605 lines · plain
1.. SPDX-License-Identifier: GPL-2.02 3===================================4Backporting and conflict resolution5===================================6 7:Author: Vegard Nossum <vegard.nossum@oracle.com>8 9.. contents::10 :local:11 :depth: 312 :backlinks: none13 14Introduction15============16 17Some developers may never really have to deal with backporting patches,18merging branches, or resolving conflicts in their day-to-day work, so19when a merge conflict does pop up, it can be daunting. Luckily,20resolving conflicts is a skill like any other, and there are many useful21techniques you can use to make the process smoother and increase your22confidence in the result.23 24This document aims to be a comprehensive, step-by-step guide to25backporting and conflict resolution.26 27Applying the patch to a tree28============================29 30Sometimes the patch you are backporting already exists as a git commit,31in which case you just cherry-pick it directly using32``git cherry-pick``. However, if the patch comes from an email, as it33often does for the Linux kernel, you will need to apply it to a tree34using ``git am``.35 36If you've ever used ``git am``, you probably already know that it is37quite picky about the patch applying perfectly to your source tree. In38fact, you've probably had nightmares about ``.rej`` files and trying to39edit the patch to make it apply.40 41It is strongly recommended to instead find an appropriate base version42where the patch applies cleanly and *then* cherry-pick it over to your43destination tree, as this will make git output conflict markers and let44you resolve conflicts with the help of git and any other conflict45resolution tools you might prefer to use. For example, if you want to46apply a patch that just arrived on LKML to an older stable kernel, you47can apply it to the most recent mainline kernel and then cherry-pick it48to your older stable branch.49 50It's generally better to use the exact same base as the one the patch51was generated from, but it doesn't really matter that much as long as it52applies cleanly and isn't too far from the original base. The only53problem with applying the patch to the "wrong" base is that it may pull54in more unrelated changes in the context of the diff when cherry-picking55it to the older branch.56 57A good reason to prefer ``git cherry-pick`` over ``git am`` is that git58knows the precise history of an existing commit, so it will know when59code has moved around and changed the line numbers; this in turn makes60it less likely to apply the patch to the wrong place (which can result61in silent mistakes or messy conflicts).62 63If you are using `b4`_. and you are applying the patch directly from an64email, you can use ``b4 am`` with the options ``-g``/``--guess-base``65and ``-3``/``--prep-3way`` to do some of this automatically (see the66`b4 presentation`_ for more information). However, the rest of this67article will assume that you are doing a plain ``git cherry-pick``.68 69.. _b4: https://people.kernel.org/monsieuricon/introducing-b4-and-patch-attestation70.. _b4 presentation: https://youtu.be/mF10hgVIx9o?t=299671 72Once you have the patch in git, you can go ahead and cherry-pick it into73your source tree. Don't forget to cherry-pick with ``-x`` if you want a74written record of where the patch came from!75 76Note that if you are submitting a patch for stable, the format is77slightly different; the first line after the subject line needs tobe78either::79 80 commit <upstream commit> upstream81 82or::83 84 [ Upstream commit <upstream commit> ]85 86Resolving conflicts87===================88 89Uh-oh; the cherry-pick failed with a vaguely threatening message::90 91 CONFLICT (content): Merge conflict92 93What to do now?94 95In general, conflicts appear when the context of the patch (i.e., the96lines being changed and/or the lines surrounding the changes) doesn't97match what's in the tree you are trying to apply the patch *to*.98 99For backports, what likely happened was that the branch you are100backporting from contains patches not in the branch you are backporting101to. However, the reverse is also possible. In any case, the result is a102conflict that needs to be resolved.103 104If your attempted cherry-pick fails with a conflict, git automatically105edits the files to include so-called conflict markers showing you where106the conflict is and how the two branches have diverged. Resolving the107conflict typically means editing the end result in such a way that it108takes into account these other commits.109 110Resolving the conflict can be done either by hand in a regular text111editor or using a dedicated conflict resolution tool.112 113Many people prefer to use their regular text editor and edit the114conflict directly, as it may be easier to understand what you're doing115and to control the final result. There are definitely pros and cons to116each method, and sometimes there's value in using both.117 118We will not cover using dedicated merge tools here beyond providing some119pointers to various tools that you could use:120 121- `Emacs Ediff mode <https://www.emacswiki.org/emacs/EdiffMode>`__122- `vimdiff/gvimdiff <https://linux.die.net/man/1/vimdiff>`__123- `KDiff3 <http://kdiff3.sourceforge.net/>`__124- `TortoiseMerge <https://tortoisesvn.net/TortoiseMerge.html>`__125- `Meld <https://meldmerge.org/help/>`__126- `P4Merge <https://www.perforce.com/products/helix-core-apps/merge-diff-tool-p4merge>`__127- `Beyond Compare <https://www.scootersoftware.com/>`__128- `IntelliJ <https://www.jetbrains.com/help/idea/resolve-conflicts.html>`__129- `VSCode <https://code.visualstudio.com/docs/editor/versioncontrol>`__130 131To configure git to work with these, see ``git mergetool --help`` or132the official `git-mergetool documentation`_.133 134.. _git-mergetool documentation: https://git-scm.com/docs/git-mergetool135 136Prerequisite patches137--------------------138 139Most conflicts happen because the branch you are backporting to is140missing some patches compared to the branch you are backporting *from*.141In the more general case (such as merging two independent branches),142development could have happened on either branch, or the branches have143simply diverged -- perhaps your older branch had some other backports144applied to it that themselves needed conflict resolutions, causing a145divergence.146 147It's important to always identify the commit or commits that caused the148conflict, as otherwise you cannot be confident in the correctness of149your resolution. As an added bonus, especially if the patch is in an150area you're not that familiar with, the changelogs of these commits will151often give you the context to understand the code and potential problems152or pitfalls with your conflict resolution.153 154git log155~~~~~~~156 157A good first step is to look at ``git log`` for the file that has the158conflict -- this is usually sufficient when there aren't a lot of159patches to the file, but may get confusing if the file is big and160frequently patched. You should run ``git log`` on the range of commits161between your currently checked-out branch (``HEAD``) and the parent of162the patch you are picking (``<commit>``), i.e.::163 164 git log HEAD..<commit>^ -- <path>165 166Even better, if you want to restrict this output to a single function167(because that's where the conflict appears), you can use the following168syntax::169 170 git log -L:'\<function\>':<path> HEAD..<commit>^171 172.. note::173 The ``\<`` and ``\>`` around the function name ensure that the174 matches are anchored on a word boundary. This is important, as this175 part is actually a regex and git only follows the first match, so176 if you use ``-L:thread_stack:kernel/fork.c`` it may only give you177 results for the function ``try_release_thread_stack_to_cache`` even178 though there are many other functions in that file containing the179 string ``thread_stack`` in their names.180 181Another useful option for ``git log`` is ``-G``, which allows you to182filter on certain strings appearing in the diffs of the commits you are183listing::184 185 git log -G'regex' HEAD..<commit>^ -- <path>186 187This can also be a handy way to quickly find when something (e.g. a188function call or a variable) was changed, added, or removed. The search189string is a regular expression, which means you can potentially search190for more specific things like assignments to a specific struct member::191 192 git log -G'\->index\>.*='193 194git blame195~~~~~~~~~196 197Another way to find prerequisite commits (albeit only the most recent198one for a given conflict) is to run ``git blame``. In this case, you199need to run it against the parent commit of the patch you are200cherry-picking and the file where the conflict appeared, i.e.::201 202 git blame <commit>^ -- <path>203 204This command also accepts the ``-L`` argument (for restricting the205output to a single function), but in this case you specify the filename206at the end of the command as usual::207 208 git blame -L:'\<function\>' <commit>^ -- <path>209 210Navigate to the place where the conflict occurred. The first column of211the blame output is the commit ID of the patch that added a given line212of code.213 214It might be a good idea to ``git show`` these commits and see if they215look like they might be the source of the conflict. Sometimes there will216be more than one of these commits, either because multiple commits217changed different lines of the same conflict area *or* because multiple218subsequent patches changed the same line (or lines) multiple times. In219the latter case, you may have to run ``git blame`` again and specify the220older version of the file to look at in order to dig further back in221the history of the file.222 223Prerequisite vs. incidental patches224~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~225 226Having found the patch that caused the conflict, you need to determine227whether it is a prerequisite for the patch you are backporting or228whether it is just incidental and can be skipped. An incidental patch229would be one that touches the same code as the patch you are230backporting, but does not change the semantics of the code in any231material way. For example, a whitespace cleanup patch is completely232incidental -- likewise, a patch that simply renames a function or a233variable would be incidental as well. On the other hand, if the function234being changed does not even exist in your current branch then this would235not be incidental at all and you need to carefully consider whether the236patch adding the function should be cherry-picked first.237 238If you find that there is a necessary prerequisite patch, then you need239to stop