Loading...
Note: File does not exist in v4.10.11.
1.. SPDX-License-Identifier: GPL-2.0-only
2
3==========
4Checkpatch
5==========
6
7Checkpatch (scripts/checkpatch.pl) is a perl script which checks for trivial
8style violations in patches and optionally corrects them. Checkpatch can
9also be run on file contexts and without the kernel tree.
10
11Checkpatch is not always right. Your judgement takes precedence over checkpatch
12messages. If your code looks better with the violations, then its probably
13best left alone.
14
15
16Options
17=======
18
19This section will describe the options checkpatch can be run with.
20
21Usage::
22
23 ./scripts/checkpatch.pl [OPTION]... [FILE]...
24
25Available options:
26
27 - -q, --quiet
28
29 Enable quiet mode.
30
31 - -v, --verbose
32 Enable verbose mode. Additional verbose test descriptions are output
33 so as to provide information on why that particular message is shown.
34
35 - --no-tree
36
37 Run checkpatch without the kernel tree.
38
39 - --no-signoff
40
41 Disable the 'Signed-off-by' line check. The sign-off is a simple line at
42 the end of the explanation for the patch, which certifies that you wrote it
43 or otherwise have the right to pass it on as an open-source patch.
44
45 Example::
46
47 Signed-off-by: Random J Developer <random@developer.example.org>
48
49 Setting this flag effectively stops a message for a missing signed-off-by
50 line in a patch context.
51
52 - --patch
53
54 Treat FILE as a patch. This is the default option and need not be
55 explicitly specified.
56
57 - --emacs
58
59 Set output to emacs compile window format. This allows emacs users to jump
60 from the error in the compile window directly to the offending line in the
61 patch.
62
63 - --terse
64
65 Output only one line per report.
66
67 - --showfile
68
69 Show the diffed file position instead of the input file position.
70
71 - -g, --git
72
73 Treat FILE as a single commit or a git revision range.
74
75 Single commit with:
76
77 - <rev>
78 - <rev>^
79 - <rev>~n
80
81 Multiple commits with:
82
83 - <rev1>..<rev2>
84 - <rev1>...<rev2>
85 - <rev>-<count>
86
87 - -f, --file
88
89 Treat FILE as a regular source file. This option must be used when running
90 checkpatch on source files in the kernel.
91
92 - --subjective, --strict
93
94 Enable stricter tests in checkpatch. By default the tests emitted as CHECK
95 do not activate by default. Use this flag to activate the CHECK tests.
96
97 - --list-types
98
99 Every message emitted by checkpatch has an associated TYPE. Add this flag
100 to display all the types in checkpatch.
101
102 Note that when this flag is active, checkpatch does not read the input FILE,
103 and no message is emitted. Only a list of types in checkpatch is output.
104
105 - --types TYPE(,TYPE2...)
106
107 Only display messages with the given types.
108
109 Example::
110
111 ./scripts/checkpatch.pl mypatch.patch --types EMAIL_SUBJECT,BRACES
112
113 - --ignore TYPE(,TYPE2...)
114
115 Checkpatch will not emit messages for the specified types.
116
117 Example::
118
119 ./scripts/checkpatch.pl mypatch.patch --ignore EMAIL_SUBJECT,BRACES
120
121 - --show-types
122
123 By default checkpatch doesn't display the type associated with the messages.
124 Set this flag to show the message type in the output.
125
126 - --max-line-length=n
127
128 Set the max line length (default 100). If a line exceeds the specified
129 length, a LONG_LINE message is emitted.
130
131
132 The message level is different for patch and file contexts. For patches,
133 a WARNING is emitted. While a milder CHECK is emitted for files. So for
134 file contexts, the --strict flag must also be enabled.
135
136 - --min-conf-desc-length=n
137
138 Set the Kconfig entry minimum description length, if shorter, warn.
139
140 - --tab-size=n
141
142 Set the number of spaces for tab (default 8).
143
144 - --root=PATH
145
146 PATH to the kernel tree root.
147
148 This option must be specified when invoking checkpatch from outside
149 the kernel root.
150
151 - --no-summary
152
153 Suppress the per file summary.
154
155 - --mailback
156
157 Only produce a report in case of Warnings or Errors. Milder Checks are
158 excluded from this.
159
160 - --summary-file
161
162 Include the filename in summary.
163
164 - --debug KEY=[0|1]
165
166 Turn on/off debugging of KEY, where KEY is one of 'values', 'possible',
167 'type', and 'attr' (default is all off).
168
169 - --fix
170
171 This is an EXPERIMENTAL feature. If correctable errors exist, a file
172 <inputfile>.EXPERIMENTAL-checkpatch-fixes is created which has the
173 automatically fixable errors corrected.
174
175 - --fix-inplace
176
177 EXPERIMENTAL - Similar to --fix but input file is overwritten with fixes.
178
179 DO NOT USE this flag unless you are absolutely sure and you have a backup
180 in place.
181
182 - --ignore-perl-version
183
184 Override checking of perl version. Runtime errors may be encountered after
185 enabling this flag if the perl version does not meet the minimum specified.
