1.. SPDX-License-Identifier: GPL-2.0
2
3KVM x86
4=======
5
6Foreword
7--------
8KVM strives to be a welcoming community; contributions from newcomers are
9valued and encouraged.  Please do not be discouraged or intimidated by the
10length of this document and the many rules/guidelines it contains.  Everyone
11makes mistakes, and everyone was a newbie at some point.  So long as you make
12an honest effort to follow KVM x86's guidelines, are receptive to feedback,
13and learn from any mistakes you make, you will be welcomed with open arms, not
14torches and pitchforks.
15
16TL;DR
17-----
18Testing is mandatory.  Be consistent with established styles and patterns.
19
20Trees
21-----
22KVM x86 is currently in a transition period from being part of the main KVM
23tree, to being "just another KVM arch".  As such, KVM x86 is split across the
24main KVM tree, ``git.kernel.org/pub/scm/virt/kvm/kvm.git``, and a KVM x86
25specific tree, ``github.com/kvm-x86/linux.git``.
26
27Generally speaking, fixes for the current cycle are applied directly to the
28main KVM tree, while all development for the next cycle is routed through the
29KVM x86 tree.  In the unlikely event that a fix for the current cycle is routed
30through the KVM x86 tree, it will be applied to the ``fixes`` branch before
31making its way to the main KVM tree.
32
33Note, this transition period is expected to last quite some time, i.e. will be
34the status quo for the foreseeable future.
35
36Branches
37~~~~~~~~
38The KVM x86 tree is organized into multiple topic branches.  The purpose of
39using finer-grained topic branches is to make it easier to keep tabs on an area
40of development, and to limit the collateral damage of human errors and/or buggy
41commits, e.g. dropping the HEAD commit of a topic branch has no impact on other
42in-flight commits' SHA1 hashes, and having to reject a pull request due to bugs
43delays only that topic branch.
44
45All topic branches, except for ``next`` and ``fixes``, are rolled into ``next``
46via a Cthulhu merge on an as-needed basis, i.e. when a topic branch is updated.
47As a result, force pushes to ``next`` are common.
48
49Lifecycle
50~~~~~~~~~
51Fixes that target the current release, a.k.a. mainline, are typically applied
52directly to the main KVM tree, i.e. do not route through the KVM x86 tree.
53
54Changes that target the next release are routed through the KVM x86 tree.  Pull
55requests (from KVM x86 to main KVM) are sent for each KVM x86 topic branch,
56typically the week before Linus' opening of the merge window, e.g. the week
57following rc7 for "normal" releases.  If all goes well, the topic branches are
58rolled into the main KVM pull request sent during Linus' merge window.
59
60The KVM x86 tree doesn't have its own official merge window, but there's a soft
61close around rc5 for new features, and a soft close around rc6 for fixes (for
62the next release; see above for fixes that target the current release).
63
64Timeline
65~~~~~~~~
66Submissions are typically reviewed and applied in FIFO order, with some wiggle
67room for the size of a series, patches that are "cache hot", etc.  Fixes,
68especially for the current release and or stable trees, get to jump the queue.
69Patches that will be taken through a non-KVM tree (most often through the tip
70tree) and/or have other acks/reviews also jump the queue to some extent.
71
72Note, the vast majority of review is done between rc1 and rc6, give or take.
73The period between rc6 and the next rc1 is used to catch up on other tasks,
74i.e. radio silence during this period isn't unusual.
75
76Pings to get a status update are welcome, but keep in mind the timing of the
77current release cycle and have realistic expectations.  If you are pinging for
78acceptance, i.e. not just for feedback or an update, please do everything you
79can, within reason, to ensure that your patches are ready to be merged!  Pings
80on series that break the build or fail tests lead to unhappy maintainers!
81
82Development
83-----------
84
85Base Tree/Branch
86~~~~~~~~~~~~~~~~
87Fixes that target the current release, a.k.a. mainline, should be based on
88``git://git.kernel.org/pub/scm/virt/kvm/kvm.git master``.  Note, fixes do not
89automatically warrant inclusion in the current release.  There is no singular
90rule, but typically only fixes for bugs that are urgent, critical, and/or were
91introduced in the current release should target the current release.
92
93Everything else should be based on ``kvm-x86/next``, i.e. there is no need to
94select a specific topic branch as the base.  If there are conflicts and/or
95dependencies across topic branches, it is the maintainer's job to sort them
96out.
