blob: 237d54c44bc5ee504dc57f8d5f65edc389081e04 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001
2 How to Get Your Change Into the Linux Kernel
3 or
4 Care And Operation Of Your Linus Torvalds
5
6
7
8For a person or company who wishes to submit a change to the Linux
9kernel, the process can sometimes be daunting if you're not familiar
10with "the system." This text is a collection of suggestions which
11can greatly increase the chances of your change being accepted.
12
13If you are submitting a driver, also read Documentation/SubmittingDrivers.
14
15
16
17--------------------------------------------
18SECTION 1 - CREATING AND SENDING YOUR CHANGE
19--------------------------------------------
20
21
22
231) "diff -up"
24------------
25
26Use "diff -up" or "diff -uprN" to create patches.
27
28All changes to the Linux kernel occur in the form of patches, as
29generated by diff(1). When creating your patch, make sure to create it
30in "unified diff" format, as supplied by the '-u' argument to diff(1).
31Also, please use the '-p' argument which shows which C function each
32change is in - that makes the resultant diff a lot easier to read.
33Patches should be based in the root kernel source directory,
34not in any lower subdirectory.
35
36To create a patch for a single file, it is often sufficient to do:
37
Randy Dunlap84da7c02005-06-28 20:45:30 -070038 SRCTREE= linux-2.6
Linus Torvalds1da177e2005-04-16 15:20:36 -070039 MYFILE= drivers/net/mydriver.c
40
41 cd $SRCTREE
42 cp $MYFILE $MYFILE.orig
43 vi $MYFILE # make your change
44 cd ..
45 diff -up $SRCTREE/$MYFILE{.orig,} > /tmp/patch
46
47To create a patch for multiple files, you should unpack a "vanilla",
48or unmodified kernel source tree, and generate a diff against your
49own source tree. For example:
50
Randy Dunlap84da7c02005-06-28 20:45:30 -070051 MYSRC= /devel/linux-2.6
Linus Torvalds1da177e2005-04-16 15:20:36 -070052
Randy Dunlap84da7c02005-06-28 20:45:30 -070053 tar xvfz linux-2.6.12.tar.gz
54 mv linux-2.6.12 linux-2.6.12-vanilla
55 diff -uprN -X linux-2.6.12-vanilla/Documentation/dontdiff \
56 linux-2.6.12-vanilla $MYSRC > /tmp/patch
Linus Torvalds1da177e2005-04-16 15:20:36 -070057
58"dontdiff" is a list of files which are generated by the kernel during
59the build process, and should be ignored in any diff(1)-generated
Randy Dunlap84da7c02005-06-28 20:45:30 -070060patch. The "dontdiff" file is included in the kernel tree in
612.6.12 and later. For earlier kernel versions, you can get it
62from <http://www.xenotime.net/linux/doc/dontdiff>.
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
64Make sure your patch does not include any extra files which do not
65belong in a patch submission. Make sure to review your patch -after-
66generated it with diff(1), to ensure accuracy.
67
68If your changes produce a lot of deltas, you may want to look into
69splitting them into individual patches which modify things in
Randy Dunlap84da7c02005-06-28 20:45:30 -070070logical stages. This will facilitate easier reviewing by other
Linus Torvalds1da177e2005-04-16 15:20:36 -070071kernel developers, very important if you want your patch accepted.
Randy Dunlap84da7c02005-06-28 20:45:30 -070072There are a number of scripts which can aid in this:
Linus Torvalds1da177e2005-04-16 15:20:36 -070073
74Quilt:
75http://savannah.nongnu.org/projects/quilt
76
77Randy Dunlap's patch scripts:
Randy Dunlap84da7c02005-06-28 20:45:30 -070078http://www.xenotime.net/linux/scripts/patching-scripts-002.tar.gz
Linus Torvalds1da177e2005-04-16 15:20:36 -070079
80Andrew Morton's patch scripts:
Randy Dunlap84da7c02005-06-28 20:45:30 -070081http://www.zip.com.au/~akpm/linux/patches/patch-scripts-0.20
82
83
Linus Torvalds1da177e2005-04-16 15:20:36 -070084
852) Describe your changes.
86
87Describe the technical detail of the change(s) your patch includes.
88
89Be as specific as possible. The WORST descriptions possible include
90things like "update driver X", "bug fix for driver X", or "this patch
91includes updates for subsystem X. Please apply."
92
93If your description starts to get long, that's a sign that you probably
94need to split up your patch. See #3, next.
95
96
97
983) Separate your changes.
99
100Separate each logical change into its own patch.