186
187 - --codespell
188
189 Use the codespell dictionary for checking spelling errors.
190
191 - --codespellfile
192
193 Use the specified codespell file.
194 Default is '/usr/share/codespell/dictionary.txt'.
195
196 - --typedefsfile
197
198 Read additional types from this file.
199
200 - --color[=WHEN]
201
202 Use colors 'always', 'never', or only when output is a terminal ('auto').
203 Default is 'auto'.
204
205 - --kconfig-prefix=WORD
206
207 Use WORD as a prefix for Kconfig symbols (default is `CONFIG_`).
208
209 - -h, --help, --version
210
211 Display the help text.
212
213Message Levels
214==============
215
216Messages in checkpatch are divided into three levels. The levels of messages
217in checkpatch denote the severity of the error. They are:
218
219 - ERROR
220
221 This is the most strict level. Messages of type ERROR must be taken
222 seriously as they denote things that are very likely to be wrong.
223
224 - WARNING
225
226 This is the next stricter level. Messages of type WARNING requires a
227 more careful review. But it is milder than an ERROR.
228
229 - CHECK
230
231 This is the mildest level. These are things which may require some thought.
232
233Type Descriptions
234=================
235
236This section contains a description of all the message types in checkpatch.
237
238.. Types in this section are also parsed by checkpatch.
239.. The types are grouped into subsections based on use.
240
241
242Allocation style
243----------------
244
245 **ALLOC_ARRAY_ARGS**
246 The first argument for kcalloc or kmalloc_array should be the
247 number of elements. sizeof() as the first argument is generally
248 wrong.
249
250 See: https://www.kernel.org/doc/html/latest/core-api/memory-allocation.html
251
252 **ALLOC_SIZEOF_STRUCT**
253 The allocation style is bad. In general for family of
254 allocation functions using sizeof() to get memory size,
255 constructs like::
256
257 p = alloc(sizeof(struct foo), ...)
258
259 should be::
260
261 p = alloc(sizeof(*p), ...)
262
263 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#allocating-memory
264
265 **ALLOC_WITH_MULTIPLY**
266 Prefer kmalloc_array/kcalloc over kmalloc/kzalloc with a
267 sizeof multiply.
268
269 See: https://www.kernel.org/doc/html/latest/core-api/memory-allocation.html
270
271
272API usage
273---------
274
275 **ARCH_DEFINES**
276 Architecture specific defines should be avoided wherever
277 possible.
278
279 **ARCH_INCLUDE_LINUX**
280 Whenever asm/file.h is included and linux/file.h exists, a
281 conversion can be made when linux/file.h includes asm/file.h.
282 However this is not always the case (See signal.h).
283 This message type is emitted only for includes from arch/.
284
285 **AVOID_BUG**
286 BUG() or BUG_ON() should be avoided totally.
287 Use WARN() and WARN_ON() instead, and handle the "impossible"
288 error condition as gracefully as possible.
289
290 See: https://www.kernel.org/doc/html/latest/process/deprecated.html#bug-and-bug-on
291
292 **CONSIDER_KSTRTO**
293 The simple_strtol(), simple_strtoll(), simple_strtoul(), and
294 simple_strtoull() functions explicitly ignore overflows, which
295 may lead to unexpected results in callers. The respective kstrtol(),
296 kstrtoll(), kstrtoul(), and kstrtoull() functions tend to be the
297 correct replacements.
298
299 See: https://www.kernel.org/doc/html/latest/process/deprecated.html#simple-strtol-simple-strtoll-simple-strtoul-simple-strtoull
300
301 **CONSTANT_CONVERSION**
302 Use of __constant_<foo> form is discouraged for the following functions::
303
304 __constant_cpu_to_be[x]
305 __constant_cpu_to_le[x]
306 __constant_be[x]_to_cpu
307 __constant_le[x]_to_cpu
308 __constant_htons
309 __constant_ntohs
310
311 Using any of these outside of include/uapi/ is not preferred as using the
312 function without __constant_ is identical when the argument is a
313 constant.
314
315 In big endian systems, the macros like __constant_cpu_to_be32(x) and
316 cpu_to_be32(x) expand to the same expression::
317
318 #define __constant_cpu_to_be32(x) ((__force __be32)(__u32)(x))
319 #define __cpu_to_be32(x) ((__force __be32)(__u32)(x))
320
321 In little endian systems, the macros __constant_cpu_to_be32(x) and
322 cpu_to_be32(x) expand to __constant_swab32 and __swab32. __swab32
323 has a __builtin_constant_p check::
324
325 #define __swab32(x) \
326 (__builtin_constant_p((__u32)(x)) ? \
327 ___constant_swab32(x) : \
328 __fswab32(x))
329
330 So ultimately they have a special case for constants.
331 Similar is the case with all of the macros in the list. Thus
332 using the __constant_... forms are unnecessarily verbose and
333 not preferred outside of include/uapi.
334
335 See: https://lore.kernel.org/lkml/1400106425.12666.6.camel@joe-AO725/
336
337 **DEPRECATED_API**
338 Usage of a deprecated RCU API is detected. It is recommended to replace
339 old flavourful RCU APIs by their new vanilla-RCU counterparts.
340
341 The full list of available RCU APIs can be viewed from the kernel docs.
342
343 See: https://www.kernel.org/doc/html/latest/RCU/whatisRCU.html#full-list-of-rcu-apis
344
345 **DEPRECATED_VARIABLE**
346 EXTRA_{A,C,CPP,LD}FLAGS are deprecated and should be replaced by the new
347 flags added via commit f77bf01425b1 ("kbuild: introduce ccflags-y,
348 asflags-y and ldflags-y").