and cherry-pick that instead. If you've already resolved some240conflicts in a different file and don't want to do it again, you can241create a temporary copy of that file.242 243To abort the current cherry-pick, go ahead and run244``git cherry-pick --abort``, then restart the cherry-picking process245with the commit ID of the prerequisite patch instead.246 247Understanding conflict markers248------------------------------249 250Combined diffs251~~~~~~~~~~~~~~252 253Let's say you've decided against picking (or reverting) additional254patches and you just want to resolve the conflict. Git will have255inserted conflict markers into your file. Out of the box, this will look256something like::257 258 <<<<<<< HEAD259 this is what's in your current tree before cherry-picking260 =======261 this is what the patch wants it to be after cherry-picking262 >>>>>>> <commit>... title263 264This is what you would see if you opened the file in your editor.265However, if you were to run ``git diff`` without any arguments, the266output would look something like this::267 268 $ git diff269 [...]270 ++<<<<<<<< HEAD271 +this is what's in your current tree before cherry-picking272 ++========273 + this is what the patch wants it to be after cherry-picking274 ++>>>>>>>> <commit>... title275 276When you are resolving a conflict, the behavior of ``git diff`` differs277from its normal behavior. Notice the two columns of diff markers278instead of the usual one; this is a so-called "`combined diff`_", here279showing the 3-way diff (or diff-of-diffs) between280 281#. the current branch (before cherry-picking) and the current working282 directory, and283#. the current branch (before cherry-picking) and the file as it looks284 after the original patch has been applied.285 286.. _combined diff: https://git-scm.com/docs/diff-format#_combined_diff_format287 288 289Better diffs290~~~~~~~~~~~~291 2923-way combined diffs include all the other changes that happened to the293file between your current branch and the branch you are cherry-picking294from. While this is useful for spotting other changes that you need to295take into account, this also makes the output of ``git diff`` somewhat296intimidating and difficult to read. You may instead prefer to run297``git diff HEAD`` (or ``git diff --ours``) which shows only the diff298between the current branch before cherry-picking and the current working299directory. It looks like this::300 301 $ git diff HEAD302 [...]303 +<<<<<<<< HEAD304 this is what's in your current tree before cherry-picking305 +========306 +this is what the patch wants it to be after cherry-picking307 +>>>>>>>> <commit>... title308 309As you can see, this reads just like any other diff and makes it clear310which lines are in the current branch and which lines are being added311because they are part of the merge conflict or the patch being312cherry-picked.313 314Merge styles and diff3315~~~~~~~~~~~~~~~~~~~~~~316 317The default conflict marker style shown above is known as the ``merge``318style. There is also another style available, known as the ``diff3``319style, which looks like this::320 321 <<<<<<< HEAD322 this is what is in your current tree before cherry-picking323 ||||||| parent of <commit> (title)324 this is what the patch expected to find there325 =======326 this is what the patch wants it to be after being applied327 >>>>>>> <commit> (title)328 329As you can see, this has 3 parts instead of 2, and includes what git330expected to find there but didn't. It is *highly recommended* to use331this conflict style as it makes it much clearer what the patch actually332changed; i.e., it allows you to compare the before-and-after versions333of the file for the commit you are cherry-picking. This allows you to334make better decisions about how to resolve the conflict.335 336To change conflict marker styles, you can use the following command::337 338 git config merge.conflictStyle diff3339 340There is a third option, ``zdiff3``, introduced in `Git 2.35`_,341which has the same 3 sections as ``diff3``, but where common lines have342been trimmed off, making the conflict area smaller in some cases.343 344.. _Git 2.35: https://github.blog/2022-01-24-highlights-from-git-2-35/345 346Iterating on conflict resolutions347---------------------------------348 349The first step in any conflict resolution process is to understand the350patch you are backporting. For the Linux kernel this is especially351important, since an incorrect change can lead to the whole system352crashing -- or worse, an undetected security vulnerability.353 354Understanding the patch can be easy or difficult depending on the patch355itself, the changelog, and your familiarity with the code being changed.356However, a good question for every change (or every hunk of the patch)357might be: "Why is this hunk in the patch?" The answers to these358questions will inform your conflict resolution.359 360Resolution process361~~~~~~~~~~~~~~~~~~362 363Sometimes the easiest thing to do is to just remove all but the first364part of the conflict, leaving the file essentially unchanged, and apply365the changes by hand. Perhaps the patch is changing a function call366argument from ``0`` to ``1`` while a conflicting change added an367entirely new (and insignificant) parameter to the end of the parameter368list; in that case, it's easy enough to change the argument