101
102For example, if your changes include both bug fixes and performance
103enhancements for a single driver, separate those changes into two
104or more patches. If your changes include an API update, and a new
105driver which uses that new API, separate those into two patches.
106
107On the other hand, if you make a single change to numerous files,
108group those changes into a single patch. Thus a single logical change
109is contained within a single patch.
110
111If one patch depends on another patch in order for a change to be
112complete, that is OK. Simply note "this patch depends on patch X"
113in your patch description.
114
115
1164) Select e-mail destination.
117
118Look through the MAINTAINERS file and the source code, and determine
119if your change applies to a specific subsystem of the kernel, with
120an assigned maintainer. If so, e-mail that person.
121
122If no maintainer is listed, or the maintainer does not respond, send
123your patch to the primary Linux kernel developer's mailing list,
124linux-kernel@vger.kernel.org. Most kernel developers monitor this
125e-mail list, and can comment on your changes.
126
127Linus Torvalds is the final arbiter of all changes accepted into the
128Linux kernel. His e-mail address is <torvalds@osdl.org>. He gets
129a lot of e-mail, so typically you should do your best to -avoid- sending
130him e-mail.
131
132Patches which are bug fixes, are "obvious" changes, or similarly
133require little discussion should be sent or CC'd to Linus. Patches
134which require discussion or do not have a clear advantage should
135usually be sent first to linux-kernel. Only after the patch is
136discussed should the patch then be submitted to Linus.
137
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138
139
1405) Select your CC (e-mail carbon copy) list.
141
142Unless you have a reason NOT to do so, CC linux-kernel@vger.kernel.org.
143
144Other kernel developers besides Linus need to be aware of your change,
145so that they may comment on it and offer code review and suggestions.
146linux-kernel is the primary Linux kernel developer mailing list.
147Other mailing lists are available for specific subsystems, such as
148USB, framebuffer devices, the VFS, the SCSI subsystem, etc. See the
149MAINTAINERS file for a mailing list that relates specifically to
150your change.
151
Paul Jackson1caf1f02005-07-31 22:34:48 -0700152If changes affect userland-kernel interfaces, please send
153the MAN-PAGES maintainer (as listed in the MAINTAINERS file)
154a man-pages patch, or at least a notification of the change,
155so that some information makes its way into the manual pages.
156
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157Even if the maintainer did not respond in step #4, make sure to ALWAYS
158copy the maintainer when you change their code.
159
160For small patches you may want to CC the Trivial Patch Monkey
161trivial@rustcorp.com.au set up by Rusty Russell; which collects "trivial"
162patches. Trivial patches must qualify for one of the following rules:
163 Spelling fixes in documentation
164 Spelling fixes which could break grep(1).
165 Warning fixes (cluttering with useless warnings is bad)
166 Compilation fixes (only if they are actually correct)
167 Runtime fixes (only if they actually fix things)
168 Removing use of deprecated functions/macros (eg. check_region).
169 Contact detail and documentation fixes
170 Non-portable code replaced by portable code (even in arch-specific,
171 since people copy, as long as it's trivial)
172 Any fix by the author/maintainer of the file. (ie. patch monkey
173 in re-transmission mode)
Randy Dunlap84da7c02005-06-28 20:45:30 -0700174URL: <http://www.kernel.org/pub/linux/kernel/people/rusty/trivial/>
175
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176
177
178
1796) No MIME, no links, no compression, no attachments. Just plain text.
180
181Linus and other kernel developers need to be able to read and comment
182on the changes you are submitting. It is important for a kernel
183developer to be able to "quote" your changes, using standard e-mail
184tools, so that they may comment on specific portions of your code.
185
186For this reason, all patches should be submitting e-mail "inline".
187WARNING: Be wary of your editor's word-wrap corrupting your patch,
188if you choose to cut-n-paste your patch.
189
190Do not attach the patch as a MIME attachment, compressed or not.
191Many popular e-mail applications will not always transmit a MIME
192attachment as plain text, making it impossible to comment on your
193code. A MIME attachment also takes Linus a bit more time to process,
194decreasing the likelihood of your MIME-attached change being accepted.
195
196Exception: If your mailer is mangling patches then someone may ask
197you to re-send them using MIME.
198
199
200
2017) E-mail size.
202
203When sending patches to Linus, always follow step #6.
204
205Large changes are not appropriate for mailing lists, and some
206maintainers. If your patch, uncompressed, exceeds 40 kB in size,
207it is preferred that you store your patch on an Internet-accessible
208server, and provide instead a URL (link) pointing to your patch.
209
210
211
2128) Name your kernel version.
