blob: 81d61c5d62ddc75cd2ccd73cddc4ac18f258d0cc [file] [log] [blame]
Mauro Carvalho Chehabf7c9fe42016-09-19 08:07:36 -03001.. _development_advancedtopics:
2
3Advanced topics
4===============
Jonathan Corbet75b02142008-09-30 15:15:56 -06005
6At this point, hopefully, you have a handle on how the development process
7works. There is still more to learn, however! This section will cover a
8number of topics which can be helpful for developers wanting to become a
9regular part of the Linux kernel development process.
10
Mauro Carvalho Chehabf7c9fe42016-09-19 08:07:36 -030011Managing patches with git
12-------------------------
Jonathan Corbet75b02142008-09-30 15:15:56 -060013
14The use of distributed version control for the kernel began in early 2002,
15when Linus first started playing with the proprietary BitKeeper
16application. While BitKeeper was controversial, the approach to software
17version management it embodied most certainly was not. Distributed version
18control enabled an immediate acceleration of the kernel development
19project. In current times, there are several free alternatives to
20BitKeeper. For better or for worse, the kernel project has settled on git
21as its tool of choice.
22
23Managing patches with git can make life much easier for the developer,
24especially as the volume of those patches grows. Git also has its rough
25edges and poses certain hazards; it is a young and powerful tool which is
26still being civilized by its developers. This document will not attempt to
27teach the reader how to use git; that would be sufficient material for a
28long document in its own right. Instead, the focus here will be on how git
29fits into the kernel development process in particular. Developers who
30wish to come up to speed with git will find more information at:
31
Randy Dunlapef0eba42010-05-23 17:02:30 -070032 http://git-scm.com/
Jonathan Corbet75b02142008-09-30 15:15:56 -060033
34 http://www.kernel.org/pub/software/scm/git/docs/user-manual.html
35
36and on various tutorials found on the web.
37
38The first order of business is to read the above sites and get a solid
39understanding of how git works before trying to use it to make patches
40available to others. A git-using developer should be able to obtain a copy
41of the mainline repository, explore the revision history, commit changes to
42the tree, use branches, etc. An understanding of git's tools for the
43rewriting of history (such as rebase) is also useful. Git comes with its
44own terminology and concepts; a new user of git should know about refs,
45remote branches, the index, fast-forward merges, pushes and pulls, detached
46heads, etc. It can all be a little intimidating at the outset, but the
47concepts are not that hard to grasp with a bit of study.
48
49Using git to generate patches for submission by email can be a good
50exercise while coming up to speed.
51
52When you are ready to start putting up git trees for others to look at, you
53will, of course, need a server that can be pulled from. Setting up such a
54server with git-daemon is relatively straightforward if you have a system
55which is accessible to the Internet. Otherwise, free, public hosting sites
56(Github, for example) are starting to appear on the net. Established
57developers can get an account on kernel.org, but those are not easy to come
58by; see http://kernel.org/faq/ for more information.
59
60The normal git workflow involves the use of a lot of branches. Each line
61of development can be separated into a separate "topic branch" and
62maintained independently. Branches in git are cheap, there is no reason to
63not make free use of them. And, in any case, you should not do your
64development in any branch which you intend to ask others to pull from.
65Publicly-available branches should be created with care; merge in patches
66from development branches when they are in complete form and ready to go -
67not before.
68
69Git provides some powerful tools which can allow you to rewrite your
70development history. An inconvenient patch (one which breaks bisection,
71say, or which has some other sort of obvious bug) can be fixed in place or
72made to disappear from the history entirely. A patch series can be
73rewritten as if it had been written on top of today's mainline, even though
74you have been working on it for months. Changes can be transparently
75shifted from one branch to another. And so on. Judicious use of git's
76ability to revise history can help in the creation of clean patch sets with
77fewer problems.
78
79Excessive use of this capability can lead to other problems, though, beyond
80a simple obsession for the creation of the perfect project history.
81Rewriting history will rewrite the changes contained in that history,
82turning a tested (hopefully) kernel tree into an untested one. But, beyond
83that, developers cannot easily collaborate if they do not have a shared
84view of the project history; if you rewrite history which other developers
85have pulled into their repositories, you will make life much more difficult
86for those developers. So a simple rule of thumb applies here: history
87which has been exported to others should generally be seen as immutable
88thereafter.
