ARM64:

* Eager page splitting optimization for dirty logging, optionally
   allowing for a VM to avoid the cost of hugepage splitting in the stage-2
   fault path.
 
 * Arm FF-A proxy for pKVM, allowing a pKVM host to safely interact with
   services that live in the Secure world. pKVM intervenes on FF-A calls
   to guarantee the host doesn't misuse memory donated to the hyp or a
   pKVM guest.
 
 * Support for running the split hypervisor with VHE enabled, known as
   'hVHE' mode. This is extremely useful for testing the split
   hypervisor on VHE-only systems, and paves the way for new use cases
   that depend on having two TTBRs available at EL2.
 
 * Generalized framework for configurable ID registers from userspace.
   KVM/arm64 currently prevents arbitrary CPU feature set configuration
   from userspace, but the intent is to relax this limitation and allow
   userspace to select a feature set consistent with the CPU.
 
 * Enable the use of Branch Target Identification (FEAT_BTI) in the
   hypervisor.
 
 * Use a separate set of pointer authentication keys for the hypervisor
   when running in protected mode, as the host is untrusted at runtime.
 
 * Ensure timer IRQs are consistently released in the init failure
   paths.
 
 * Avoid trapping CTR_EL0 on systems with Enhanced Virtualization Traps
   (FEAT_EVT), as it is a register commonly read from userspace.
 
 * Erratum workaround for the upcoming AmpereOne part, which has broken
   hardware A/D state management.
 
 RISC-V:
 
 * Redirect AMO load/store misaligned traps to KVM guest
 
 * Trap-n-emulate AIA in-kernel irqchip for KVM guest
 
 * Svnapot support for KVM Guest
 
 s390:
 
 * New uvdevice secret API
 
 * CMM selftest and fixes
 
 * fix racy access to target CPU for diag 9c
 
 x86:
 
 * Fix missing/incorrect #GP checks on ENCLS
 
 * Use standard mmu_notifier hooks for handling APIC access page
 
 * Drop now unnecessary TR/TSS load after VM-Exit on AMD
 
 * Print more descriptive information about the status of SEV and SEV-ES during
   module load
 
 * Add a test for splitting and reconstituting hugepages during and after
   dirty logging
 
 * Add support for CPU pinning in demand paging test
 
 * Add support for AMD PerfMonV2, with a variety of cleanups and minor fixes
   included along the way
 
 * Add a "nx_huge_pages=never" option to effectively avoid creating NX hugepage
   recovery threads (because nx_huge_pages=off can be toggled at runtime)
 
 * Move handling of PAT out of MTRR code and dedup SVM+VMX code
 
 * Fix output of PIC poll command emulation when there's an interrupt
 
 * Add a maintainer's handbook to document KVM x86 processes, preferred coding
   style, testing expectations, etc.
 
 * Misc cleanups, fixes and comments
 
 Generic:
 
 * Miscellaneous bugfixes and cleanups
 
 Selftests:
 
 * Generate dependency files so that partial rebuilds work as expected
 -----BEGIN PGP SIGNATURE-----
 
 iQFIBAABCAAyFiEE8TM4V0tmI4mGbHaCv/vSX3jHroMFAmSgHrIUHHBib256aW5p
 QHJlZGhhdC5jb20ACgkQv/vSX3jHroORcAf+KkBlXwQMf+Q0Hy6Mfe0OtkKmh0Ae
 6HJ6dsuMfOHhWv5kgukh+qvuGUGzHq+gpVKmZg2yP3h3cLHOLUAYMCDm+rjXyjsk
 F4DbnJLfxq43Pe9PHRKFxxSecRcRYCNox0GD5UYL4PLKcH0FyfQrV+HVBK+GI8L3
 FDzUcyJkR12Lcj1qf++7fsbzfOshL0AJPmidQCoc6wkLJpUEr/nYUqlI1Kx3YNuQ
 LKmxFHS4l4/O/px3GKNDrLWDbrVlwciGIa3GZLS52PZdW3mAqT+cqcPcYK6SW71P
 m1vE80VbNELX5q3YSRoOXtedoZ3Pk97LEmz/xQAsJ/jri0Z5Syk0Ok0m/Q==
 =AMXp
 -----END PGP SIGNATURE-----

Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm

Pull kvm updates from Paolo Bonzini:
 "ARM64:

   - Eager page splitting optimization for dirty logging, optionally
     allowing for a VM to avoid the cost of hugepage splitting in the
     stage-2 fault path.

   - Arm FF-A proxy for pKVM, allowing a pKVM host to safely interact
     with services that live in the Secure world. pKVM intervenes on
     FF-A calls to guarantee the host doesn't misuse memory donated to
     the hyp or a pKVM guest.

   - Support for running the split hypervisor with VHE enabled, known as
     'hVHE' mode. This is extremely useful for testing the split
     hypervisor on VHE-only systems, and paves the way for new use cases
     that depend on having two TTBRs available at EL2.

   - Generalized framework for configurable ID registers from userspace.
     KVM/arm64 currently prevents arbitrary CPU feature set
     configuration from userspace, but the intent is to relax this
     limitation and allow userspace to select a feature set consistent
     with the CPU.

   - Enable the use of Branch Target Identification (FEAT_BTI) in the
     hypervisor.

   - Use a separate set of pointer authentication keys for the
     hypervisor when running in protected mode, as the host is untrusted
     at runtime.

   - Ensure timer IRQs are consistently released in the init failure
     paths.

   - Avoid trapping CTR_EL0 on systems with Enhanced Virtualization
     Traps (FEAT_EVT), as it is a register commonly read from userspace.

   - Erratum workaround for the upcoming AmpereOne part, which has
     broken hardware A/D state management.

  RISC-V:

   - Redirect AMO load/store misaligned traps to KVM guest

   - Trap-n-emulate AIA in-kernel irqchip for KVM guest

   - Svnapot support for KVM Guest

  s390:

   - New uvdevice secret API

   - CMM selftest and fixes

   - fix racy access to target CPU for diag 9c

  x86:

   - Fix missing/incorrect #GP checks on ENCLS

   - Use standard mmu_notifier hooks for handling APIC access page

   - Drop now unnecessary TR/TSS load after VM-Exit on AMD

   - Print more descriptive information about the status of SEV and
     SEV-ES during module load

   - Add a test for splitting and reconstituting hugepages during and
     after dirty logging

   - Add support for CPU pinning in demand paging test

   - Add support for AMD PerfMonV2, with a variety of cleanups and minor
     fixes included along the way

   - Add a "nx_huge_pages=never" option to effectively avoid creating NX
     hugepage recovery threads (because nx_huge_pages=off can be toggled
     at runtime)

   - Move handling of PAT out of MTRR code and dedup SVM+VMX code

   - Fix output of PIC poll command emulation when there's an interrupt

   - Add a maintainer's handbook to document KVM x86 processes,
     preferred coding style, testing expectations, etc.

   - Misc cleanups, fixes and comments

  Generic:

   - Miscellaneous bugfixes and cleanups

  Selftests:

   - Generate dependency files so that partial rebuilds work as
     expected"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (153 commits)
  Documentation/process: Add a maintainer handbook for KVM x86
  Documentation/process: Add a label for the tip tree handbook's coding style
  KVM: arm64: Fix misuse of KVM_ARM_VCPU_POWER_OFF bit index
  RISC-V: KVM: Remove unneeded semicolon
  RISC-V: KVM: Allow Svnapot extension for Guest/VM
  riscv: kvm: define vcpu_sbi_ext_pmu in header
  RISC-V: KVM: Expose IMSIC registers as attributes of AIA irqchip
  RISC-V: KVM: Add in-kernel virtualization of AIA IMSIC
  RISC-V: KVM: Expose APLIC registers as attributes of AIA irqchip
  RISC-V: KVM: Add in-kernel emulation of AIA APLIC
  RISC-V: KVM: Implement device interface for AIA irqchip
  RISC-V: KVM: Skeletal in-kernel AIA irqchip support
  RISC-V: KVM: Set kvm_riscv_aia_nr_hgei to zero
  RISC-V: KVM: Add APLIC related defines
  RISC-V: KVM: Add IMSIC related defines
  RISC-V: KVM: Implement guest external interrupt line management
  KVM: x86: Remove PRIx* definitions as they are solely for user space
  s390/uv: Update query for secret-UVCs
  s390/uv: replace scnprintf with sysfs_emit
  s390/uvdevice: Add 'Lock Secret Store' UVC
  ...

Documentation/arch/arm64/silicon-errata.rst

@@ -52,6 +52,9 @@ stable kernels.
 | Allwinner      | A64/R18         | UNKNOWN1        | SUN50I_ERRATUM_UNKNOWN1     |
 +----------------+-----------------+-----------------+-----------------------------+
 +----------------+-----------------+-----------------+-----------------------------+
+| Ampere         | AmpereOne       | AC03_CPU_38     | AMPERE_ERRATUM_AC03_CPU_38  |
++----------------+-----------------+-----------------+-----------------------------+
++----------------+-----------------+-----------------+-----------------------------+
 | ARM            | Cortex-A510     | #2457168        | ARM64_ERRATUM_2457168       |
 +----------------+-----------------+-----------------+-----------------------------+
 | ARM            | Cortex-A510     | #2064142        | ARM64_ERRATUM_2064142       |

Documentation/process/maintainer-handbooks.rst

@@ -18,3 +18,4 @@ Contents:
    maintainer-netdev
    maintainer-soc
    maintainer-tip
+   maintainer-kvm-x86

Documentation/process/maintainer-kvm-x86.rst

@@ -0,0 +1,390 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+KVM x86
+=======
+
+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.
+
+TL;DR
+-----
+Testing is mandatory.  Be consistent with established styles and patterns.
+
+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.
+
+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.
+
+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).
+
+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!
+
+Development
+-----------
+
+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.
+
+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
+:ref:`maintainer-tip-coding-style`, 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).
+
+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.
+
+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.
+
+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 :ref:`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.
+
+Changelog
+~~~~~~~~~
+Most importantly, write changelogs using imperative mood and avoid pronouns.
+
+See :ref:`describe_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".
+
+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).
+
+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.
+
+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.
+
+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.
+
+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.
+
+Posting
+-------
+
+Links
+~~~~~
+Do not explicitly reference bug reports, prior versions of a patch/series, etc.
+via ``In-Reply-To:`` headers.  Using ``In-Reply-To:`` becomes an unholy mess
+for large series and/or when the version count gets high, and ``In-Reply-To:``
+is useless for anyone that doesn't have the original message, e.g. if someone
+wasn't Cc'd on the bug report or if the list of recipients changes between
+versions.
+
+To link to a bug report, previous version, or anything of interest, use lore
+links.  For referencing previous version(s), generally speaking do not include
+a Link: in the changelog as there is no need to record the history in git, i.e.
+put the link in the cover letter or in the section git ignores.  Do provide a
+formal Link: for bug reports and/or discussions that led to the patch.  The
+context of why a change was made is highly valuable for future readers.
+
+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.
+
+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).
+
+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.
+
+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.
+
+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
+:ref:`securitybugs` if you suspect a bug can lead to an escape, data leak, etc.
+

Documentation/process/maintainer-tip.rst

@@ -455,6 +455,8 @@ and can be added to an existing kernel config by running:
 Some of these options are x86-specific and can be left out when testing
 on other architectures.
 