213
214It is important to note, either in the subject line or in the patch
215description, the kernel version to which this patch applies.
216
217If the patch does not apply cleanly to the latest kernel version,
218Linus will not apply it.
219
220
221
2229) Don't get discouraged. Re-submit.
223
224After you have submitted your change, be patient and wait. If Linus
225likes your change and applies it, it will appear in the next version
226of the kernel that he releases.
227
228However, if your change doesn't appear in the next version of the
229kernel, there could be any number of reasons. It's YOUR job to
230narrow down those reasons, correct what was wrong, and submit your
231updated change.
232
233It is quite common for Linus to "drop" your patch without comment.
234That's the nature of the system. If he drops your patch, it could be
235due to
236* Your patch did not apply cleanly to the latest kernel version
237* Your patch was not sufficiently discussed on linux-kernel.
238* A style issue (see section 2),
239* An e-mail formatting issue (re-read this section)
240* A technical problem with your change
241* He gets tons of e-mail, and yours got lost in the shuffle
242* You are being annoying (See Figure 1)
243
244When in doubt, solicit comments on linux-kernel mailing list.
245
246
247
24810) Include PATCH in the subject
249
250Due to high e-mail traffic to Linus, and to linux-kernel, it is common
251convention to prefix your subject line with [PATCH]. This lets Linus
252and other kernel developers more easily distinguish patches from other
253e-mail discussions.
254
255
256
25711) Sign your work
258
259To improve tracking of who did what, especially with patches that can
260percolate to their final resting place in the kernel through several
261layers of maintainers, we've introduced a "sign-off" procedure on
262patches that are being emailed around.
263
264The sign-off is a simple line at the end of the explanation for the
265patch, which certifies that you wrote it or otherwise have the right to
266pass it on as a open-source patch. The rules are pretty simple: if you
267can certify the below:
268
Linus Torvaldscbd83da2005-06-13 17:51:55 -0700269 Developer's Certificate of Origin 1.1
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270
271 By making a contribution to this project, I certify that:
272
273 (a) The contribution was created in whole or in part by me and I
274 have the right to submit it under the open source license
275 indicated in the file; or
276
277 (b) The contribution is based upon previous work that, to the best
278 of my knowledge, is covered under an appropriate open source
279 license and I have the right under that license to submit that
280 work with modifications, whether created in whole or in part
281 by me, under the same open source license (unless I am
282 permitted to submit under a different license), as indicated
283 in the file; or
284
285 (c) The contribution was provided directly to me by some other
286 person who certified (a), (b) or (c) and I have not modified
287 it.
288
Linus Torvaldscbd83da2005-06-13 17:51:55 -0700289 (d) I understand and agree that this project and the contribution
290 are public and that a record of the contribution (including all
291 personal information I submit with it, including my sign-off) is
292 maintained indefinitely and may be redistributed consistent with
293 this project or the open source license(s) involved.
294
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295then you just add a line saying
296
Alexey Dobriyan9fd55592005-06-25 14:59:34 -0700297 Signed-off-by: Random J Developer <random@developer.example.org>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298
299Some people also put extra tags at the end. They'll just be ignored for
300now, but you can do this to mark internal company procedures or just
301point out some special detail about the sign-off.
302
303
Paul Jackson75f84262005-10-02 18:01:42 -070030412) The canonical patch format
Randy Dunlap84da7c02005-06-28 20:45:30 -0700305
Paul Jackson75f84262005-10-02 18:01:42 -0700306The canonical patch subject line is:
307
Paul Jacksond6b9acc2005-10-03 00:29:10 -0700308 Subject: [PATCH 001/123] subsystem: summary phrase
Paul Jackson75f84262005-10-02 18:01:42 -0700309
310The canonical patch message body contains the following:
311
312 - A "from" line specifying the patch author.
313
314 - An empty line.
315
316 - The body of the explanation, which will be copied to the
317 permanent changelog to describe this patch.
318
319 - The "Signed-off-by:" lines, described above, which will
320 also go in the changelog.
321
322 - A marker line containing simply "---".
323
324 - Any additional comments not suitable for the changelog.
325
326 - The actual patch (diff output).
327
328The Subject line format makes it very easy to sort the emails
329alphabetically by subject line - pretty much any email reader will
330support that - since because the sequence number is zero-padded,
331the numerical and alphabetic sort is the same.
332
Paul Jacksond6b9acc2005-10-03 00:29:10 -0700333The "subsystem" in the email's Subject should identify which
334area or subsystem of the kernel is being patched.
335
336The "summary phrase" in the email's Subject should concisely
337describe the patch which that email contains. The "summary
338phrase" should not be a filename. Do not use the same "summary
339phrase" for every patch in a whole patch series.
