English

5. KVM x86

5.1. Foreword

KVM strives to be a welcoming community; contributions from newcomers are valued and encouraged. Please do not be discouraged or intimidated by the length of this document and the many rules/guidelines it contains. Everyone makes mistakes, and everyone was a newbie at some point. So long as you make an honest effort to follow KVM x86’s guidelines, are receptive to feedback, and learn from any mistakes you make, you will be welcomed with open arms, not torches and pitchforks.

5.2. TL;DR

Testing is mandatory. Be consistent with established styles and patterns.

5.3. Trees

KVM x86 is currently in a transition period from being part of the main KVM tree, to being “just another KVM arch”. As such, KVM x86 is split across the main KVM tree, git.kernel.org/pub/scm/virt/kvm/kvm.git, and a KVM x86 specific tree, github.com/kvm-x86/linux.git.

Generally speaking, fixes for the current cycle are applied directly to the main KVM tree, while all development for the next cycle is routed through the KVM x86 tree. In the unlikely event that a fix for the current cycle is routed through the KVM x86 tree, it will be applied to the fixes branch before making its way to the main KVM tree.

Note, this transition period is expected to last quite some time, i.e. will be the status quo for the foreseeable future.

5.3.1. Branches

The KVM x86 tree is organized into multiple topic branches. The purpose of using finer-grained topic branches is to make it easier to keep tabs on an area of development, and to limit the collateral damage of human errors and/or buggy commits, e.g. dropping the HEAD commit of a topic branch has no impact on other in-flight commits’ SHA1 hashes, and having to reject a pull request due to bugs delays only that topic branch.

All topic branches, except for next and fixes, are rolled into next via a Cthulhu merge on an as-needed basis, i.e. when a topic branch is updated. As a result, force pushes to next are common.

5.3.2. Lifecycle

Fixes that target the current release, a.k.a. mainline, are typically applied directly to the main KVM tree, i.e. do not route through the KVM x86 tree.

Changes that target the next release are routed through the KVM x86 tree. Pull requests (from KVM x86 to main KVM) are sent for each KVM x86 topic branch, typically the week before Linus’ opening of the merge window, e.g. the week following rc7 for “normal” releases. If all goes well, the topic branches are rolled into the main KVM pull request sent during Linus’ merge window.

The KVM x86 tree doesn’t have its own official merge window, but there’s a soft close around rc5 for new features, and a soft close around rc6 for fixes (for the next release; see above for fixes that target the current release).

5.3.3. Timeline

Submissions are typically reviewed and applied in FIFO order, with some wiggle room for the size of a series, patches that are “cache hot”, etc. Fixes, especially for the current release and or stable trees, get to jump the queue. Patches that will be taken through a non-KVM tree (most often through the tip tree) and/or have other acks/reviews also jump the queue to some extent.

Note, the vast majority of review is done between rc1 and rc6, give or take. The period between rc6 and the next rc1 is used to catch up on other tasks, i.e. radio silence during this period isn’t unusual.

Pings to get a status update are welcome, but keep in mind the timing of the current release cycle and have realistic expectations. If you are pinging for acceptance, i.e. not just for feedback or an update, please do everything you can, within reason, to ensure that your patches are ready to be merged! Pings on series that break the build or fail tests lead to unhappy maintainers!

5.4. Development

5.4.1. Base Tree/Branch

Fixes that target the current release, a.k.a. mainline, should be based on git://git.kernel.org/pub/scm/virt/kvm/kvm.git master. Note, fixes do not automatically warrant inclusion in the current release. There is no singular rule, but typically only fixes for bugs that are urgent, critical, and/or were introduced in the current release should target the current release.

Everything else should be based on kvm-x86/next, i.e. there is no need to select a specific topic branch as the base. If there are conflicts and/or dependencies across topic branches, it is the maintainer’s job to sort them out.

The only exception to using kvm-x86/next as the base is if a patch/series is a multi-arch series, i.e. has non-trivial modifications to common KVM code and/or has more than superficial changes to other architectures’ code. Multi- arch patch/series should instead be based on a common, stable point in KVM’s history, e.g. the release candidate upon which kvm-x86 next is based. If you’re unsure whether a patch/series is truly multi-arch, err on the side of caution and treat it as multi-arch, i.e. use a common base.

5.4.2. Coding Style

When it comes to style, naming, patterns, etc., consistency is the number one priority in KVM x86. If all else fails, match what already exists.

With a few caveats listed below, follow the tip tree maintainers’ preferred Coding style notes, as patches/series often touch both KVM and non-KVM x86 files, i.e. draw the attention of KVM and tip tree maintainers.

Using reverse fir tree, a.k.a. reverse Christmas tree or reverse XMAS tree, for variable declarations isn’t strictly required, though it is still preferred.

Except for a handful of special snowflakes, do not use kernel-doc comments for functions. The vast majority of “public” KVM functions aren’t truly public as they are intended only for KVM-internal consumption (there are plans to privatize KVM’s headers and exports to enforce this).