+.. _maintainer-tip-coding-style:
+
 Coding style notes
 ------------------
 

Documentation/virt/kvm/api.rst

@@ -8445,6 +8445,33 @@ structure.
 When getting the Modified Change Topology Report value, the attr->addr
 must point to a byte where the value will be stored or retrieved from.
 
+8.40 KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE
+---------------------------------------
+
+:Capability: KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE
+:Architectures: arm64
+:Type: vm
+:Parameters: arg[0] is the new split chunk size.
+:Returns: 0 on success, -EINVAL if any memslot was already created.
+
+This capability sets the chunk size used in Eager Page Splitting.
+
+Eager Page Splitting improves the performance of dirty-logging (used
+in live migrations) when guest memory is backed by huge-pages.  It
+avoids splitting huge-pages (into PAGE_SIZE pages) on fault, by doing
+it eagerly when enabling dirty logging (with the
+KVM_MEM_LOG_DIRTY_PAGES flag for a memory region), or when using
+KVM_CLEAR_DIRTY_LOG.
+
+The chunk size specifies how many pages to break at a time, using a
+single allocation for each chunk. Bigger the chunk size, more pages
+need to be allocated ahead of time.
+
+The chunk size needs to be a valid block size. The list of acceptable
+block sizes is exposed in KVM_CAP_ARM_SUPPORTED_BLOCK_SIZES as a
+64-bit bitmap (each bit describing a block size). The default value is
+0, to disable the eager page splitting.
+
 9. Known KVM API problems
 =========================
 

Documentation/virt/kvm/x86/mmu.rst

@@ -205,7 +205,7 @@ Shadow pages contain the following information:
   role.passthrough:
     The page is not backed by a guest page table, but its first entry
     points to one.  This is set if NPT uses 5-level page tables (host
-    CR4.LA57=1) and is shadowing L1's 4-level NPT (L1 CR4.LA57=1).
+    CR4.LA57=1) and is shadowing L1's 4-level NPT (L1 CR4.LA57=0).
   gfn:
     Either the guest page table containing the translations shadowed by this
     page, or the base page frame for linear translations.  See role.direct.

MAINTAINERS

@@ -11546,6 +11546,7 @@ M:	Sean Christopherson <seanjc@google.com>
 M:	Paolo Bonzini <pbonzini@redhat.com>
 L:	kvm@vger.kernel.org
 S:	Supported
+P:	Documentation/process/maintainer-kvm-x86.rst
 T:	git git://git.kernel.org/pub/scm/virt/kvm/kvm.git
 F:	arch/x86/include/asm/kvm*
 F:	arch/x86/include/asm/svm.h

arch/arm64/Kconfig

@@ -414,6 +414,25 @@ menu "Kernel Features"
 
 menu "ARM errata workarounds via the alternatives framework"
 
+config AMPERE_ERRATUM_AC03_CPU_38
+        bool "AmpereOne: AC03_CPU_38: Certain bits in the Virtualization Translation Control Register and Translation Control Registers do not follow RES0 semantics"
+	default y
+	help
+	  This option adds an alternative code sequence to work around Ampere
+	  erratum AC03_CPU_38 on AmpereOne.
+
+	  The affected design reports FEAT_HAFDBS as not implemented in
+	  ID_AA64MMFR1_EL1.HAFDBS, but (V)TCR_ELx.{HA,HD} are not RES0
+	  as required by the architecture. The unadvertised HAFDBS
+	  implementation suffers from an additional erratum where hardware
+	  A/D updates can occur after a PTE has been marked invalid.
+
+	  The workaround forces KVM to explicitly set VTCR_EL2.HA to 0,
+	  which avoids enabling unadvertised hardware Access Flag management
+	  at stage-2.
+
+	  If unsure, say Y.
+
 config ARM64_WORKAROUND_CLEAN_CACHE
 	bool
 

arch/arm64/include/asm/cpufeature.h

@@ -15,6 +15,9 @@
 #define MAX_CPU_FEATURES	128
 #define cpu_feature(x)		KERNEL_HWCAP_ ## x
 