89
90So, once you push a set of changes to your publicly-available server, those
91changes should not be rewritten. Git will attempt to enforce this rule if
92you try to push changes which do not result in a fast-forward merge
93(i.e. changes which do not share the same history). It is possible to
94override this check, and there may be times when it is necessary to rewrite
95an exported tree. Moving changesets between trees to avoid conflicts in
96linux-next is one example. But such actions should be rare. This is one
97of the reasons why development should be done in private branches (which
98can be rewritten if necessary) and only moved into public branches when
99it's in a reasonably advanced state.
100
101As the mainline (or other tree upon which a set of changes is based)
102advances, it is tempting to merge with that tree to stay on the leading
103edge. For a private branch, rebasing can be an easy way to keep up with
104another tree, but rebasing is not an option once a tree is exported to the
105world. Once that happens, a full merge must be done. Merging occasionally
106makes good sense, but overly frequent merges can clutter the history
107needlessly. Suggested technique in this case is to merge infrequently, and
108generally only at specific release points (such as a mainline -rc
109release). If you are nervous about specific changes, you can always
110perform test merges in a private branch. The git "rerere" tool can be
111useful in such situations; it remembers how merge conflicts were resolved
112so that you don't have to do the same work twice.
113
114One of the biggest recurring complaints about tools like git is this: the
115mass movement of patches from one repository to another makes it easy to
116slip in ill-advised changes which go into the mainline below the review
117radar. Kernel developers tend to get unhappy when they see that kind of
118thing happening; putting up a git tree with unreviewed or off-topic patches
119can affect your ability to get trees pulled in the future. Quoting Linus:
120
Mauro Carvalho Chehabf7c9fe42016-09-19 08:07:36 -0300121::
122
Jonathan Corbet75b02142008-09-30 15:15:56 -0600123 You can send me patches, but for me to pull a git patch from you, I
124 need to know that you know what you're doing, and I need to be able
125 to trust things *without* then having to go and check every
126 individual change by hand.
127
Jonathan Corbet5c050fb2011-03-25 12:17:53 -0600128(http://lwn.net/Articles/224135/).
Jonathan Corbet75b02142008-09-30 15:15:56 -0600129
130To avoid this kind of situation, ensure that all patches within a given
131branch stick closely to the associated topic; a "driver fixes" branch
132should not be making changes to the core memory management code. And, most
133importantly, do not use a git tree to bypass the review process. Post an
134occasional summary of the tree to the relevant list, and, when the time is
135right, request that the tree be included in linux-next.
136
137If and when others start to send patches for inclusion into your tree,
138don't forget to review them. Also ensure that you maintain the correct
139authorship information; the git "am" tool does its best in this regard, but
140you may have to add a "From:" line to the patch if it has been relayed to
141you via a third party.
142
143When requesting a pull, be sure to give all the relevant information: where
144your tree is, what branch to pull, and what changes will result from the
145pull. The git request-pull command can be helpful in this regard; it will
146format the request as other developers expect, and will also check to be
Jonathan Corbet5c050fb2011-03-25 12:17:53 -0600147sure that you have remembered to push those changes to the public server.
Jonathan Corbet75b02142008-09-30 15:15:56 -0600148
149
Mauro Carvalho Chehabf7c9fe42016-09-19 08:07:36 -0300150Reviewing patches
151-----------------
Jonathan Corbet75b02142008-09-30 15:15:56 -0600152
153Some readers will certainly object to putting this section with "advanced
154topics" on the grounds that even beginning kernel developers should be
155reviewing patches. It is certainly true that there is no better way to
156learn how to program in the kernel environment than by looking at code
157posted by others. In addition, reviewers are forever in short supply; by
158looking at code you can make a significant contribution to the process as a
159whole.
160
161Reviewing code can be an intimidating prospect, especially for a new kernel
162developer who may well feel nervous about questioning code - in public -
163which has been posted by those with more experience. Even code written by
164the most experienced developers can be improved, though. Perhaps the best
165piece of advice for reviewers (all reviewers) is this: phrase review
166comments as questions rather than criticisms. Asking "how does the lock
167get released in this path?" will always work better than stating "the
168locking here is wrong."
169
170Different developers will review code from different points of view. Some
171are mostly concerned with coding style and whether code lines have trailing
172white space. Others will focus primarily on whether the change implemented
173by the patch as a whole is a good thing for the kernel or not. Yet others
174will check for problematic locking, excessive stack usage, possible
175security issues, duplication of code found elsewhere, adequate
176documentation, adverse effects on performance, user-space ABI changes, etc.
177All types of review, if they lead to better code going into the kernel, are
178welcome and worthwhile.
179
180