349
350 The following conversion scheme maybe used::
351
352 EXTRA_AFLAGS -> asflags-y
353 EXTRA_CFLAGS -> ccflags-y
354 EXTRA_CPPFLAGS -> cppflags-y
355 EXTRA_LDFLAGS -> ldflags-y
356
357 See:
358
359 1. https://lore.kernel.org/lkml/20070930191054.GA15876@uranus.ravnborg.org/
360 2. https://lore.kernel.org/lkml/1313384834-24433-12-git-send-email-lacombar@gmail.com/
361 3. https://www.kernel.org/doc/html/latest/kbuild/makefiles.html#compilation-flags
362
363 **DEVICE_ATTR_FUNCTIONS**
364 The function names used in DEVICE_ATTR is unusual.
365 Typically, the store and show functions are used with <attr>_store and
366 <attr>_show, where <attr> is a named attribute variable of the device.
367
368 Consider the following examples::
369
370 static DEVICE_ATTR(type, 0444, type_show, NULL);
371 static DEVICE_ATTR(power, 0644, power_show, power_store);
372
373 The function names should preferably follow the above pattern.
374
375 See: https://www.kernel.org/doc/html/latest/driver-api/driver-model/device.html#attributes
376
377 **DEVICE_ATTR_RO**
378 The DEVICE_ATTR_RO(name) helper macro can be used instead of
379 DEVICE_ATTR(name, 0444, name_show, NULL);
380
381 Note that the macro automatically appends _show to the named
382 attribute variable of the device for the show method.
383
384 See: https://www.kernel.org/doc/html/latest/driver-api/driver-model/device.html#attributes
385
386 **DEVICE_ATTR_RW**
387 The DEVICE_ATTR_RW(name) helper macro can be used instead of
388 DEVICE_ATTR(name, 0644, name_show, name_store);
389
390 Note that the macro automatically appends _show and _store to the
391 named attribute variable of the device for the show and store methods.
392
393 See: https://www.kernel.org/doc/html/latest/driver-api/driver-model/device.html#attributes
394
395 **DEVICE_ATTR_WO**
396 The DEVICE_AATR_WO(name) helper macro can be used instead of
397 DEVICE_ATTR(name, 0200, NULL, name_store);
398
399 Note that the macro automatically appends _store to the
400 named attribute variable of the device for the store method.
401
402 See: https://www.kernel.org/doc/html/latest/driver-api/driver-model/device.html#attributes
403
404 **DUPLICATED_SYSCTL_CONST**
405 Commit d91bff3011cf ("proc/sysctl: add shared variables for range
406 check") added some shared const variables to be used instead of a local
407 copy in each source file.
408
409 Consider replacing the sysctl range checking value with the shared
410 one in include/linux/sysctl.h. The following conversion scheme may
411 be used::
412
413 &zero -> SYSCTL_ZERO
414 &one -> SYSCTL_ONE
415 &int_max -> SYSCTL_INT_MAX
416
417 See:
418
419 1. https://lore.kernel.org/lkml/20190430180111.10688-1-mcroce@redhat.com/
420 2. https://lore.kernel.org/lkml/20190531131422.14970-1-mcroce@redhat.com/
421
422 **ENOSYS**
423 ENOSYS means that a nonexistent system call was called.
424 Earlier, it was wrongly used for things like invalid operations on
425 otherwise valid syscalls. This should be avoided in new code.
426
427 See: https://lore.kernel.org/lkml/5eb299021dec23c1a48fa7d9f2c8b794e967766d.1408730669.git.luto@amacapital.net/
428
429 **ENOTSUPP**
430 ENOTSUPP is not a standard error code and should be avoided in new patches.
431 EOPNOTSUPP should be used instead.
432
433 See: https://lore.kernel.org/netdev/20200510182252.GA411829@lunn.ch/
434
435 **EXPORT_SYMBOL**
436 EXPORT_SYMBOL should immediately follow the symbol to be exported.
437
438 **IN_ATOMIC**
439 in_atomic() is not for driver use so any such use is reported as an ERROR.
440 Also in_atomic() is often used to determine if sleeping is permitted,
441 but it is not reliable in this use model. Therefore its use is
442 strongly discouraged.
443
444 However, in_atomic() is ok for core kernel use.
445
446 See: https://lore.kernel.org/lkml/20080320201723.b87b3732.akpm@linux-foundation.org/
447
448 **LOCKDEP**
449 The lockdep_no_validate class was added as a temporary measure to
450 prevent warnings on conversion of device->sem to device->mutex.
451 It should not be used for any other purpose.
452
453 See: https://lore.kernel.org/lkml/1268959062.9440.467.camel@laptop/
454
455 **MALFORMED_INCLUDE**
456 The #include statement has a malformed path. This has happened
457 because the author has included a double slash "//" in the pathname
458 accidentally.
459
460 **USE_LOCKDEP**
461 lockdep_assert_held() annotations should be preferred over
462 assertions based on spin_is_locked()
463
464 See: https://www.kernel.org/doc/html/latest/locking/lockdep-design.html#annotations
465
466 **UAPI_INCLUDE**
467 No #include statements in include/uapi should use a uapi/ path.
468
469 **USLEEP_RANGE**
470 usleep_range() should be preferred over udelay(). The proper way of
471 using usleep_range() is mentioned in the kernel docs.