+#define ARM64_SW_FEATURE_OVERRIDE_NOKASLR	0
+#define ARM64_SW_FEATURE_OVERRIDE_HVHE		4
+
 #ifndef __ASSEMBLY__
 
 #include <linux/bug.h>
@@ -905,6 +908,7 @@ static inline unsigned int get_vmid_bits(u64 mmfr1)
 	return 8;
 }
 
+s64 arm64_ftr_safe_value(const struct arm64_ftr_bits *ftrp, s64 new, s64 cur);
 struct arm64_ftr_reg *get_arm64_ftr_reg(u32 sys_id);
 
 extern struct arm64_ftr_override id_aa64mmfr1_override;
@@ -915,6 +919,8 @@ extern struct arm64_ftr_override id_aa64smfr0_override;
 extern struct arm64_ftr_override id_aa64isar1_override;
 extern struct arm64_ftr_override id_aa64isar2_override;
 
+extern struct arm64_ftr_override arm64_sw_feature_override;
+
 u32 get_kvm_ipa_limit(void);
 void dump_cpu_features(void);
 

arch/arm64/include/asm/el2_setup.h

@@ -43,6 +43,11 @@
  */
 .macro __init_el2_timers
 	mov	x0, #3				// Enable EL1 physical timers
+	mrs	x1, hcr_el2
+	and	x1, x1, #HCR_E2H
+	cbz	x1, .LnVHE_\@
+	lsl	x0, x0, #10
+.LnVHE_\@:
 	msr	cnthctl_el2, x0
 	msr	cntvoff_el2, xzr		// Clear virtual offset
 .endm
@@ -133,8 +138,15 @@
 .endm
 
 /* Coprocessor traps */
-.macro __init_el2_nvhe_cptr
+.macro __init_el2_cptr
+	mrs	x1, hcr_el2
+	and	x1, x1, #HCR_E2H
+	cbz	x1, .LnVHE_\@
+	mov	x0, #(CPACR_EL1_FPEN_EL1EN | CPACR_EL1_FPEN_EL0EN)
+	b	.Lset_cptr_\@
+.LnVHE_\@:
 	mov	x0, #0x33ff
+.Lset_cptr_\@:
 	msr	cptr_el2, x0			// Disable copro. traps to EL2
 .endm
 
@@ -210,9 +222,8 @@
 	__init_el2_gicv3
 	__init_el2_hstr
 	__init_el2_nvhe_idregs
-	__init_el2_nvhe_cptr
+	__init_el2_cptr
 	__init_el2_fgt
-	__init_el2_nvhe_prepare_eret
 .endm
 
 #ifndef __KVM_NVHE_HYPERVISOR__
@@ -258,7 +269,17 @@
 
 .Linit_sve_\@:	/* SVE register access */
 	mrs	x0, cptr_el2			// Disable SVE traps
+	mrs	x1, hcr_el2
+	and	x1, x1, #HCR_E2H
+	cbz	x1, .Lcptr_nvhe_\@
+
+	// VHE case
+	orr	x0, x0, #(CPACR_EL1_ZEN_EL1EN | CPACR_EL1_ZEN_EL0EN)
+	b	.Lset_cptr_\@
+
+.Lcptr_nvhe_\@: // nVHE case
 	bic	x0, x0, #CPTR_EL2_TZ
+.Lset_cptr_\@:
 	msr	cptr_el2, x0
 	isb
 	mov	x1, #ZCR_ELx_LEN_MASK		// SVE: Enable full vector

