TEXT   8

CodingStyle

Guest on 26th May 2021 07:36:38 PM

  1.  
  2.             Linux kernel coding style
  3.  
  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.
  9.  
  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.
  12.  
  13. Anyway, here goes:
  14.  
  15.  
  16.             Chapter 1: Indentation
  17.  
  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
  21. be 3.
  22.  
  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.
  27.  
  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
  32. your program.
  33.  
  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.
  36. Heed that warning.
  37.  
  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.:
  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.  
  60. Don't put multiple statements on a single line unless you have
  61. something to hide:
  62.  
  63.       if (condition) do_this;
  64.         do_something_everytime;
  65.  
  66. Don't put multiple assignments on a single line either.  Kernel coding style
  67. is super simple.  Avoid tricky expressions.
  68.  
  69. Outside of comments, documentation and except in Kconfig, spaces are never
  70. used for indentation, and the above example is deliberately broken.
  71.  
  72. Get a decent editor and don't leave whitespace at the end of lines.
  73.  
  74.  
  75.             Chapter 2: Breaking long lines and strings
  76.  
  77. Coding style is all about readability and maintainability using commonly
  78. available tools.
  79.  
  80. The limit on the length of lines is 80 columns and this is a strongly
  81. preferred limit.
  82.  
  83. Statements longer than 80 columns will be broken into sensible chunks.
  84. Descendants are always substantially shorter than the parent and are placed
  85. substantially to the right. The same applies to function headers with a long
  86. argument list. Long strings are as well broken into shorter strings. The
  87. only exception to this is where exceeding 80 columns significantly increases
  88. readability and does not hide information.
  89.  
  90. void fun(int a, int b, int c)
  91. {
  92.       if (condition)
  93.             printk(KERN_WARNING "Warning this is a long printk with "
  94.                                     "3 parameters a: %u b: %u "
  95.                                     "c: %u \n", a, b, c);
  96.       else
  97.             next_statement;
  98. }
  99.  
  100.             Chapter 3: Placing Braces and Spaces
  101.  
  102. The other issue that always comes up in C styling is the placement of
  103. braces.  Unlike the indent size, there are few technical reasons to
  104. choose one placement strategy over the other, but the preferred way, as
  105. shown to us by the prophets Kernighan and Ritchie, is to put the opening
  106. brace last on the line, and put the closing brace first, thusly:
  107.  
  108.       if (x is true) {
  109.             we do y
  110.       }
  111.  
  112. This applies to all non-function statement blocks (if, switch, for,
  113. while, do).  E.g.:
  114.  
  115.       switch (action) {
  116.       case KOBJ_ADD:
  117.             return "add";
  118.       case KOBJ_REMOVE:
  119.             return "remove";
  120.       case KOBJ_CHANGE:
  121.             return "change";
  122.       default:
  123.             return NULL;
  124.       }
  125.  
  126. However, there is one special case, namely functions: they have the
  127. opening brace at the beginning of the next line, thus:
  128.  
  129.       int function(int x)
  130.       {
  131.             body of function
  132.       }
  133.  
  134. Heretic people all over the world have claimed that this inconsistency
  135. is ...  well ...  inconsistent, but all right-thinking people know that
  136. (a) K&R are _right_ and (b) K&R are right.  Besides, functions are
  137. special anyway (you can't nest them in C).
  138.  
  139. Note that the closing brace is empty on a line of its own, _except_ in
  140. the cases where it is followed by a continuation of the same statement,
  141. ie a "while" in a do-statement or an "else" in an if-statement, like
  142. this:
  143.  
  144.       do {
  145.             body of do-loop
  146.       } while (condition);
  147.  
  148. and
  149.  
  150.       if (x == y) {
  151.             ..
  152.       } else if (x > y) {
  153.             ...
  154.       } else {
  155.             ....
  156.       }
  157.  
  158. Rationale: K&R.
  159.  