340
341Bear in mind that the "summary phrase" of your email becomes
342a globally-unique identifier for that patch. It propagates
343all the way into the git changelog. The "summary phrase" may
344later be used in developer discussions which refer to the patch.
345People will want to google for the "summary phrase" to read
346discussion regarding that patch.
347
348A couple of example Subjects:
349
350 Subject: [patch 2/5] ext2: improve scalability of bitmap searching
351 Subject: [PATCHv2 001/207] x86: fix eflags tracking
Paul Jackson75f84262005-10-02 18:01:42 -0700352
353The "from" line must be the very first line in the message body,
354and has the form:
355
356 From: Original Author <author@example.com>
357
358The "from" line specifies who will be credited as the author of the
359patch in the permanent changelog. If the "from" line is missing,
360then the "From:" line from the email header will be used to determine
361the patch author in the changelog.
362
363The explanation body will be committed to the permanent source
364changelog, so should make sense to a competent reader who has long
365since forgotten the immediate details of the discussion that might
366have led to this patch.
367
368The "---" marker line serves the essential purpose of marking for patch
369handling tools where the changelog message ends.
370
371One good use for the additional comments after the "---" marker is for
372a diffstat, to show what files have changed, and the number of inserted
373and deleted lines per file. A diffstat is especially useful on bigger
374patches. Other comments relevant only to the moment or the maintainer,
375not suitable for the permanent changelog, should also go here.
376
377See more details on the proper patch format in the following
378references.
379
380
38113) More references for submitting patches
Randy Dunlap84da7c02005-06-28 20:45:30 -0700382
383Andrew Morton, "The perfect patch" (tpp).
384 <http://www.zip.com.au/~akpm/linux/patches/stuff/tpp.txt>
385
386Jeff Garzik, "Linux kernel patch submission format."
387 <http://linux.yyz.us/patch-format.html>
388
Paul Jackson75f84262005-10-02 18:01:42 -0700389Greg KH, "How to piss off a kernel subsystem maintainer"
390 <http://www.kroah.com/log/2005/03/31/>
391
392Kernel Documentation/CodingStyle
393 <http://sosdg.org/~coywolf/lxr/source/Documentation/CodingStyle>
394
395Linus Torvald's mail on the canonical patch format:
396 <http://lkml.org/lkml/2005/4/7/183>
Randy Dunlap84da7c02005-06-28 20:45:30 -0700397
398
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399-----------------------------------
400SECTION 2 - HINTS, TIPS, AND TRICKS
401-----------------------------------
402
403This section lists many of the common "rules" associated with code
404submitted to the kernel. There are always exceptions... but you must
405have a really good reason for doing so. You could probably call this
406section Linus Computer Science 101.
407
408
409
4101) Read Documentation/CodingStyle
411
412Nuff said. If your code deviates too much from this, it is likely
413to be rejected without further review, and without comment.
414
415
416
4172) #ifdefs are ugly
418
419Code cluttered with ifdefs is difficult to read and maintain. Don't do
420it. Instead, put your ifdefs in a header, and conditionally define
421'static inline' functions, or macros, which are used in the code.
422Let the compiler optimize away the "no-op" case.
423
424Simple example, of poor code:
425
426 dev = alloc_etherdev (sizeof(struct funky_private));
427 if (!dev)
428 return -ENODEV;
429 #ifdef CONFIG_NET_FUNKINESS
430 init_funky_net(dev);
431 #endif
432
433Cleaned-up example:
434
435(in header)
436 #ifndef CONFIG_NET_FUNKINESS
437 static inline void init_funky_net (struct net_device *d) {}
438 #endif
439
440(in the code itself)
441 dev = alloc_etherdev (sizeof(struct funky_private));
442 if (!dev)
443 return -ENODEV;
444 init_funky_net(dev);
445
446
447
4483) 'static inline' is better than a macro
449
450Static inline functions are greatly preferred over macros.
451They provide type safety, have no length limitations, no formatting
452limitations, and under gcc they are as cheap as macros.
453
454Macros should only be used for cases where a static inline is clearly
455suboptimal [there a few, isolated cases of this in fast paths],
456or where it is impossible to use a static inline function [such as
457string-izing].
458
459'static inline' is preferred over 'static __inline__', 'extern inline',
460and 'extern __inline__'.
461
462
463
4644) Don't over-design.
465
466Don't try to anticipate nebulous future cases which may or may not
Randy Dunlap84da7c02005-06-28 20:45:30 -0700467be useful: "Make it as simple as you can, and no simpler."
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468