472
473
474Comments
475--------
476
477 **BLOCK_COMMENT_STYLE**
478 The comment style is incorrect. The preferred style for multi-
479 line comments is::
480
481 /*
482 * This is the preferred style
483 * for multi line comments.
484 */
485
486 The networking comment style is a bit different, with the first line
487 not empty like the former::
488
489 /* This is the preferred comment style
490 * for files in net/ and drivers/net/
491 */
492
493 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#commenting
494
495 **C99_COMMENTS**
496 C99 style single line comments (//) should not be used.
497 Prefer the block comment style instead.
498
499 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#commenting
500
501 **DATA_RACE**
502 Applications of data_race() should have a comment so as to document the
503 reasoning behind why it was deemed safe.
504
505 See: https://lore.kernel.org/lkml/20200401101714.44781-1-elver@google.com/
506
507 **FSF_MAILING_ADDRESS**
508 Kernel maintainers reject new instances of the GPL boilerplate paragraph
509 directing people to write to the FSF for a copy of the GPL, since the
510 FSF has moved in the past and may do so again.
511 So do not write paragraphs about writing to the Free Software Foundation's
512 mailing address.
513
514 See: https://lore.kernel.org/lkml/20131006222342.GT19510@leaf/
515
516
517Commit message
518--------------
519
520 **BAD_SIGN_OFF**
521 The signed-off-by line does not fall in line with the standards
522 specified by the community.
523
524 See: https://www.kernel.org/doc/html/latest/process/submitting-patches.html#developer-s-certificate-of-origin-1-1
525
526 **BAD_STABLE_ADDRESS_STYLE**
527 The email format for stable is incorrect.
528 Some valid options for stable address are::
529
530 1. stable@vger.kernel.org
531 2. stable@kernel.org
532
533 For adding version info, the following comment style should be used::
534
535 stable@vger.kernel.org # version info
536
537 **COMMIT_COMMENT_SYMBOL**
538 Commit log lines starting with a '#' are ignored by git as
539 comments. To solve this problem addition of a single space
540 infront of the log line is enough.
541
542 **COMMIT_MESSAGE**
543 The patch is missing a commit description. A brief
544 description of the changes made by the patch should be added.
545
546 See: https://www.kernel.org/doc/html/latest/process/submitting-patches.html#describe-your-changes
547
548 **EMAIL_SUBJECT**
549 Naming the tool that found the issue is not very useful in the
550 subject line. A good subject line summarizes the change that
551 the patch brings.
552
553 See: https://www.kernel.org/doc/html/latest/process/submitting-patches.html#describe-your-changes
554
555 **FROM_SIGN_OFF_MISMATCH**
556 The author's email does not match with that in the Signed-off-by:
557 line(s). This can be sometimes caused due to an improperly configured
558 email client.
559
560 This message is emitted due to any of the following reasons::
561
562 - The email names do not match.
563 - The email addresses do not match.
564 - The email subaddresses do not match.
565 - The email comments do not match.
566
567 **MISSING_SIGN_OFF**
568 The patch is missing a Signed-off-by line. A signed-off-by
569 line should be added according to Developer's certificate of
570 Origin.
571
572 See: https://www.kernel.org/doc/html/latest/process/submitting-patches.html#sign-your-work-the-developer-s-certificate-of-origin
573
574 **NO_AUTHOR_SIGN_OFF**
575 The author of the patch has not signed off the patch. It is
576 required that a simple sign off line should be present at the
577 end of explanation of the patch to denote that the author has
578 written it or otherwise has the rights to pass it on as an open
579 source patch.
580
581 See: https://www.kernel.org/doc/html/latest/process/submitting-patches.html#sign-your-work-the-developer-s-certificate-of-origin
582
583 **DIFF_IN_COMMIT_MSG**
584 Avoid having diff content in commit message.
585 This causes problems when one tries to apply a file containing both
586 the changelog and the diff because patch(1) tries to apply the diff
587 which it found in the changelog.
588
589 See: https://lore.kernel.org/lkml/20150611134006.9df79a893e3636019ad2759e@linux-foundation.org/
590
591 **GERRIT_CHANGE_ID**
592 To be picked up by gerrit, the footer of the commit message might
593 have a Change-Id like::
594
595 Change-Id: Ic8aaa0728a43936cd4c6e1ed590e01ba8f0fbf5b
596 Signed-off-by: A. U. Thor <author@example.com>
597
598 The Change-Id line must be removed before submitting.
599
600 **GIT_COMMIT_ID**
601 The proper way to reference a commit id is:
602 commit <12+ chars of sha1> ("<title line>")
603
604 An example may be::
605
606 Commit e21d2170f36602ae2708 ("video: remove unnecessary
607 platform_set_drvdata()") removed the unnecessary
608 platform_set_drvdata(), but left the variable "dev" unused,
609 delete it.
610
611 See: https://www.kernel.org/doc/html/latest/process/submitting-patches.html#describe-your-changes
612
613 **BAD_FIXES_TAG**
614 The Fixes: tag is malformed or does not follow the community conventions.
615 This can occur if the tag have been split into multiple lines (e.g., when
616 pasted in an email program with word wrapping enabled).