  160. Also, note that this brace-placement also minimizes the number of empty
  161. (or almost empty) lines, without any loss of readability.  Thus, as the
  162. supply of new-lines on your screen is not a renewable resource (think
  163. 25-line terminal screens here), you have more empty lines to put
  164. comments on.
  165.  
  166. Do not unnecessarily use braces where a single statement will do.
  167.  
  168. if (condition)
  169.       action();
  170.  
  171. This does not apply if one branch of a conditional statement is a single
  172. statement. Use braces in both branches.
  173.  
  174. if (condition) {
  175.       do_this();
  176.       do_that();
  177. } else {
  178.       otherwise();
  179. }
  180.  
  181.             3.1:  Spaces
  182.  
  183. Linux kernel style for use of spaces depends (mostly) on
  184. function-versus-keyword usage.  Use a space after (most) keywords.  The
  185. notable exceptions are sizeof, typeof, alignof, and __attribute__, which look
  186. somewhat like functions (and are usually used with parentheses in Linux,
  187. although they are not required in the language, as in: "sizeof info" after
  188. "struct fileinfo info;" is declared).
  189.  
  190. So use a space after these keywords:
  191.       if, switch, case, for, do, while
  192. but not with sizeof, typeof, alignof, or __attribute__.  E.g.,
  193.       s = sizeof(struct file);
  194.  
  195. Do not add spaces around (inside) parenthesized expressions.  This example is
  196. *bad*:
  197.  
  198.       s = sizeof( struct file );
  199.  
  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:
  203.  
  204.       char *linux_banner;
  205.       unsigned long long memparse(char *ptr, char **retptr);
  206.       char *match_strdup(substring_t *s);
  207.  
  208. Use one space around (on each side of) most binary and ternary operators,
  209. such as any of these:
  210.  
  211.       =  +  -  <  >  *  /  %  |  &  ^  <=  >=  ==  !=  ?  :
  212.  
  213. but no space after unary operators:
  214.       &  *  +  -  ~  !  sizeof  typeof  alignof  __attribute__  defined
  215.  
  216. no space before the postfix increment & decrement unary operators:
  217.       ++  --
  218.  
  219. no space after the prefix increment & decrement unary operators:
  220.       ++  --
  221.  
  222. and no space around the '.' and "->" structure member operators.
  223.  
  224. Do not leave trailing whitespace at the ends of lines.  Some editors with
  225. "smart" indentation will insert whitespace at the beginning of new lines as
  226. appropriate, so you can start typing the next line of code right away.
  227. However, some such editors do not remove the whitespace if you end up not
  228. putting a line of code there, such as if you leave a blank line.  As a result,
  229. you end up with lines containing trailing whitespace.
  230.  
  231. Git will warn you about patches that introduce trailing whitespace, and can
  232. optionally strip the trailing whitespace for you; however, if applying a series
  233. of patches, this may make later patches in the series fail by changing their
  234. context lines.
  235.  
  236.  
  237.             Chapter 4: Naming
  238.  
  239. C is a Spartan language, and so should your naming be.  Unlike Modula-2
  240. and Pascal programmers, C programmers do not use cute names like
  241. ThisVariableIsATemporaryCounter.  A C programmer would call that
  242. variable "tmp", which is much easier to write, and not the least more
  243. difficult to understand.
  244.  
  245. HOWEVER, while mixed-case names are frowned upon, descriptive names for
  246. global variables are a must.  To call a global function "foo" is a
  247. shooting offense.
  248.  
  249. GLOBAL variables (to be used only if you _really_ need them) need to
  250. have descriptive names, as do global functions.  If you have a function
  251. that counts the number of active users, you should call that
  252. "count_active_users()" or similar, you should _not_ call it "cntusr()".
  253.  
  254. Encoding the type of a function into the name (so-called Hungarian
  255. notation) is brain damaged - the compiler knows the types anyway and can
  256. check those, and it only confuses the programmer.  No wonder MicroSoft
  257. makes buggy programs.