5.4.3. Comments

Write comments using imperative mood and avoid pronouns. Use comments to provide a high level overview of the code, and/or to explain why the code does what it does. Do not reiterate what the code literally does; let the code speak for itself. If the code itself is inscrutable, comments will not help.

5.4.4. SDM and APM References

Much of KVM’s code base is directly tied to architectural behavior defined in Intel’s Software Development Manual (SDM) and AMD’s Architecture Programmer’s Manual (APM). Use of “Intel’s SDM” and “AMD’s APM”, or even just “SDM” or “APM”, without additional context is a-ok.

Do not reference specific sections, tables, figures, etc. by number, especially not in comments. Instead, if necessary (see below), copy-paste the relevant snippet and reference sections/tables/figures by name. The layouts of the SDM and APM are constantly changing, and so the numbers/labels aren’t stable.

Generally speaking, do not explicitly reference or copy-paste from the SDM or APM in comments. With few exceptions, KVM must honor architectural behavior, therefore it’s implied that KVM behavior is emulating SDM and/or APM behavior. Note, referencing the SDM/APM in changelogs to justify the change and provide context is perfectly ok and encouraged.

5.4.5. Shortlog

The preferred prefix format is KVM: <topic>:, where <topic> is one of:

- x86
- x86/mmu
- x86/pmu
- x86/xen
- selftests
- SVM
- nSVM
- VMX
- nVMX

DO NOT use x86/kvm! x86/kvm is used exclusively for Linux-as-a-KVM-guest changes, i.e. for arch/x86/kernel/kvm.c. Do not use file names or complete file paths as the subject/shortlog prefix.

Note, these don’t align with the topics branches (the topic branches care much more about code conflicts).

All names are case sensitive! KVM: x86: is good, kvm: vmx: is not.

Capitalize the first word of the condensed patch description, but omit ending punctionation. E.g.:

KVM: x86: Fix a null pointer dereference in function_xyz()

not:

kvm: x86: fix a null pointer dereference in function_xyz.

If a patch touches multiple topics, traverse up the conceptual tree to find the first common parent (which is often simply x86). When in doubt, git log path/to/file should provide a reasonable hint.

New topics do occasionally pop up, but please start an on-list discussion if you want to propose introducing a new topic, i.e. don’t go rogue.

See The canonical patch format for more information, with one amendment: do not treat the 70-75 character limit as an absolute, hard limit. Instead, use 75 characters as a firm-but-not-hard limit, and use 80 characters as a hard limit. I.e. let the shortlog run a few characters over the standard limit if you have good reason to do so.

5.4.6. Changelog

Most importantly, write changelogs using imperative mood and avoid pronouns.

See Describe your changes for more information, with one amendment: lead with a short blurb on the actual changes, and then follow up with the context and background. Note! This order directly conflicts with the tip tree’s preferred approach! Please follow the tip tree’s preferred style when sending patches that primarily target arch/x86 code that is _NOT_ KVM code.

Stating what a patch does before diving into details is preferred by KVM x86 for several reasons. First and foremost, what code is actually being changed is arguably the most important information, and so that info should be easy to find. Changelogs that bury the “what’s actually changing” in a one-liner after 3+ paragraphs of background make it very hard to find that information.

For initial review, one could argue the “what’s broken” is more important, but for skimming logs and git archaeology, the gory details matter less and less. E.g. when doing a series of “git blame”, the details of each change along the way are useless, the details only matter for the culprit. Providing the “what changed” makes it easy to quickly determine whether or not a commit might be of interest.

Another benefit of stating “what’s changing” first is that it’s almost always possible to state “what’s changing” in a single sentence. Conversely, all but the most simple bugs require multiple sentences or paragraphs to fully describe the problem. If both the “what’s changing” and “what’s the bug” are super short then the order doesn’t matter. But if one is shorter (almost always the “what’s changing), then covering the shorter one first is advantageous because it’s less of an inconvenience for readers/reviewers that have a strict ordering preference. E.g. having to skip one sentence to get to the context is less painful than having to skip three paragraphs to get to “what’s changing”.

5.4.7. Fixes

If a change fixes a KVM/kernel bug, add a Fixes: tag even if the change doesn’t need to be backported to stable kernels, and even if the change fixes a bug in an older release.

Conversely, if a fix does need to be backported, explicitly tag the patch with “Cc: stable@vger.kernel” (though the email itself doesn’t need to Cc: stable); KVM x86 opts out of backporting Fixes: by default. Some auto-selected patches do get backported, but require explicit maintainer approval (search MANUALSEL).

5.4.8. Function References

When a function is mentioned in a comment, changelog, or shortlog (or anywhere for that matter), use the format function_name(). The parentheses provide context and disambiguate the reference.

5.5. Testing

At a bare minimum, all patches in a series must build cleanly for KVM_INTEL=m KVM_AMD=m, and KVM_WERROR=y. Building every possible combination of Kconfigs isn’t feasible, but the more the merrier. KVM_SMM, KVM_XEN, PROVE_LOCKING, and X86_64 are particularly interesting knobs to turn.