617
618 See: https://www.kernel.org/doc/html/latest/process/submitting-patches.html#describe-your-changes
619
620
621Comparison style
622----------------
623
624 **ASSIGN_IN_IF**
625 Do not use assignments in if condition.
626 Example::
627
628 if ((foo = bar(...)) < BAZ) {
629
630 should be written as::
631
632 foo = bar(...);
633 if (foo < BAZ) {
634
635 **BOOL_COMPARISON**
636 Comparisons of A to true and false are better written
637 as A and !A.
638
639 See: https://lore.kernel.org/lkml/1365563834.27174.12.camel@joe-AO722/
640
641 **COMPARISON_TO_NULL**
642 Comparisons to NULL in the form (foo == NULL) or (foo != NULL)
643 are better written as (!foo) and (foo).
644
645 **CONSTANT_COMPARISON**
646 Comparisons with a constant or upper case identifier on the left
647 side of the test should be avoided.
648
649
650Indentation and Line Breaks
651---------------------------
652
653 **CODE_INDENT**
654 Code indent should use tabs instead of spaces.
655 Outside of comments, documentation and Kconfig,
656 spaces are never used for indentation.
657
658 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#indentation
659
660 **DEEP_INDENTATION**
661 Indentation with 6 or more tabs usually indicate overly indented
662 code.
663
664 It is suggested to refactor excessive indentation of
665 if/else/for/do/while/switch statements.
666
667 See: https://lore.kernel.org/lkml/1328311239.21255.24.camel@joe2Laptop/
668
669 **SWITCH_CASE_INDENT_LEVEL**
670 switch should be at the same indent as case.
671 Example::
672
673 switch (suffix) {
674 case 'G':
675 case 'g':
676 mem <<= 30;
677 break;
678 case 'M':
679 case 'm':
680 mem <<= 20;
681 break;
682 case 'K':
683 case 'k':
684 mem <<= 10;
685 fallthrough;
686 default:
687 break;
688 }
689
690 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#indentation
691
692 **LONG_LINE**
693 The line has exceeded the specified maximum length.
694 To use a different maximum line length, the --max-line-length=n option
695 may be added while invoking checkpatch.
696
697 Earlier, the default line length was 80 columns. Commit bdc48fa11e46
698 ("checkpatch/coding-style: deprecate 80-column warning") increased the
699 limit to 100 columns. This is not a hard limit either and it's
700 preferable to stay within 80 columns whenever possible.
701
702 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#breaking-long-lines-and-strings
703
704 **LONG_LINE_STRING**
705 A string starts before but extends beyond the maximum line length.
706 To use a different maximum line length, the --max-line-length=n option
707 may be added while invoking checkpatch.
708
709 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#breaking-long-lines-and-strings
710
711 **LONG_LINE_COMMENT**
712 A comment starts before but extends beyond the maximum line length.
713 To use a different maximum line length, the --max-line-length=n option
714 may be added while invoking checkpatch.
715
716 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#breaking-long-lines-and-strings
717
718 **SPLIT_STRING**
719 Quoted strings that appear as messages in userspace and can be
720 grepped, should not be split across multiple lines.
721
722 See: https://lore.kernel.org/lkml/20120203052727.GA15035@leaf/
723
724 **MULTILINE_DEREFERENCE**
725 A single dereferencing identifier spanned on multiple lines like::
726
727 struct_identifier->member[index].
728 member = <foo>;
729
730 is generally hard to follow. It can easily lead to typos and so makes
731 the code vulnerable to bugs.
732
733 If fixing the multiple line dereferencing leads to an 80 column
734 violation, then either rewrite the code in a more simple way or if the
735 starting part of the dereferencing identifier is the same and used at
736 multiple places then store it in a temporary variable, and use that
737 temporary variable only at all the places. For example, if there are
738 two dereferencing identifiers::
739
740 member1->member2->member3.foo1;
741 member1->member2->member3.foo2;
742
743 then store the member1->member2->member3 part in a temporary variable.
744 It not only helps to avoid the 80 column violation but also reduces
745 the program size by removing the unnecessary dereferences.
746
747 But if none of the above methods work then ignore the 80 column
748 violation because it is much easier to read a dereferencing identifier
749 on a single line.
750
751 **TRAILING_STATEMENTS**
752 Trailing statements (for example after any conditional) should be
753 on the next line.
754 Statements, such as::
755
756 if (x == y) break;
757
758 should be::
759
760 if (x == y)
761 break;
762
763
764Macros, Attributes and Symbols
765------------------------------
766
767 **ARRAY_SIZE**
768 The ARRAY_SIZE(foo) macro should be preferred over
769 sizeof(foo)/sizeof(foo[0]) for finding number of elements in an
770 array.
771
772 The macro is defined in include/linux/kernel.h::
773
774 #define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
775
776 **AVOID_EXTERNS**
777 Function prototypes don't need to be declared extern in .h
778 files. It's assumed by the compiler and is unnecessary.
779
780 **AVOID_L_PREFIX**
781 Local symbol names that are prefixed with `.L` should be avoided,
782 as this has special meaning for the assembler; a symbol entry will
783 not be emitted into the symbol table. This can prevent `objtool`
784 from generating correct unwind info.