97
98The only exception to using ``kvm-x86/next`` as the base is if a patch/series
99is a multi-arch series, i.e. has non-trivial modifications to common KVM code
100and/or has more than superficial changes to other architectures' code.  Multi-
101arch patch/series should instead be based on a common, stable point in KVM's
102history, e.g. the release candidate upon which ``kvm-x86 next`` is based.  If
103you're unsure whether a patch/series is truly multi-arch, err on the side of
104caution and treat it as multi-arch, i.e. use a common base.
105
106Coding Style
107~~~~~~~~~~~~
108When it comes to style, naming, patterns, etc., consistency is the number one
109priority in KVM x86.  If all else fails, match what already exists.
110
111With a few caveats listed below, follow the tip tree maintainers' preferred
112:ref:`maintainer-tip-coding-style`, as patches/series often touch both KVM and
113non-KVM x86 files, i.e. draw the attention of KVM *and* tip tree maintainers.
114
115Using reverse fir tree, a.k.a. reverse Christmas tree or reverse XMAS tree, for
116variable declarations isn't strictly required, though it is still preferred.
117
118Except for a handful of special snowflakes, do not use kernel-doc comments for
119functions.  The vast majority of "public" KVM functions aren't truly public as
120they are intended only for KVM-internal consumption (there are plans to
121privatize KVM's headers and exports to enforce this).
122
123Comments
124~~~~~~~~
125Write comments using imperative mood and avoid pronouns.  Use comments to
126provide a high level overview of the code, and/or to explain why the code does
127what it does.  Do not reiterate what the code literally does; let the code
128speak for itself.  If the code itself is inscrutable, comments will not help.
129
130SDM and APM References
131~~~~~~~~~~~~~~~~~~~~~~
132Much of KVM's code base is directly tied to architectural behavior defined in
133Intel's Software Development Manual (SDM) and AMD's Architecture Programmer’s
134Manual (APM).  Use of "Intel's SDM" and "AMD's APM", or even just "SDM" or
135"APM", without additional context is a-ok.
136
137Do not reference specific sections, tables, figures, etc. by number, especially
138not in comments.  Instead, if necessary (see below), copy-paste the relevant
139snippet and reference sections/tables/figures by name.  The layouts of the SDM
140and APM are constantly changing, and so the numbers/labels aren't stable.
141
142Generally speaking, do not explicitly reference or copy-paste from the SDM or
143APM in comments.  With few exceptions, KVM *must* honor architectural behavior,
144therefore it's implied that KVM behavior is emulating SDM and/or APM behavior.
145Note, referencing the SDM/APM in changelogs to justify the change and provide
146context is perfectly ok and encouraged.
147
148Shortlog
149~~~~~~~~
150The preferred prefix format is ``KVM: <topic>:``, where ``<topic>`` is one of::
151
152  - x86
153  - x86/mmu
154  - x86/pmu
155  - x86/xen
156  - selftests
157  - SVM
158  - nSVM
159  - VMX
160  - nVMX
161
162**DO NOT use x86/kvm!**  ``x86/kvm`` is used exclusively for Linux-as-a-KVM-guest
163changes, i.e. for arch/x86/kernel/kvm.c.  Do not use file names or complete file
164paths as the subject/shortlog prefix.
165
166Note, these don't align with the topics branches (the topic branches care much
167more about code conflicts).
168
169All names are case sensitive!  ``KVM: x86:`` is good, ``kvm: vmx:`` is not.
170
171Capitalize the first word of the condensed patch description, but omit ending
172punctionation.  E.g.::
173
174    KVM: x86: Fix a null pointer dereference in function_xyz()
175
176not::
177
178    kvm: x86: fix a null pointer dereference in function_xyz.
179
180If a patch touches multiple topics, traverse up the conceptual tree to find the
181first common parent (which is often simply ``x86``).  When in doubt,
182``git log path/to/file`` should provide a reasonable hint.
183
184New topics do occasionally pop up, but please start an on-list discussion if
185you want to propose introducing a new topic, i.e. don't go rogue.
186
187See :ref:`the_canonical_patch_format` for more information, with one amendment:
188do not treat the 70-75 character limit as an absolute, hard limit.  Instead,
189use 75 characters as a firm-but-not-hard limit, and use 80 characters as a hard
190limit.  I.e. let the shortlog run a few characters over the standard limit if
191you have good reason to do so.
192
193Changelog
194~~~~~~~~~
195Most importantly, write changelogs using imperative mood and avoid pronouns.
196
197See :ref:`describe_changes` for more information, with one amendment: lead with
198a short blurb on the actual changes, and then follow up with the context and
199background.  Note!  This order directly conflicts with the tip tree's preferred
200approach!  Please follow the tip tree's preferred style when sending patches
201that primarily target arch/x86 code that is _NOT_ KVM code.