arch/arm64/include/asm/kvm_arm.h

@@ -19,6 +19,7 @@
 #define HCR_ATA_SHIFT	56
 #define HCR_ATA		(UL(1) << HCR_ATA_SHIFT)
 #define HCR_AMVOFFEN	(UL(1) << 51)
+#define HCR_TID4	(UL(1) << 49)
 #define HCR_FIEN	(UL(1) << 47)
 #define HCR_FWB		(UL(1) << 46)
 #define HCR_API		(UL(1) << 41)
@@ -87,7 +88,7 @@
 #define HCR_GUEST_FLAGS (HCR_TSC | HCR_TSW | HCR_TWE | HCR_TWI | HCR_VM | \
 			 HCR_BSU_IS | HCR_FB | HCR_TACR | \
 			 HCR_AMO | HCR_SWIO | HCR_TIDCP | HCR_RW | HCR_TLOR | \
-			 HCR_FMO | HCR_IMO | HCR_PTW | HCR_TID3 | HCR_TID2)
+			 HCR_FMO | HCR_IMO | HCR_PTW | HCR_TID3)
 #define HCR_VIRT_EXCP_MASK (HCR_VSE | HCR_VI | HCR_VF)
 #define HCR_HOST_NVHE_FLAGS (HCR_RW | HCR_API | HCR_APK | HCR_ATA)
 #define HCR_HOST_NVHE_PROTECTED_FLAGS (HCR_HOST_NVHE_FLAGS | HCR_TSC)
@@ -289,7 +290,6 @@
 #define CPTR_EL2_TFP	(1 << CPTR_EL2_TFP_SHIFT)
 #define CPTR_EL2_TZ	(1 << 8)
 #define CPTR_NVHE_EL2_RES1	0x000032ff /* known RES1 bits in CPTR_EL2 (nVHE) */
-#define CPTR_EL2_DEFAULT	CPTR_NVHE_EL2_RES1
 #define CPTR_NVHE_EL2_RES0	(GENMASK(63, 32) |	\
 				 GENMASK(29, 21) |	\
 				 GENMASK(19, 14) |	\
@@ -351,8 +351,7 @@
 	ECN(SOFTSTP_CUR), ECN(WATCHPT_LOW), ECN(WATCHPT_CUR), \
 	ECN(BKPT32), ECN(VECTOR32), ECN(BRK64), ECN(ERET)
 
-#define CPACR_EL1_DEFAULT	(CPACR_EL1_FPEN_EL0EN | CPACR_EL1_FPEN_EL1EN |\
+#define CPACR_EL1_TTA		(1 << 28)
-				 CPACR_EL1_ZEN_EL1EN)
 
 #define kvm_mode_names				\
 	{ PSR_MODE_EL0t,	"EL0t" },	\

arch/arm64/include/asm/kvm_asm.h

@@ -68,6 +68,7 @@ enum __kvm_host_smccc_func {
 	__KVM_HOST_SMCCC_FUNC___kvm_vcpu_run,
 	__KVM_HOST_SMCCC_FUNC___kvm_flush_vm_context,
 	__KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_ipa,
+	__KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_ipa_nsh,
 	__KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid,
 	__KVM_HOST_SMCCC_FUNC___kvm_flush_cpu_context,
 	__KVM_HOST_SMCCC_FUNC___kvm_timer_set_cntvoff,
@@ -225,6 +226,9 @@ extern void __kvm_flush_vm_context(void);
 extern void __kvm_flush_cpu_context(struct kvm_s2_mmu *mmu);
 extern void __kvm_tlb_flush_vmid_ipa(struct kvm_s2_mmu *mmu, phys_addr_t ipa,
 				     int level);
+extern void __kvm_tlb_flush_vmid_ipa_nsh(struct kvm_s2_mmu *mmu,
+					 phys_addr_t ipa,
+					 int level);
 extern void __kvm_tlb_flush_vmid(struct kvm_s2_mmu *mmu);
 
 extern void __kvm_timer_set_cntvoff(u64 cntvoff);

