Based on kernel version 4.9. Page generated on 2016-12-21 14:28 EST.
1 .. _development_advancedtopics: 2 3 Advanced topics 4 =============== 5 6 At this point, hopefully, you have a handle on how the development process 7 works. There is still more to learn, however! This section will cover a 8 number of topics which can be helpful for developers wanting to become a 9 regular part of the Linux kernel development process. 10 11 Managing patches with git 12 ------------------------- 13 14 The use of distributed version control for the kernel began in early 2002, 15 when Linus first started playing with the proprietary BitKeeper 16 application. While BitKeeper was controversial, the approach to software 17 version management it embodied most certainly was not. Distributed version 18 control enabled an immediate acceleration of the kernel development 19 project. In current times, there are several free alternatives to 20 BitKeeper. For better or for worse, the kernel project has settled on git 21 as its tool of choice. 22 23 Managing patches with git can make life much easier for the developer, 24 especially as the volume of those patches grows. Git also has its rough 25 edges and poses certain hazards; it is a young and powerful tool which is 26 still being civilized by its developers. This document will not attempt to 27 teach the reader how to use git; that would be sufficient material for a 28 long document in its own right. Instead, the focus here will be on how git 29 fits into the kernel development process in particular. Developers who 30 wish to come up to speed with git will find more information at: 31 32 http://git-scm.com/ 33 34 http://www.kernel.org/pub/software/scm/git/docs/user-manual.html 35 36 and on various tutorials found on the web. 37 38 The first order of business is to read the above sites and get a solid 39 understanding of how git works before trying to use it to make patches 40 available to others. A git-using developer should be able to obtain a copy 41 of the mainline repository, explore the revision history, commit changes to 42 the tree, use branches, etc. An understanding of git's tools for the 43 rewriting of history (such as rebase) is also useful. Git comes with its 44 own terminology and concepts; a new user of git should know about refs, 45 remote branches, the index, fast-forward merges, pushes and pulls, detached 46 heads, etc. It can all be a little intimidating at the outset, but the 47 concepts are not that hard to grasp with a bit of study. 48 49 Using git to generate patches for submission by email can be a good 50 exercise while coming up to speed. 51 52 When you are ready to start putting up git trees for others to look at, you 53 will, of course, need a server that can be pulled from. Setting up such a 54 server with git-daemon is relatively straightforward if you have a system 55 which is accessible to the Internet. Otherwise, free, public hosting sites 56 (Github, for example) are starting to appear on the net. Established 57 developers can get an account on kernel.org, but those are not easy to come 58 by; see http://kernel.org/faq/ for more information. 59 60 The normal git workflow involves the use of a lot of branches. Each line 61 of development can be separated into a separate "topic branch" and 62 maintained independently. Branches in git are cheap, there is no reason to 63 not make free use of them. And, in any case, you should not do your 64 development in any branch which you intend to ask others to pull from. 65 Publicly-available branches should be created with care; merge in patches 66 from development branches when they are in complete form and ready to go - 67 not before. 68 69 Git provides some powerful tools which can allow you to rewrite your 70 development history. An inconvenient patch (one which breaks bisection, 71 say, or which has some other sort of obvious bug) can be fixed in place or 72 made to disappear from the history entirely. A patch series can be 73 rewritten as if it had been written on top of today's mainline, even though 74 you have been working on it for months. Changes can be transparently 75 shifted from one branch to another. And so on. Judicious use of git's 76 ability to revise history can help in the creation of clean patch sets with 77 fewer problems. 78 79 Excessive use of this capability can lead to other problems, though, beyond 80 a simple obsession for the creation of the perfect project history. 81 Rewriting history will rewrite the changes contained in that history, 82 turning a tested (hopefully) kernel tree into an untested one. But, beyond 83 that, developers cannot easily collaborate if they do not have a shared 84 view of the project history; if you rewrite history which other developers 85 have pulled into their repositories, you will make life much more difficult 86 for those developers. So a simple rule of thumb applies here: history 87 which has been exported to others should generally be seen as immutable 88 thereafter. 