202
203Stating what a patch does before diving into details is preferred by KVM x86
204for several reasons.  First and foremost, what code is actually being changed
205is arguably the most important information, and so that info should be easy to
206find. Changelogs that bury the "what's actually changing" in a one-liner after
2073+ paragraphs of background make it very hard to find that information.
208
209For initial review, one could argue the "what's broken" is more important, but
210for skimming logs and git archaeology, the gory details matter less and less.
211E.g. when doing a series of "git blame", the details of each change along the
212way are useless, the details only matter for the culprit.  Providing the "what
213changed" makes it easy to quickly determine whether or not a commit might be of
214interest.
215
216Another benefit of stating "what's changing" first is that it's almost always
217possible to state "what's changing" in a single sentence.  Conversely, all but
218the most simple bugs require multiple sentences or paragraphs to fully describe
219the problem.  If both the "what's changing" and "what's the bug" are super
220short then the order doesn't matter.  But if one is shorter (almost always the
221"what's changing), then covering the shorter one first is advantageous because
222it's less of an inconvenience for readers/reviewers that have a strict ordering
223preference.  E.g. having to skip one sentence to get to the context is less
224painful than having to skip three paragraphs to get to "what's changing".
225
226Fixes
227~~~~~
228If a change fixes a KVM/kernel bug, add a Fixes: tag even if the change doesn't
229need to be backported to stable kernels, and even if the change fixes a bug in
230an older release.
231
232Conversely, if a fix does need to be backported, explicitly tag the patch with
233"Cc: stable@vger.kernel" (though the email itself doesn't need to Cc: stable);
234KVM x86 opts out of backporting Fixes: by default.  Some auto-selected patches
235do get backported, but require explicit maintainer approval (search MANUALSEL).
236
237Function References
238~~~~~~~~~~~~~~~~~~~
239When a function is mentioned in a comment, changelog, or shortlog (or anywhere
240for that matter), use the format ``function_name()``.  The parentheses provide
241context and disambiguate the reference.
242
243Testing
244-------
245At a bare minimum, *all* patches in a series must build cleanly for KVM_INTEL=m
246KVM_AMD=m, and KVM_WERROR=y.  Building every possible combination of Kconfigs
247isn't feasible, but the more the merrier.  KVM_SMM, KVM_XEN, PROVE_LOCKING, and
248X86_64 are particularly interesting knobs to turn.
249
250Running KVM selftests and KVM-unit-tests is also mandatory (and stating the
251obvious, the tests need to pass).  The only exception is for changes that have
252negligible probability of affecting runtime behavior, e.g. patches that only
253modify comments.  When possible and relevant, testing on both Intel and AMD is
254strongly preferred.  Booting an actual VM is encouraged, but not mandatory.
255
256For changes that touch KVM's shadow paging code, running with TDP (EPT/NPT)
257disabled is mandatory.  For changes that affect common KVM MMU code, running
258with TDP disabled is strongly encouraged.  For all other changes, if the code
259being modified depends on and/or interacts with a module param, testing with
260the relevant settings is mandatory.
261
262Note, KVM selftests and KVM-unit-tests do have known failures.  If you suspect
263a failure is not due to your changes, verify that the *exact same* failure
264occurs with and without your changes.
265
266Changes that touch reStructured Text documentation, i.e. .rst files, must build
267htmldocs cleanly, i.e. with no new warnings or errors.
268
269If you can't fully test a change, e.g. due to lack of hardware, clearly state
270what level of testing you were able to do, e.g. in the cover letter.
271
272New Features
273~~~~~~~~~~~~
274With one exception, new features *must* come with test coverage.  KVM specific
275tests aren't strictly required, e.g. if coverage is provided by running a
276sufficiently enabled guest VM, or by running a related kernel selftest in a VM,
277but dedicated KVM tests are preferred in all cases.  Negative testcases in
278particular are mandatory for enabling of new hardware features as error and
279exception flows are rarely exercised simply by running a VM.
280
281The only exception to this rule is if KVM is simply advertising support for a
282feature via KVM_GET_SUPPORTED_CPUID, i.e. for instructions/features that KVM
283can't prevent a guest from using and for which there is no true enabling.
284
285Note, "new features" does not just mean "new hardware features"!  New features
286that can't be well validated using existing KVM selftests and/or KVM-unit-tests
287must come with tests.
288
289Posting new feature development without tests to get early feedback is more
290than welcome, but such submissions should be tagged RFC, and the cover letter
291should clearly state what type of feedback is requested/expected.  Do not abuse
292the RFC process; RFCs will typically not receive in-depth review.