785
786 Symbols with STB_LOCAL binding may still be used, and `.L` prefixed
787 local symbol names are still generally usable within a function,
788 but `.L` prefixed local symbol names should not be used to denote
789 the beginning or end of code regions via
790 `SYM_CODE_START_LOCAL`/`SYM_CODE_END`
791
792 **BIT_MACRO**
793 Defines like: 1 << <digit> could be BIT(digit).
794 The BIT() macro is defined via include/linux/bits.h::
795
796 #define BIT(nr) (1UL << (nr))
797
798 **CONST_READ_MOSTLY**
799 When a variable is tagged with the __read_mostly annotation, it is a
800 signal to the compiler that accesses to the variable will be mostly
801 reads and rarely(but NOT never) a write.
802
803 const __read_mostly does not make any sense as const data is already
804 read-only. The __read_mostly annotation thus should be removed.
805
806 **DATE_TIME**
807 It is generally desirable that building the same source code with
808 the same set of tools is reproducible, i.e. the output is always
809 exactly the same.
810
811 The kernel does *not* use the ``__DATE__`` and ``__TIME__`` macros,
812 and enables warnings if they are used as they can lead to
813 non-deterministic builds.
814
815 See: https://www.kernel.org/doc/html/latest/kbuild/reproducible-builds.html#timestamps
816
817 **DEFINE_ARCH_HAS**
818 The ARCH_HAS_xyz and ARCH_HAVE_xyz patterns are wrong.
819
820 For big conceptual features use Kconfig symbols instead. And for
821 smaller things where we have compatibility fallback functions but
822 want architectures able to override them with optimized ones, we
823 should either use weak functions (appropriate for some cases), or
824 the symbol that protects them should be the same symbol we use.
825
826 See: https://lore.kernel.org/lkml/CA+55aFycQ9XJvEOsiM3txHL5bjUc8CeKWJNR_H+MiicaddB42Q@mail.gmail.com/
827
828 **DO_WHILE_MACRO_WITH_TRAILING_SEMICOLON**
829 do {} while(0) macros should not have a trailing semicolon.
830
831 **INIT_ATTRIBUTE**
832 Const init definitions should use __initconst instead of
833 __initdata.
834
835 Similarly init definitions without const require a separate
836 use of const.
837
838 **INLINE_LOCATION**
839 The inline keyword should sit between storage class and type.
840
841 For example, the following segment::
842
843 inline static int example_function(void)
844 {
845 ...
846 }
847
848 should be::
849
850 static inline int example_function(void)
851 {
852 ...
853 }
854
855 **MISPLACED_INIT**
856 It is possible to use section markers on variables in a way
857 which gcc doesn't understand (or at least not the way the
858 developer intended)::
859
860 static struct __initdata samsung_pll_clock exynos4_plls[nr_plls] = {
861
862 does not put exynos4_plls in the .initdata section. The __initdata
863 marker can be virtually anywhere on the line, except right after
864 "struct". The preferred location is before the "=" sign if there is
865 one, or before the trailing ";" otherwise.
866
867 See: https://lore.kernel.org/lkml/1377655732.3619.19.camel@joe-AO722/
868
869 **MULTISTATEMENT_MACRO_USE_DO_WHILE**
870 Macros with multiple statements should be enclosed in a
871 do - while block. Same should also be the case for macros
872 starting with `if` to avoid logic defects::
873
874 #define macrofun(a, b, c) \
875 do { \
876 if (a == 5) \
877 do_this(b, c); \
878 } while (0)
879
880 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#macros-enums-and-rtl
881
882 **PREFER_FALLTHROUGH**
883 Use the `fallthrough;` pseudo keyword instead of
884 `/* fallthrough */` like comments.
885
886 **TRAILING_SEMICOLON**
887 Macro definition should not end with a semicolon. The macro
888 invocation style should be consistent with function calls.
889 This can prevent any unexpected code paths::
890
891 #define MAC do_something;
892
893 If this macro is used within a if else statement, like::
894
895 if (some_condition)
896 MAC;
897
898 else
899 do_something;
900
901 Then there would be a compilation error, because when the macro is
902 expanded there are two trailing semicolons, so the else branch gets
903 orphaned.
904
905 See: https://lore.kernel.org/lkml/1399671106.2912.21.camel@joe-AO725/
906
907 **MACRO_ARG_UNUSED**
908 If function-like macros do not utilize a parameter, it might result
909 in a build warning. We advocate for utilizing static inline functions
910 to replace such macros.
911 For example, for a macro such as the one below::
912
913 #define test(a) do { } while (0)
914
915 there would be a warning like below::
916
917 WARNING: Argument 'a' is not used in function-like macro.
918
919 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#macros-enums-and-rtl
920
921 **SINGLE_STATEMENT_DO_WHILE_MACRO**
922 For the multi-statement macros, it is necessary to use the do-while
923 loop to avoid unpredictable code paths. The do-while loop helps to
924 group the multiple statements into a single one so that a
925 function-like macro can be used as a function only.
926
927 But for the single statement macros, it is unnecessary to use the
928 do-while loop. Although the code is syntactically correct but using
929 the do-while loop is redundant. So remove the do-while loop for single
930 statement macros.
931
932 **WEAK_DECLARATION**
933 Using weak declarations like __attribute__((weak)) or __weak
934 can have unintended link defects. Avoid using them.
935
936
937Functions and Variables
938-----------------------
939
940 **CAMELCASE**
941 Avoid CamelCase Identifiers.
942
943 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#naming
944
945 **CONST_CONST**
946 Using `const <type> const *` is generally meant to be
947 written `const <type> * const`.