arch/arm64/include/asm/kvm_emulate.h

@@ -62,19 +62,14 @@ static __always_inline bool vcpu_el1_is_32bit(struct kvm_vcpu *vcpu)
 #else
 static __always_inline bool vcpu_el1_is_32bit(struct kvm_vcpu *vcpu)
 {
-	struct kvm *kvm = vcpu->kvm;
+	return test_bit(KVM_ARM_VCPU_EL1_32BIT, vcpu->arch.features);
-
-	WARN_ON_ONCE(!test_bit(KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED,
-			       &kvm->arch.flags));
-
-	return test_bit(KVM_ARCH_FLAG_EL1_32BIT, &kvm->arch.flags);
 }
 #endif
 
 static inline void vcpu_reset_hcr(struct kvm_vcpu *vcpu)
 {
 	vcpu->arch.hcr_el2 = HCR_GUEST_FLAGS;
-	if (is_kernel_in_hyp_mode())
+	if (has_vhe() || has_hvhe())
 		vcpu->arch.hcr_el2 |= HCR_E2H;
 	if (cpus_have_const_cap(ARM64_HAS_RAS_EXTN)) {
 		/* route synchronous external abort exceptions to EL2 */
@@ -95,6 +90,12 @@ static inline void vcpu_reset_hcr(struct kvm_vcpu *vcpu)
 		vcpu->arch.hcr_el2 |= HCR_TVM;
 	}
 
+	if (cpus_have_final_cap(ARM64_HAS_EVT) &&
+	    !cpus_have_final_cap(ARM64_MISMATCHED_CACHE_TYPE))
+		vcpu->arch.hcr_el2 |= HCR_TID4;
+	else
+		vcpu->arch.hcr_el2 |= HCR_TID2;
+
 	if (vcpu_el1_is_32bit(vcpu))
 		vcpu->arch.hcr_el2 &= ~HCR_RW;
 
@@ -570,4 +571,35 @@ static inline bool vcpu_has_feature(struct kvm_vcpu *vcpu, int feature)
 	return test_bit(feature, vcpu->arch.features);
 }
 
+static __always_inline u64 kvm_get_reset_cptr_el2(struct kvm_vcpu *vcpu)
+{
+	u64 val;
+
+	if (has_vhe()) {
+		val = (CPACR_EL1_FPEN_EL0EN | CPACR_EL1_FPEN_EL1EN |
+		       CPACR_EL1_ZEN_EL1EN);
+	} else if (has_hvhe()) {
+		val = (CPACR_EL1_FPEN_EL0EN | CPACR_EL1_FPEN_EL1EN);
+	} else {
+		val = CPTR_NVHE_EL2_RES1;
+
+		if (vcpu_has_sve(vcpu) &&
+		    (vcpu->arch.fp_state == FP_STATE_GUEST_OWNED))
+			val |= CPTR_EL2_TZ;
+		if (cpus_have_final_cap(ARM64_SME))
+			val &= ~CPTR_EL2_TSM;
+	}
+
+	return val;
+}
+
+static __always_inline void kvm_reset_cptr_el2(struct kvm_vcpu *vcpu)
+{
+	u64 val = kvm_get_reset_cptr_el2(vcpu);
+
+	if (has_vhe() || has_hvhe())
+		write_sysreg(val, cpacr_el1);
+	else
+		write_sysreg(val, cptr_el2);
+}
 #endif /* __ARM64_KVM_EMULATE_H__ */

arch/arm64/include/asm/kvm_host.h

@@ -39,6 +39,7 @@
 #define KVM_MAX_VCPUS VGIC_V3_MAX_CPUS
 
 #define KVM_VCPU_MAX_FEATURES 7
+#define KVM_VCPU_VALID_FEATURES	(BIT(KVM_VCPU_MAX_FEATURES) - 1)
 
 #define KVM_REQ_SLEEP \
 	KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
@@ -159,6 +160,21 @@ struct kvm_s2_mmu {
 	/* The last vcpu id that ran on each physical CPU */
 	int __percpu *last_vcpu_ran;
 