Running KVM selftests and KVM-unit-tests is also mandatory (and stating the obvious, the tests need to pass). The only exception is for changes that have negligible probability of affecting runtime behavior, e.g. patches that only modify comments. When possible and relevant, testing on both Intel and AMD is strongly preferred. Booting an actual VM is encouraged, but not mandatory.

For changes that touch KVM’s shadow paging code, running with TDP (EPT/NPT) disabled is mandatory. For changes that affect common KVM MMU code, running with TDP disabled is strongly encouraged. For all other changes, if the code being modified depends on and/or interacts with a module param, testing with the relevant settings is mandatory.

Note, KVM selftests and KVM-unit-tests do have known failures. If you suspect a failure is not due to your changes, verify that the exact same failure occurs with and without your changes.

Changes that touch reStructured Text documentation, i.e. .rst files, must build htmldocs cleanly, i.e. with no new warnings or errors.

If you can’t fully test a change, e.g. due to lack of hardware, clearly state what level of testing you were able to do, e.g. in the cover letter.

5.5.1. New Features

With one exception, new features must come with test coverage. KVM specific tests aren’t strictly required, e.g. if coverage is provided by running a sufficiently enabled guest VM, or by running a related kernel selftest in a VM, but dedicated KVM tests are preferred in all cases. Negative testcases in particular are mandatory for enabling of new hardware features as error and exception flows are rarely exercised simply by running a VM.

The only exception to this rule is if KVM is simply advertising support for a feature via KVM_GET_SUPPORTED_CPUID, i.e. for instructions/features that KVM can’t prevent a guest from using and for which there is no true enabling.

Note, “new features” does not just mean “new hardware features”! New features that can’t be well validated using existing KVM selftests and/or KVM-unit-tests must come with tests.

Posting new feature development without tests to get early feedback is more than welcome, but such submissions should be tagged RFC, and the cover letter should clearly state what type of feedback is requested/expected. Do not abuse the RFC process; RFCs will typically not receive in-depth review.

5.5.2. Bug Fixes

Except for “obvious” found-by-inspection bugs, fixes must be accompanied by a reproducer for the bug being fixed. In many cases the reproducer is implicit, e.g. for build errors and test failures, but it should still be clear to readers what is broken and how to verify the fix. Some leeway is given for bugs that are found via non-public workloads/tests, but providing regression tests for such bugs is strongly preferred.

In general, regression tests are preferred for any bug that is not trivial to hit. E.g. even if the bug was originally found by a fuzzer such as syzkaller, a targeted regression test may be warranted if the bug requires hitting a one-in-a-million type race condition.

Note, KVM bugs are rarely urgent and non-trivial to reproduce. Ask yourself if a bug is really truly the end of the world before posting a fix without a reproducer.

5.6. Posting

5.6.2. Git Base

If you are using git version 2.9.0 or later (Googlers, this is all of you!), use git format-patch with the --base flag to automatically include the base tree information in the generated patches.

Note, --base=auto works as expected if and only if a branch’s upstream is set to the base topic branch, e.g. it will do the wrong thing if your upstream is set to your personal repository for backup purposes. An alternative “auto” solution is to derive the names of your development branches based on their KVM x86 topic, and feed that into --base. E.g. x86/pmu/my_branch_name, and then write a small wrapper to extract pmu from the current branch name to yield --base=x/pmu, where x is whatever name your repository uses to track the KVM x86 remote.

5.6.3. Co-Posting Tests

KVM selftests that are associated with KVM changes, e.g. regression tests for bug fixes, should be posted along with the KVM changes as a single series. The standard kernel rules for bisection apply, i.e. KVM changes that result in test failures should be ordered after the selftests updates, and vice versa, new tests that fail due to KVM bugs should be ordered after the KVM fixes.

KVM-unit-tests should always be posted separately. Tools, e.g. b4 am, don’t know that KVM-unit-tests is a separate repository and get confused when patches in a series apply on different trees. To tie KVM-unit-tests patches back to KVM patches, first post the KVM changes and then provide a lore Link: to the KVM patch/series in the KVM-unit-tests patch(es).

5.7. Notifications

When a patch/series is officially accepted, a notification email will be sent in reply to the original posting (cover letter for multi-patch series). The notification will include the tree and topic branch, along with the SHA1s of the commits of applied patches.

If a subset of patches is applied, this will be clearly stated in the notification. Unless stated otherwise, it’s implied that any patches in the series that were not accepted need more work and should be submitted in a new version.

If for some reason a patch is dropped after officially being accepted, a reply will be sent to the notification email explaining why the patch was dropped, as well as the next steps.

5.7.1. SHA1 Stability

SHA1s are not 100% guaranteed to be stable until they land in Linus’ tree! A SHA1 is usually stable once a notification has been sent, but things happen. In most cases, an update to the notification email be provided if an applied patch’s SHA1 changes. However, in some scenarios, e.g. if all KVM x86 branches need to be rebased, individual notifications will not be given.

5.8. Vulnerabilities

Bugs that can be exploited by the guest to attack the host (kernel or userspace), or that can be exploited by a nested VM to its host (L2 attacking L1), are of particular interest to KVM. Please follow the protocol for Security bugs if you suspect a bug can lead to an escape, data leak, etc.