948
949 **CONST_STRUCT**
950 Using const is generally a good idea. Checkpatch reads
951 a list of frequently used structs that are always or
952 almost always constant.
953
954 The existing structs list can be viewed from
955 `scripts/const_structs.checkpatch`.
956
957 See: https://lore.kernel.org/lkml/alpine.DEB.2.10.1608281509480.3321@hadrien/
958
959 **EMBEDDED_FUNCTION_NAME**
960 Embedded function names are less appropriate to use as
961 refactoring can cause function renaming. Prefer the use of
962 "%s", __func__ to embedded function names.
963
964 Note that this does not work with -f (--file) checkpatch option
965 as it depends on patch context providing the function name.
966
967 **FUNCTION_ARGUMENTS**
968 This warning is emitted due to any of the following reasons:
969
970 1. Arguments for the function declaration do not follow
971 the identifier name. Example::
972
973 void foo
974 (int bar, int baz)
975
976 This should be corrected to::
977
978 void foo(int bar, int baz)
979
980 2. Some arguments for the function definition do not
981 have an identifier name. Example::
982
983 void foo(int)
984
985 All arguments should have identifier names.
986
987 **FUNCTION_WITHOUT_ARGS**
988 Function declarations without arguments like::
989
990 int foo()
991
992 should be::
993
994 int foo(void)
995
996 **GLOBAL_INITIALISERS**
997 Global variables should not be initialized explicitly to
998 0 (or NULL, false, etc.). Your compiler (or rather your
999 loader, which is responsible for zeroing out the relevant
1000 sections) automatically does it for you.
1001
1002 **INITIALISED_STATIC**
1003 Static variables should not be initialized explicitly to zero.
1004 Your compiler (or rather your loader) automatically does
1005 it for you.
1006
1007 **MULTIPLE_ASSIGNMENTS**
1008 Multiple assignments on a single line makes the code unnecessarily
1009 complicated. So on a single line assign value to a single variable
1010 only, this makes the code more readable and helps avoid typos.
1011
1012 **RETURN_PARENTHESES**
1013 return is not a function and as such doesn't need parentheses::
1014
1015 return (bar);
1016
1017 can simply be::
1018
1019 return bar;
1020
1021
1022Permissions
1023-----------
1024
1025 **DEVICE_ATTR_PERMS**
1026 The permissions used in DEVICE_ATTR are unusual.
1027 Typically only three permissions are used - 0644 (RW), 0444 (RO)
1028 and 0200 (WO).
1029
1030 See: https://www.kernel.org/doc/html/latest/filesystems/sysfs.html#attributes
1031
1032 **EXECUTE_PERMISSIONS**
1033 There is no reason for source files to be executable. The executable
1034 bit can be removed safely.
1035
1036 **EXPORTED_WORLD_WRITABLE**
1037 Exporting world writable sysfs/debugfs files is usually a bad thing.
1038 When done arbitrarily they can introduce serious security bugs.
1039 In the past, some of the debugfs vulnerabilities would seemingly allow
1040 any local user to write arbitrary values into device registers - a
1041 situation from which little good can be expected to emerge.
1042
1043 See: https://lore.kernel.org/linux-arm-kernel/cover.1296818921.git.segoon@openwall.com/
1044
1045 **NON_OCTAL_PERMISSIONS**
1046 Permission bits should use 4 digit octal permissions (like 0700 or 0444).
1047 Avoid using any other base like decimal.
1048
1049 **SYMBOLIC_PERMS**
1050 Permission bits in the octal form are more readable and easier to
1051 understand than their symbolic counterparts because many command-line
1052 tools use this notation. Experienced kernel developers have been using
1053 these traditional Unix permission bits for decades and so they find it
1054 easier to understand the octal notation than the symbolic macros.
1055 For example, it is harder to read S_IWUSR|S_IRUGO than 0644, which
1056 obscures the developer's intent rather than clarifying it.
1057
1058 See: https://lore.kernel.org/lkml/CA+55aFw5v23T-zvDZp-MmD_EYxF8WbafwwB59934FV7g21uMGQ@mail.gmail.com/
1059
1060
1061Spacing and Brackets
1062--------------------
1063
1064 **ASSIGNMENT_CONTINUATIONS**
1065 Assignment operators should not be written at the start of a
1066 line but should follow the operand at the previous line.
1067
1068 **BRACES**
1069 The placement of braces is stylistically incorrect.
1070 The preferred way is to put the opening brace last on the line,
1071 and put the closing brace first::
1072
1073 if (x is true) {
1074 we do y
1075 }
1076
1077 This applies for all non-functional blocks.
1078 However, there is one special case, namely functions: they have the
1079 opening brace at the beginning of the next line, thus::
1080
1081 int function(int x)
1082 {
1083 body of function
1084 }
1085
1086 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#placing-braces-and-spaces
1087
1088 **BRACKET_SPACE**
1089 Whitespace before opening bracket '[' is prohibited.
1090 There are some exceptions:
1091
1092 1. With a type on the left::
1093
1094 int [] a;
1095
1096 2. At the beginning of a line for slice initialisers::
1097
1098 [0...10] = 5,
1099
1100 3. Inside a curly brace::
1101
1102 = { [0...10] = 5 }
1103
1104 **CONCATENATED_STRING**
1105 Concatenated elements should have a space in between.
1106 Example::
1107
1108 printk(KERN_INFO"bar");
1109
1110 should be::
1111
1112 printk(KERN_INFO "bar");
1113
1114 **ELSE_AFTER_BRACE**
1115 `else {` should follow the closing block `}` on the same line.