  258.  
  259. LOCAL variable names should be short, and to the point.  If you have
  260. some random integer loop counter, it should probably be called "i".
  261. Calling it "loop_counter" is non-productive, if there is no chance of it
  262. being mis-understood.  Similarly, "tmp" can be just about any type of
  263. variable that is used to hold a temporary value.
  264.  
  265. If you are afraid to mix up your local variable names, you have another
  266. problem, which is called the function-growth-hormone-imbalance syndrome.
  267. See chapter 6 (Functions).
  268.  
  269.  
  270.             Chapter 5: Typedefs
  271.  
  272. Please don't use things like "vps_t".
  273.  
  274. It's a _mistake_ to use typedef for structures and pointers. When you see a
  275.  
  276.       vps_t a;
  277.  
  278. in the source, what does it mean?
  279.  
  280. In contrast, if it says
  281.  
  282.       struct virtual_container *a;
  283.  
  284. you can actually tell what "a" is.
  285.  
  286. Lots of people think that typedefs "help readability". Not so. They are
  287. useful only for:
  288.  
  289.  (a) totally opaque objects (where the typedef is actively used to _hide_
  290.      what the object is).
  291.  
  292.      Example: "pte_t" etc. opaque objects that you can only access using
  293.      the proper accessor functions.
  294.  
  295.      NOTE! Opaqueness and "accessor functions" are not good in themselves.
  296.      The reason we have them for things like pte_t etc. is that there
  297.      really is absolutely _zero_ portably accessible information there.
  298.  
  299.  (b) Clear integer types, where the abstraction _helps_ avoid confusion
  300.      whether it is "int" or "long".
  301.  
  302.      u8/u16/u32 are perfectly fine typedefs, although they fit into
  303.      category (d) better than here.
  304.  
  305.      NOTE! Again - there needs to be a _reason_ for this. If something is
  306.      "unsigned long", then there's no reason to do
  307.  
  308.       typedef unsigned long myflags_t;
  309.  
  310.      but if there is a clear reason for why it under certain circumstances
  311.      might be an "unsigned int" and under other configurations might be
  312.      "unsigned long", then by all means go ahead and use a typedef.
  313.  
  314.  (c) when you use sparse to literally create a _new_ type for
  315.      type-checking.
  316.  
  317.  (d) New types which are identical to standard C99 types, in certain
  318.      exceptional circumstances.
  319.  
  320.      Although it would only take a short amount of time for the eyes and
  321.      brain to become accustomed to the standard types like 'uint32_t',
  322.      some people object to their use anyway.
  323.  
  324.      Therefore, the Linux-specific 'u8/u16/u32/u64' types and their
  325.      signed equivalents which are identical to standard types are
  326.      permitted -- although they are not mandatory in new code of your
  327.      own.
  328.  
  329.      When editing existing code which already uses one or the other set
  330.      of types, you should conform to the existing choices in that code.
  331.  
  332.  (e) Types safe for use in userspace.
  333.  
  334.      In certain structures which are visible to userspace, we cannot
  335.      require C99 types and cannot use the 'u32' form above. Thus, we
  336.      use __u32 and similar types in all structures which are shared
  337.      with userspace.
  338.  
  339. Maybe there are other cases too, but the rule should basically be to NEVER
  340. EVER use a typedef unless you can clearly match one of those rules.
  341.  
  342. In general, a pointer, or a struct that has elements that can reasonably
  343. be directly accessed should _never_ be a typedef.
  344.  
  345.  
  346.             Chapter 6: Functions
  347.  
  348. Functions should be short and sweet, and do just one thing.  They should
  349. fit on one or two screenfuls of text (the ISO/ANSI screen size is 80x24,
  350. as we all know), and do one thing and do that well.
  351.  