89 90 So, once you push a set of changes to your publicly-available server, those 91 changes should not be rewritten. Git will attempt to enforce this rule if 92 you 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 94 override this check, and there may be times when it is necessary to rewrite 95 an exported tree. Moving changesets between trees to avoid conflicts in 96 linux-next is one example. But such actions should be rare. This is one 97 of the reasons why development should be done in private branches (which 98 can be rewritten if necessary) and only moved into public branches when 99 it's in a reasonably advanced state. 100 101 As the mainline (or other tree upon which a set of changes is based) 102 advances, it is tempting to merge with that tree to stay on the leading 103 edge. For a private branch, rebasing can be an easy way to keep up with 104 another tree, but rebasing is not an option once a tree is exported to the 105 world. Once that happens, a full merge must be done. Merging occasionally 106 makes good sense, but overly frequent merges can clutter the history 107 needlessly. Suggested technique in this case is to merge infrequently, and 108 generally only at specific release points (such as a mainline -rc 109 release). If you are nervous about specific changes, you can always 110 perform test merges in a private branch. The git "rerere" tool can be 111 useful in such situations; it remembers how merge conflicts were resolved 112 so that you don't have to do the same work twice. 113 114 One of the biggest recurring complaints about tools like git is this: the 115 mass movement of patches from one repository to another makes it easy to 116 slip in ill-advised changes which go into the mainline below the review 117 radar. Kernel developers tend to get unhappy when they see that kind of 118 thing happening; putting up a git tree with unreviewed or off-topic patches 119 can affect your ability to get trees pulled in the future. Quoting Linus: 120 121 :: 122 123 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. 127 128 (http://lwn.net/Articles/224135/). 129 130 To avoid this kind of situation, ensure that all patches within a given 131 branch stick closely to the associated topic; a "driver fixes" branch 132 should not be making changes to the core memory management code. And, most 133 importantly, do not use a git tree to bypass the review process. Post an 134 occasional summary of the tree to the relevant list, and, when the time is 135 right, request that the tree be included in linux-next. 136 137 If and when others start to send patches for inclusion into your tree, 138 don't forget to review them. Also ensure that you maintain the correct 139 authorship information; the git "am" tool does its best in this regard, but 140 you may have to add a "From:" line to the patch if it has been relayed to 141 you via a third party. 142 143 When requesting a pull, be sure to give all the relevant information: where 144 your tree is, what branch to pull, and what changes will result from the 145 pull. The git request-pull command can be helpful in this regard; it will 146 format the request as other developers expect, and will also check to be 147 sure that you have remembered to push those changes to the public server. 148 149 150 Reviewing patches 151 ----------------- 152 153 Some readers will certainly object to putting this section with "advanced 154 topics" on the grounds that even beginning kernel developers should be 155 reviewing patches. It is certainly true that there is no better way to 156 learn how to program in the kernel environment than by looking at code 157 posted by others. In addition, reviewers are forever in short supply; by 158 looking at code you can make a significant contribution to the process as a 159 whole. 160 161 Reviewing code can be an intimidating prospect, especially for a new kernel 162 developer who may well feel nervous about questioning code - in public - 163 which has been posted by those with more experience. Even code written by 164 the most experienced developers can be improved, though. Perhaps the best 165 piece of advice for reviewers (all reviewers) is this: phrase review 166 comments as questions rather than criticisms. Asking "how does the lock 167 get released in this path?" will always work better than stating "the 168 locking here is wrong." 169 170 Different developers will review code from different points of view. Some 171 are mostly concerned with coding style and whether code lines have trailing 172 white space. Others will focus primarily on whether the change implemented 173 by the patch as a whole is a good thing for the kernel or not. Yet others 174 will check for problematic locking, excessive stack usage, possible 175 security issues, duplication of code found elsewhere, adequate 176 documentation, adverse effects on performance, user-space ABI changes, etc. 177 All types of review, if they lead to better code going into the kernel, are 178 welcome and worthwhile. 179