293
294Bug Fixes
295~~~~~~~~~
296Except for "obvious" found-by-inspection bugs, fixes must be accompanied by a
297reproducer for the bug being fixed.  In many cases the reproducer is implicit,
298e.g. for build errors and test failures, but it should still be clear to
299readers what is broken and how to verify the fix.  Some leeway is given for
300bugs that are found via non-public workloads/tests, but providing regression
301tests for such bugs is strongly preferred.
302
303In general, regression tests are preferred for any bug that is not trivial to
304hit.  E.g. even if the bug was originally found by a fuzzer such as syzkaller,
305a targeted regression test may be warranted if the bug requires hitting a
306one-in-a-million type race condition.
307
308Note, KVM bugs are rarely urgent *and* non-trivial to reproduce.  Ask yourself
309if a bug is really truly the end of the world before posting a fix without a
310reproducer.
311
312Posting
313-------
314
315Links
316~~~~~
317Do not explicitly reference bug reports, prior versions of a patch/series, etc.
318via ``In-Reply-To:`` headers.  Using ``In-Reply-To:`` becomes an unholy mess
319for large series and/or when the version count gets high, and ``In-Reply-To:``
320is useless for anyone that doesn't have the original message, e.g. if someone
321wasn't Cc'd on the bug report or if the list of recipients changes between
322versions.
323
324To link to a bug report, previous version, or anything of interest, use lore
325links.  For referencing previous version(s), generally speaking do not include
326a Link: in the changelog as there is no need to record the history in git, i.e.
327put the link in the cover letter or in the section git ignores.  Do provide a
328formal Link: for bug reports and/or discussions that led to the patch.  The
329context of why a change was made is highly valuable for future readers.
330
331Git Base
332~~~~~~~~
333If you are using git version 2.9.0 or later (Googlers, this is all of you!),
334use ``git format-patch`` with the ``--base`` flag to automatically include the
335base tree information in the generated patches.
336
337Note, ``--base=auto`` works as expected if and only if a branch's upstream is
338set to the base topic branch, e.g. it will do the wrong thing if your upstream
339is set to your personal repository for backup purposes.  An alternative "auto"
340solution is to derive the names of your development branches based on their
341KVM x86 topic, and feed that into ``--base``.  E.g. ``x86/pmu/my_branch_name``,
342and then write a small wrapper to extract ``pmu`` from the current branch name
343to yield ``--base=x/pmu``, where ``x`` is whatever name your repository uses to
344track the KVM x86 remote.
345
346Co-Posting Tests
347~~~~~~~~~~~~~~~~
348KVM selftests that are associated with KVM changes, e.g. regression tests for
349bug fixes, should be posted along with the KVM changes as a single series.  The
350standard kernel rules for bisection apply, i.e. KVM changes that result in test
351failures should be ordered after the selftests updates, and vice versa, new
352tests that fail due to KVM bugs should be ordered after the KVM fixes.
353
354KVM-unit-tests should *always* be posted separately.  Tools, e.g. b4 am, don't
355know that KVM-unit-tests is a separate repository and get confused when patches
356in a series apply on different trees.  To tie KVM-unit-tests patches back to
357KVM patches, first post the KVM changes and then provide a lore Link: to the
358KVM patch/series in the KVM-unit-tests patch(es).
359
360Notifications
361-------------
362When a patch/series is officially accepted, a notification email will be sent
363in reply to the original posting (cover letter for multi-patch series).  The
364notification will include the tree and topic branch, along with the SHA1s of
365the commits of applied patches.
366
367If a subset of patches is applied, this will be clearly stated in the
368notification.  Unless stated otherwise, it's implied that any patches in the
369series that were not accepted need more work and should be submitted in a new
370version.
371
372If for some reason a patch is dropped after officially being accepted, a reply
373will be sent to the notification email explaining why the patch was dropped, as
374well as the next steps.
375
376SHA1 Stability
377~~~~~~~~~~~~~~
378SHA1s are not 100% guaranteed to be stable until they land in Linus' tree!  A
379SHA1 is *usually* stable once a notification has been sent, but things happen.
380In most cases, an update to the notification email be provided if an applied
381patch's SHA1 changes.  However, in some scenarios, e.g. if all KVM x86 branches
382need to be rebased, individual notifications will not be given.
383
384Vulnerabilities
385---------------
386Bugs that can be exploited by the guest to attack the host (kernel or
387userspace), or that can be exploited by a nested VM to *its* host (L2 attacking
388L1), are of particular interest to KVM.  Please follow the protocol for
389:ref:`securitybugs` if you suspect a bug can lead to an escape, data leak, etc.
390
391