  352. The maximum length of a function is inversely proportional to the
  353. complexity and indentation level of that function.  So, if you have a
  354. conceptually simple function that is just one long (but simple)
  355. case-statement, where you have to do lots of small things for a lot of
  356. different cases, it's OK to have a longer function.
  357.  
  358. However, if you have a complex function, and you suspect that a
  359. less-than-gifted first-year high-school student might not even
  360. understand what the function is all about, you should adhere to the
  361. maximum limits all the more closely.  Use helper functions with
  362. descriptive names (you can ask the compiler to in-line them if you think
  363. it's performance-critical, and it will probably do a better job of it
  364. than you would have done).
  365.  
  366. Another measure of the function is the number of local variables.  They
  367. shouldn't exceed 5-10, or you're doing something wrong.  Re-think the
  368. function, and split it into smaller pieces.  A human brain can
  369. generally easily keep track of about 7 different things, anything more
  370. and it gets confused.  You know you're brilliant, but maybe you'd like
  371. to understand what you did 2 weeks from now.
  372.  
  373. In source files, separate functions with one blank line.  If the function is
  374. exported, the EXPORT* macro for it should follow immediately after the closing
  375. function brace line.  E.g.:
  376.  
  377. int system_is_up(void)
  378. {
  379.       return system_state == SYSTEM_RUNNING;
  380. }
  381. EXPORT_SYMBOL(system_is_up);
  382.  
  383. In function prototypes, include parameter names with their data types.
  384. Although this is not required by the C language, it is preferred in Linux
  385. because it is a simple way to add valuable information for the reader.
  386.  
  387.  
  388.             Chapter 7: Centralized exiting of functions
  389.  
  390. Albeit deprecated by some people, the equivalent of the goto statement is
  391. used frequently by compilers in form of the unconditional jump instruction.
  392.  
  393. The goto statement comes in handy when a function exits from multiple
  394. locations and some common work such as cleanup has to be done.
  395.  
  396. The rationale is:
  397.  
  398. - unconditional statements are easier to understand and follow
  399. - nesting is reduced
  400. - errors by not updating individual exit points when making
  401.     modifications are prevented
  402. - saves the compiler work to optimize redundant code away ;)
  403.  
  404. int fun(int a)
  405. {
  406.       int result = 0;
  407.       char *buffer = kmalloc(SIZE);
  408.  
  409.       if (buffer == NULL)
  410.             return -ENOMEM;
  411.  
  412.       if (condition1) {
  413.             while (loop1) {
  414.                   ...
  415.             }
  416.             result = 1;
  417.             goto out;
  418.       }
  419.       ...
  420. out:
  421.       kfree(buffer);
  422.       return result;
  423. }
  424.  
  425.             Chapter 8: Commenting
  426.  
  427. Comments are good, but there is also a danger of over-commenting.  NEVER
  428. try to explain HOW your code works in a comment: it's much better to
  429. write the code so that the _working_ is obvious, and it's a waste of
  430. time to explain badly written code.
  431.  
  432. Generally, you want your comments to tell WHAT your code does, not HOW.
  433. Also, try to avoid putting comments inside a function body: if the
  434. function is so complex that you need to separately comment parts of it,
  435. you should probably go back to chapter 6 for a while.  You can make
  436. small comments to note or warn about something particularly clever (or
  437. ugly), but try to avoid excess.  Instead, put the comments at the head
  438. of the function, telling people what it does, and possibly WHY it does
  439. it.
  440.  
  441. When commenting the kernel API functions, please use the kernel-doc format.
  442. See the files Documentation/kernel-doc-nano-HOWTO.txt and scripts/kernel-doc
  443. for details.
  444.  
  445. Linux style for comments is the C89 "/* ... */" style.
  446. Don't use C99-style "// ..." comments.
  447.  
  448. The preferred style for long (multi-line) comments is:
  449.  
  450.       /*
  451.        * This is the preferred style for multi-line
  452.        * comments in the Linux kernel source code.
  453.        * Please use it consistently.