from ``0``369to ``1`` by hand and leave the rest of the arguments alone. This370technique of manually applying changes is mostly useful if the conflict371pulled in a lot of unrelated context that you don't really need to care372about.373 374For particularly nasty conflicts with many conflict markers, you can use375``git add`` or ``git add -i`` to selectively stage your resolutions to376get them out of the way; this also lets you use ``git diff HEAD`` to377always see what remains to be resolved or ``git diff --cached`` to see378what your patch looks like so far.379 380Dealing with file renames381~~~~~~~~~~~~~~~~~~~~~~~~~382 383One of the most annoying things that can happen while backporting a384patch is discovering that one of the files being patched has been385renamed, as that typically means git won't even put in conflict markers,386but will just throw up its hands and say (paraphrased): "Unmerged path!387You do the work..."388 389There are generally a few ways to deal with this. If the patch to the390renamed file is small, like a one-line change, the easiest thing is to391just go ahead and apply the change by hand and be done with it. On the392other hand, if the change is big or complicated, you definitely don't393want to do it by hand.394 395As a first pass, you can try something like this, which will lower the396rename detection threshold to 30% (by default, git uses 50%, meaning397that two files need to have at least 50% in common for it to consider398an add-delete pair to be a potential rename)::399 400 git cherry-pick -strategy=recursive -Xrename-threshold=30401 402Sometimes the right thing to do will be to also backport the patch that403did the rename, but that's definitely not the most common case. Instead,404what you can do is to temporarily rename the file in the branch you're405backporting to (using ``git mv`` and committing the result), restart the406attempt to cherry-pick the patch, rename the file back (``git mv`` and407committing again), and finally squash the result using ``git rebase -i``408(see the `rebase tutorial`_) so it appears as a single commit when you409are done.410 411.. _rebase tutorial: https://medium.com/@slamflipstrom/a-beginners-guide-to-squashing-commits-with-git-rebase-8185cf6e62ec412 413Gotchas414-------415 416Function arguments417~~~~~~~~~~~~~~~~~~418 419Pay attention to changing function arguments! It's easy to gloss over420details and think that two lines are the same but actually they differ421in some small detail like which variable was passed as an argument422(especially if the two variables are both a single character that look423the same, like i and j).424 425Error handling426~~~~~~~~~~~~~~427 428If you cherry-pick a patch that includes a ``goto`` statement (typically429for error handling), it is absolutely imperative to double check that430the target label is still correct in the branch you are backporting to.431The same goes for added ``return``, ``break``, and ``continue``432statements.433 434Error handling is typically located at the bottom of the function, so it435may not be part of the conflict even though could have been changed by436other patches.437 438A good way to ensure that you review the error paths is to always use439``git diff -W`` and ``git show -W`` (AKA ``--function-context``) when440inspecting your changes. For C code, this will show you the whole441function that's being changed in a patch. One of the things that often442go wrong during backports is that something else in the function changed443on either of the branches that you're backporting from or to. By444including the whole function in the diff you get more context and can445more easily spot problems that might otherwise go unnoticed.446 447Refactored code448~~~~~~~~~~~~~~~449 450Something that happens quite often is that code gets refactored by451"factoring out" a common code sequence or pattern into a helper452function. When backporting patches to an area where such a refactoring453has taken place, you effectively need to do the reverse when454backporting: a patch to a single location may need to be applied to455multiple locations in the backported version. (One giveaway for this456scenario is that a function was renamed -- but that's not always the457case.)458 459To avoid incomplete backports, it's worth trying to figure out if the460patch fixes a bug that appears in more than one place. One way to do461this would be to use ``git grep``. (This is actually a good idea to do462in general, not just for backports.) If you do find that the same kind463of fix would apply to other places, it's also worth seeing if those464places exist upstream -- if they don't, it's likely the patch may need465to be adjusted. ``git log`` is your friend to figure out what happened466to these areas as ``git blame`` won't show you code that has been467removed.468 469If you do find other instances of the same pattern in the upstream tree470and you're not sure whether it's also a bug, it may be worth asking the471patch author. It's not uncommon to find new bugs during backporting!472 473Verifying the result474====================475 476colordiff477---------478 479Having committed a conflict-free new patch, you can now compare your480patch to the original patch. It is highly recommended that you use a481tool such as `colordiff`_ that can show two files side by side and color482them