc., 1999.
741ISBN 0-201-61586-X.
742URL: http://cm.bell-labs.com/cm/cs/tpop/
743
744GNU manuals - where in compliance with K&R and this text - for cpp, gcc,
Xose Vazquez Perez5b0ed2c2006-01-08 01:02:49 -0800745gcc internals and indent, all available from http://www.gnu.org/manual/
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746
747WG14 is the international standardization working group for the programming
Xose Vazquez Perez5b0ed2c2006-01-08 01:02:49 -0800748language C, URL: http://www.open-std.org/JTC1/SC22/WG14/
749
750Kernel CodingStyle, by greg@kroah.com at OLS 2002:
751http://www.kroah.com/linux/talks/ols_2002_kernel_codingstyle_talk/html/
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752
753--
Randy Dunlapb3fc9942006-12-10 02:18:56 -0800754Last updated on 2006-December-06.