  454.        *
  455.        * Description:  A column of asterisks on the left side,
  456.        * with beginning and ending almost-blank lines.
  457.        */
  458.  
  459. It's also important to comment data, whether they are basic types or derived
  460. types.  To this end, use just one data declaration per line (no commas for
  461. multiple data declarations).  This leaves you room for a small comment on each
  462. item, explaining its use.
  463.  
  464.  
  465.             Chapter 9: You've made a mess of it
  466.  
  467. That's OK, we all do.  You've probably been told by your long-time Unix
  468. user helper that "GNU emacs" automatically formats the C sources for
  469. you, and you've noticed that yes, it does do that, but the defaults it
  470. uses are less than desirable (in fact, they are worse than random
  471. typing - an infinite number of monkeys typing into GNU emacs would never
  472. make a good program).
  473.  
  474. So, you can either get rid of GNU emacs, or change it to use saner
  475. values.  To do the latter, you can stick the following in your .emacs file:
  476.  
  477. (defun linux-c-mode ()
  478.   "C mode with adjusted defaults for use with the Linux kernel."
  479.   (interactive)
  480.   (c-mode)
  481.   (c-set-style "K&R")
  482.   (setq tab-width 8)
  483.   (setq indent-tabs-mode t)
  484.   (setq c-basic-offset 8))
  485.  
  486. This will define the M-x linux-c-mode command.  When hacking on a
  487. module, if you put the string -*- linux-c -*- somewhere on the first
  488. two lines, this mode will be automatically invoked. Also, you may want
  489. to add
  490.  
  491. (setq auto-mode-alist (cons '("/usr/src/linux.*/.*\\.[ch]$" . linux-c-mode)
  492.                   auto-mode-alist))
  493.  
  494. to your .emacs file if you want to have linux-c-mode switched on
  495. automagically when you edit source files under /usr/src/linux.
  496.  
  497. But even if you fail in getting emacs to do sane formatting, not
  498. everything is lost: use "indent".
  499.  
  500. Now, again, GNU indent has the same brain-dead settings that GNU emacs
  501. has, which is why you need to give it a few command line options.
  502. However, that's not too bad, because even the makers of GNU indent
  503. recognize the authority of K&R (the GNU people aren't evil, they are
  504. just severely misguided in this matter), so you just give indent the
  505. options "-kr -i8" (stands for "K&R, 8 character indents"), or use
  506. "scripts/Lindent", which indents in the latest style.
  507.  
  508. "indent" has a lot of options, and especially when it comes to comment
  509. re-formatting you may want to take a look at the man page.  But
  510. remember: "indent" is not a fix for bad programming.
  511.  
  512.  
  513.             Chapter 10: Kconfig configuration files
  514.  
  515. For all of the Kconfig* configuration files throughout the source tree,
  516. the indentation is somewhat different.  Lines under a "config" definition
  517. are indented with one tab, while help text is indented an additional two
  518. spaces.  Example:
  519.  
  520. config AUDIT
  521.       bool "Auditing support"
  522.       depends on NET
  523.       help
  524.         Enable auditing infrastructure that can be used with another
  525.         kernel subsystem, such as SELinux (which requires this for
  526.         logging of avc messages output).  Does not do system-call
  527.         auditing without CONFIG_AUDITSYSCALL.
  528.  
  529. Features that might still be considered unstable should be defined as
  530. dependent on "EXPERIMENTAL":
  531.  
  532. config SLUB
  533.       depends on EXPERIMENTAL && !ARCH_USES_SLAB_PAGE_STRUCT
  534.       bool "SLUB (Unqueued Allocator)"
  535.       ...
  536.  
  537. while seriously dangerous features (such as write support for certain
  538. filesystems) should advertise this prominently in their prompt string:
  539.  
  540. config ADFS_FS_RW
  541.       bool "ADFS write support (DANGEROUS)"
  542.       depends on ADFS_FS
  543.       ...