1116
1117 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#placing-braces-and-spaces
1118
1119 **LINE_SPACING**
1120 Vertical space is wasted given the limited number of lines an
1121 editor window can display when multiple blank lines are used.
1122
1123 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#spaces
1124
1125 **OPEN_BRACE**
1126 The opening brace should be following the function definitions on the
1127 next line. For any non-functional block it should be on the same line
1128 as the last construct.
1129
1130 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#placing-braces-and-spaces
1131
1132 **POINTER_LOCATION**
1133 When using pointer data or a function that returns a pointer type,
1134 the preferred use of * is adjacent to the data name or function name
1135 and not adjacent to the type name.
1136 Examples::
1137
1138 char *linux_banner;
1139 unsigned long long memparse(char *ptr, char **retptr);
1140 char *match_strdup(substring_t *s);
1141
1142 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#spaces
1143
1144 **SPACING**
1145 Whitespace style used in the kernel sources is described in kernel docs.
1146
1147 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#spaces
1148
1149 **TRAILING_WHITESPACE**
1150 Trailing whitespace should always be removed.
1151 Some editors highlight the trailing whitespace and cause visual
1152 distractions when editing files.
1153
1154 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#spaces
1155
1156 **UNNECESSARY_PARENTHESES**
1157 Parentheses are not required in the following cases:
1158
1159 1. Function pointer uses::
1160
1161 (foo->bar)();
1162
1163 could be::
1164
1165 foo->bar();
1166
1167 2. Comparisons in if::
1168
1169 if ((foo->bar) && (foo->baz))
1170 if ((foo == bar))
1171
1172 could be::
1173
1174 if (foo->bar && foo->baz)
1175 if (foo == bar)
1176
1177 3. addressof/dereference single Lvalues::
1178
1179 &(foo->bar)
1180 *(foo->bar)
1181
1182 could be::
1183
1184 &foo->bar
1185 *foo->bar
1186
1187 **WHILE_AFTER_BRACE**
1188 while should follow the closing bracket on the same line::
1189
1190 do {
1191 ...
1192 } while(something);
1193
1194 See: https://www.kernel.org/doc/html/latest/process/coding-style.html#placing-braces-and-spaces
1195
1196
1197Others
1198------
1199
1200 **CONFIG_DESCRIPTION**
1201 Kconfig symbols should have a help text which fully describes
1202 it.
1203
1204 **CORRUPTED_PATCH**
1205 The patch seems to be corrupted or lines are wrapped.
1206 Please regenerate the patch file before sending it to the maintainer.
1207
1208 **CVS_KEYWORD**
1209 Since linux moved to git, the CVS markers are no longer used.
1210 So, CVS style keywords ($Id$, $Revision$, $Log$) should not be
1211 added.
1212
1213 **DEFAULT_NO_BREAK**
1214 switch default case is sometimes written as "default:;". This can
1215 cause new cases added below default to be defective.
1216
1217 A "break;" should be added after empty default statement to avoid
1218 unwanted fallthrough.
1219
1220 **DOS_LINE_ENDINGS**
1221 For DOS-formatted patches, there are extra ^M symbols at the end of
1222 the line. These should be removed.
1223
1224 **DT_SCHEMA_BINDING_PATCH**
1225 DT bindings moved to a json-schema based format instead of
1226 freeform text.
1227
1228 See: https://www.kernel.org/doc/html/latest/devicetree/bindings/writing-schema.html
1229
1230 **DT_SPLIT_BINDING_PATCH**
1231 Devicetree bindings should be their own patch. This is because
1232 bindings are logically independent from a driver implementation,
1233 they have a different maintainer (even though they often
1234 are applied via the same tree), and it makes for a cleaner history in the
1235 DT only tree created with git-filter-branch.
1236
1237 See: https://www.kernel.org/doc/html/latest/devicetree/bindings/submitting-patches.html#i-for-patch-submitters
1238
1239 **EMBEDDED_FILENAME**
1240 Embedding the complete filename path inside the file isn't particularly
1241 useful as often the path is moved around and becomes incorrect.
1242
1243 **FILE_PATH_CHANGES**
1244 Whenever files are added, moved, or deleted, the MAINTAINERS file
1245 patterns can be out of sync or outdated.
1246
1247 So MAINTAINERS might need updating in these cases.
1248
1249 **MEMSET**
1250 The memset use appears to be incorrect. This may be caused due to
1251 badly ordered parameters. Please recheck the usage.
1252
1253 **NOT_UNIFIED_DIFF**
1254 The patch file does not appear to be in unified-diff format. Please
1255 regenerate the patch file before sending it to the maintainer.
1256
1257 **PRINTF_0XDECIMAL**
1258 Prefixing 0x with decimal output is defective and should be corrected.
1259
1260 **SPDX_LICENSE_TAG**
1261 The source file is missing or has an improper SPDX identifier tag.
1262 The Linux kernel requires the precise SPDX identifier in all source files,
1263 and it is thoroughly documented in the kernel docs.
1264
1265 See: https://www.kernel.org/doc/html/latest/process/license-rules.html
1266
1267 **TYPO_SPELLING**
1268 Some words may have been misspelled. Consider reviewing them.