according to the changes between them::483 484 colordiff -yw -W 200 <(git diff -W <upstream commit>^-) <(git diff -W HEAD^-) | less -SR485 486.. _colordiff: https://www.colordiff.org/487 488Here, ``-y`` means to do a side-by-side comparison; ``-w`` ignores489whitespace, and ``-W 200`` sets the width of the output (as otherwise it490will use 130 by default, which is often a bit too little).491 492The ``rev^-`` syntax is a handy shorthand for ``rev^..rev``, essentially493giving you just the diff for that single commit; also see494the official `git rev-parse documentation`_.495 496.. _git rev-parse documentation: https://git-scm.com/docs/git-rev-parse#_other_rev_parent_shorthand_notations497 498Again, note the inclusion of ``-W`` for ``git diff``; this ensures that499you will see the full function for any function that has changed.500 501One incredibly important thing that colordiff does is to highlight lines502that are different. For example, if an error-handling ``goto`` has503changed labels between the original and backported patch, colordiff will504show these side-by-side but highlighted in a different color. Thus, it505is easy to see that the two ``goto`` statements are jumping to different506labels. Likewise, lines that were not modified by either patch but507differ in the context will also be highlighted and thus stand out during508a manual inspection.509 510Of course, this is just a visual inspection; the real test is building511and running the patched kernel (or program).512 513Build testing514-------------515 516We won't cover runtime testing here, but it can be a good idea to build517just the files touched by the patch as a quick sanity check. For the518Linux kernel you can build single files like this, assuming you have the519``.config`` and build environment set up correctly::520 521 make path/to/file.o522 523Note that this won't discover linker errors, so you should still do a524full build after verifying that the single file compiles. By compiling525the single file first you can avoid having to wait for a full build *in526case* there are compiler errors in any of the files you've changed.527 528Runtime testing529---------------530 531Even a successful build or boot test is not necessarily enough to rule532out a missing dependency somewhere. Even though the chances are small,533there could be code changes where two independent changes to the same534file result in no conflicts, no compile-time errors, and runtime errors535only in exceptional cases.536 537One concrete example of this was a pair of patches to the system call538entry code where the first patch saved/restored a register and a later539patch made use of the same register somewhere in the middle of this540sequence. Since there was no overlap between the changes, one could541cherry-pick the second patch, have no conflicts, and believe that542everything was fine, when in fact the code was now scribbling over an543unsaved register.544 545Although the vast majority of errors will be caught during compilation546or by superficially exercising the code, the only way to *really* verify547a backport is to review the final patch with the same level of scrutiny548as you would (or should) give to any other patch. Having unit tests and549regression tests or other types of automatic testing can help increase550the confidence in the correctness of a backport.551 552Submitting backports to stable553==============================554 555As the stable maintainers try to cherry-pick mainline fixes onto their556stable kernels, they may send out emails asking for backports when when557encountering conflicts, see e.g.558<https://lore.kernel.org/stable/2023101528-jawed-shelving-071a@gregkh/>.559These emails typically include the exact steps you need to cherry-pick560the patch to the correct tree and submit the patch.561 562One thing to make sure is that your changelog conforms to the expected563format::564 565 <original patch title>566 567 [ Upstream commit <mainline rev> ]568 569 <rest of the original changelog>570 [ <summary of the conflicts and their resolutions> ]571 Signed-off-by: <your name and email>572 573The "Upstream commit" line is sometimes slightly different depending on574the stable version. Older version used this format::575 576 commit <mainline rev> upstream.577 578It is most common to indicate the kernel version the patch applies to579in the email subject line (using e.g.580``git send-email --subject-prefix='PATCH 6.1.y'``), but you can also put581it in the Signed-off-by:-area or below the ``---`` line.582 583The stable maintainers expect separate submissions for each active584stable version, and each submission should also be tested separately.585 586A few final words of advice587===========================588 5891) Approach the backporting process with humility.5902) Understand the patch you are backporting; this means reading both591 the changelog and the code.5923) Be honest about your confidence in the result when submitting the593 patch.5944) Ask relevant maintainers for explicit acks.595 596Examples597========598 599The above shows roughly the idealized process of backporting a patch.600For a more concrete example, see this video tutorial where two patches601are backported from mainline to stable:602`Backporting Linux Kernel Patches`_.603 604.. _Backporting Linux Kernel Patches: https://youtu.be/sBR7R1V2FeA605