  544.  
  545. For full documentation on the configuration files, see the file
  546. Documentation/kbuild/kconfig-language.txt.
  547.  
  548.  
  549.             Chapter 11: Data structures
  550.  
  551. Data structures that have visibility outside the single-threaded
  552. environment they are created and destroyed in should always have
  553. reference counts.  In the kernel, garbage collection doesn't exist (and
  554. outside the kernel garbage collection is slow and inefficient), which
  555. means that you absolutely _have_ to reference count all your uses.
  556.  
  557. Reference counting means that you can avoid locking, and allows multiple
  558. users to have access to the data structure in parallel - and not having
  559. to worry about the structure suddenly going away from under them just
  560. because they slept or did something else for a while.
  561.  
  562. Note that locking is _not_ a replacement for reference counting.
  563. Locking is used to keep data structures coherent, while reference
  564. counting is a memory management technique.  Usually both are needed, and
  565. they are not to be confused with each other.
  566.  
  567. Many data structures can indeed have two levels of reference counting,
  568. when there are users of different "classes".  The subclass count counts
  569. the number of subclass users, and decrements the global count just once
  570. when the subclass count goes to zero.
  571.  
  572. Examples of this kind of "multi-level-reference-counting" can be found in
  573. memory management ("struct mm_struct": mm_users and mm_count), and in
  574. filesystem code ("struct super_block": s_count and s_active).
  575.  
  576. Remember: if another thread can find your data structure, and you don't
  577. have a reference count on it, you almost certainly have a bug.
  578.  
  579.  
  580.             Chapter 12: Macros, Enums and RTL
  581.  
  582. Names of macros defining constants and labels in enums are capitalized.
  583.  
  584. #define CONSTANT 0x12345
  585.  
  586. Enums are preferred when defining several related constants.
  587.  
  588. CAPITALIZED macro names are appreciated but macros resembling functions
  589. may be named in lower case.
  590.  
  591. Generally, inline functions are preferable to macros resembling functions.
  592.  
  593. Macros with multiple statements should be enclosed in a do - while block:
  594.  
  595. #define macrofun(a, b, c)                 \
  596.       do {                          \
  597.             if (a == 5)             \
  598.                   do_this(b, c);          \
  599.       } while (0)
  600.  
  601. Things to avoid when using macros:
  602.  
  603. 1) macros that affect control flow:
  604.  
  605. #define FOO(x)                            \
  606.       do {                          \
  607.             if (blah(x) < 0)        \
  608.                   return -EBUGGERED;      \
  609.       } while(0)
  610.  
  611. is a _very_ bad idea.  It looks like a function call but exits the "calling"
  612. function; don't break the internal parsers of those who will read the code.
  613.  
  614. 2) macros that depend on having a local variable with a magic name:
  615.  
  616. #define FOO(val) bar(index, val)
  617.  
  618. might look like a good thing, but it's confusing as hell when one reads the
  619. code and it's prone to breakage from seemingly innocent changes.
  620.  
  621. 3) macros with arguments that are used as l-values: FOO(x) = y; will
  622. bite you if somebody e.g. turns FOO into an inline function.
  623.  
  624. 4) forgetting about precedence: macros defining constants using expressions
  625. must enclose the expression in parentheses. Beware of similar issues with
  626. macros using parameters.
  627.  
  628. #define CONSTANT 0x4000
  629. #define CONSTEXP (CONSTANT | 3)
  630.  
  631. The cpp manual deals with macros exhaustively. The gcc internals manual also
  632. covers RTL which is used frequently with assembly language in the kernel.
  633.  
  634.  
  635.             Chapter 13: Printing kernel messages
  636.  
  637. Kernel developers like to be seen as literate. Do mind the spelling
  638. of kernel messages to make a good impression. Do not use crippled
  639. words like "dont"; use "do not" or "don't" instead.  Make the messages
  640. concise, clear, and unambiguous.
  641.  
  642. Kernel messages do not have to be terminated with a period.
  643.  
  644. Printing numbers in parentheses (%d) adds no value and should be avoided.
  645.  
  646. There are a number of driver model diagnostic macros in <linux/device.h>
  647. which you should use to make sure messages are matched to the right device
  648. and driver, and are tagged with the right level:  dev_err(), dev_warn(),
  649. dev_info(), and so forth.  For messages that aren't associated with a
  650. particular device, <linux/kernel.h> defines pr_debug() and pr_info().
  651.  
  652. Coming up with good debugging messages can be quite a challenge; and once
  653. you have them, they can be a huge help for remote troubleshooting.  Such
  654. messages should be compiled out when the DEBUG symbol is not defined (that
  655. is, by default they are not included).  When you use dev_dbg() or pr_debug(),
  656. that's automatic.  Many subsystems have Kconfig options to turn on -DDEBUG.
  657. A related convention uses VERBOSE_DEBUG to add dev_vdbg() messages to the
  658. ones already enabled by DEBUG.
  659.  
  660.  
  661.             Chapter 14: Allocating memory
  662.  
  663. The kernel provides the following general purpose memory allocators:
  664. kmalloc(), kzalloc(), kcalloc(), and vmalloc().  Please refer to the API
  665. documentation for further information about them.
  666.  
  667. The preferred form for passing a size of a struct is the following:
  668.  
  669.       p = kmalloc(sizeof(*p), ...);
  670.  
  671. The alternative form where struct name is spelled out hurts readability and
  672. introduces an opportunity for a bug when the pointer variable type is changed
  673. but the corresponding sizeof that is passed to a memory allocator is not.
  674.  
  675. Casting the return value which is a void pointer is redundant. The conversion
  676. from void pointer to any other pointer type is guaranteed by the C programming
  677. language.
  678.  
  679.  
  680.             Chapter 15: The inline disease
  681.  
  682. There appears to be a common misperception that gcc has a magic "make me
  683. faster" speedup option called "inline". While the use of inlines can be
  684. appropriate (for example as a means of replacing macros, see Chapter 12), it
  685. very often is not. Abundant use of the inline keyword leads to a much bigger
  686. kernel, which in turn slows the system as a whole down, due to a bigger
  687. icache footprint for the CPU and simply because there is less memory
  688. available for the pagecache. Just think about it; a pagecache miss causes a
  689. disk seek, which easily takes 5 miliseconds. There are a LOT of cpu cycles
  690. that can go into these 5 miliseconds.
  691.  
  692. A reasonable rule of thumb is to not put inline at functions that have more
  693. than 3 lines of code in them. An exception to this rule are the cases where
  694. a parameter is known to be a compiletime constant, and as a result of this
  695. constantness you *know* the compiler will be able to optimize most of your
  696. function away at compile time. For a good example of this later case, see
  697. the kmalloc() inline function.
  698.  
  699. Often people argue that adding inline to functions that are static and used
  700. only once is always a win since there is no space tradeoff. While this is
  701. technically correct, gcc is capable of inlining these automatically without
  702. help, and the maintenance issue of removing the inline when a second user
  703. appears outweighs the potential value of the hint that tells gcc to do
  704. something it would have done anyway.
  705.  
  706.  
  707.             Chapter 16: Function return values and names
  708.  
  709. Functions can return values of many different kinds, and one of the
  710. most common is a value indicating whether the function succeeded or
  711. failed.  Such a value can be represented as an error-code integer
  712. (-Exxx = failure, 0 = success) or a "succeeded" boolean (0 = failure,
  713. non-zero = success).
  714.  
  715. Mixing up these two sorts of representations is a fertile source of
  716. difficult-to-find bugs.  If the C language included a strong distinction
  717. between integers and booleans then the compiler would find these mistakes
  718. for us... but it doesn't.  To help prevent such bugs, always follow this
  719. convention:
  720.  
  721.       If the name of a function is an action or an imperative command,
  722.       the function should return an error-code integer.  If the name
  723.       is a predicate, the function should return a "succeeded" boolean.
  724.  
  725. For example, "add work" is a command, and the add_work() function returns 0
  726. for success or -EBUSY for failure.  In the same way, "PCI device present" is
  727. a predicate, and the pci_dev_present() function returns 1 if it succeeds in
  728. finding a matching device or 0 if it doesn't.
  729.  
  730. All EXPORTed functions must respect this convention, and so should all
  731. public functions.  Private (static) functions need not, but it is
  732. recommended that they do.
  733.  
  734. Functions whose return value is the actual result of a computation, rather
  735. than an indication of whether the computation succeeded, are not subject to
  736. this rule.  Generally they indicate failure by returning some out-of-range
  737. result.  Typical examples would be functions that return pointers; they use
  738. NULL or the ERR_PTR mechanism to report failure.
  739.  
  740.  
  741.             Chapter 17:  Don't re-invent the kernel macros
  742.  
  743. The header file include/linux/kernel.h contains a number of macros that
  744. you should use, rather than explicitly coding some variant of them yourself.
  745. For example, if you need to calculate the length of an array, take advantage
  746. of the macro
  747.  
  748.   #define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
  749.  
  750. Similarly, if you need to calculate the size of some structure member, use
  751.  
  752.   #define FIELD_SIZEOF(t, f) (sizeof(((t*)0)->f))
  753.  
  754. There are also min() and max() macros that do strict type checking if you
  755. need them.  Feel free to peruse that header file to see what else is already
  756. defined that you shouldn't reproduce in your code.
  757.  
  758.  
  759.             Chapter 18:  Editor modelines and other cruft
  760.  
  761. Some editors can interpret configuration information embedded in source files,
  762. indicated with special markers.  For example, emacs interprets lines marked
  763. like this:
  764.  
  765. -*- mode: c -*-
  766.  
  767. Or like this:
  768.  
  769. /*
  770. Local Variables:
  771. compile-command: "gcc -DMAGIC_DEBUG_FLAG foo.c"
  772. End:
  773. */
  774.  
  775. Vim interprets markers that look like this:
  776.  
  777. /* vim:set sw=8 noet */
  778.  
  779. Do not include any of these in source files.  People have their own personal
  780. editor configurations, and your source files should not override them.  This
  781. includes markers for indentation and mode configuration.  People may use their
  782. own custom mode, or may have some other magic method for making indentation
  783. work correctly.
  784.  
  785.  
  786.  
  787.             Appendix I: References
  788.  
  789. The C Programming Language, Second Edition
  790. by Brian W. Kernighan and Dennis M. Ritchie.
  791. Prentice Hall, Inc., 1988.
  792. ISBN 0-13-110362-8 (paperback), 0-13-110370-9 (hardback).
  793. URL: http://cm.bell-labs.com/cm/cs/cbook/
  794.  
  795. The Practice of Programming
  796. by Brian W. Kernighan and Rob Pike.
  797. Addison-Wesley, Inc., 1999.
  798. ISBN 0-201-61586-X.
  799. URL: http://cm.bell-labs.com/cm/cs/tpop/
  800.  
  801. GNU manuals - where in compliance with K&R and this text - for cpp, gcc,
  802. gcc internals and indent, all available from http://www.gnu.org/manual/
  803.  
  804. WG14 is the international standardization working group for the programming
  805. language C, URL: http://www.open-std.org/JTC1/SC22/WG14/
  806.  
  807. Kernel CodingStyle, by greg@kroah.com at OLS 2002:
  808. http://www.kroah.com/linux/talks/ols_2002_kernel_codingstyle_talk/html/
  809.  
  810. --
  811. Last updated on 2007-July-13.

Raw Paste


Login or Register to edit or fork this paste. It's free.