+#define KVM_ARM_EAGER_SPLIT_CHUNK_SIZE_DEFAULT 0
+	/*
+	 * Memory cache used to split
+	 * KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE worth of huge pages. It
+	 * is used to allocate stage2 page tables while splitting huge
+	 * pages. The choice of KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE
+	 * influences both the capacity of the split page cache, and
+	 * how often KVM reschedules. Be wary of raising CHUNK_SIZE
+	 * too high.
+	 *
+	 * Protected by kvm->slots_lock.
+	 */
+	struct kvm_mmu_memory_cache split_page_cache;
+	uint64_t split_page_chunk_size;
+
 	struct kvm_arch *arch;
 };
 
@@ -214,25 +230,23 @@ struct kvm_arch {
 #define KVM_ARCH_FLAG_MTE_ENABLED			1
 	/* At least one vCPU has ran in the VM */
 #define KVM_ARCH_FLAG_HAS_RAN_ONCE			2
-	/*
+	/* The vCPU feature set for the VM is configured */
-	 * The following two bits are used to indicate the guest's EL1
+#define KVM_ARCH_FLAG_VCPU_FEATURES_CONFIGURED		3
-	 * register width configuration. A value of KVM_ARCH_FLAG_EL1_32BIT
-	 * bit is valid only when KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED is set.
-	 * Otherwise, the guest's EL1 register width has not yet been
-	 * determined yet.
-	 */
-#define KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED		3
-#define KVM_ARCH_FLAG_EL1_32BIT				4
 	/* PSCI SYSTEM_SUSPEND enabled for the guest */
-#define KVM_ARCH_FLAG_SYSTEM_SUSPEND_ENABLED		5
+#define KVM_ARCH_FLAG_SYSTEM_SUSPEND_ENABLED		4
 	/* VM counter offset */
-#define KVM_ARCH_FLAG_VM_COUNTER_OFFSET			6
+#define KVM_ARCH_FLAG_VM_COUNTER_OFFSET			5
 	/* Timer PPIs made immutable */
-#define KVM_ARCH_FLAG_TIMER_PPIS_IMMUTABLE		7
+#define KVM_ARCH_FLAG_TIMER_PPIS_IMMUTABLE		6
 	/* SMCCC filter initialized for the VM */
-#define KVM_ARCH_FLAG_SMCCC_FILTER_CONFIGURED		8
+#define KVM_ARCH_FLAG_SMCCC_FILTER_CONFIGURED		7
+	/* Initial ID reg values loaded */
+#define KVM_ARCH_FLAG_ID_REGS_INITIALIZED		8
 	unsigned long flags;
 
+	/* VM-wide vCPU feature set */
+	DECLARE_BITMAP(vcpu_features, KVM_VCPU_MAX_FEATURES);
+
 	/*
 	 * VM-wide PMU filter, implemented as a bitmap and big enough for
 	 * up to 2^10 events (ARMv8.0) or 2^16 events (ARMv8.1+).
@@ -242,17 +256,23 @@ struct kvm_arch {
 
 	cpumask_var_t supported_cpus;
 
-	u8 pfr0_csv2;
-	u8 pfr0_csv3;
-	struct {
-		u8 imp:4;
-		u8 unimp:4;
-	} dfr0_pmuver;
-
 	/* Hypercall features firmware registers' descriptor */
 	struct kvm_smccc_features smccc_feat;
 	struct maple_tree smccc_filter;
 
+	/*
+	 * Emulated CPU ID registers per VM
+	 * (Op0, Op1, CRn, CRm, Op2) of the ID registers to be saved in it
+	 * is (3, 0, 0, crm, op2), where 1<=crm<8, 0<=op2<8.
+	 *
+	 * These emulated idregs are VM-wide, but accessed from the context of a vCPU.
+	 * Atomic access to multiple idregs are guarded by kvm_arch.config_lock.
+	 */
+#define IDREG_IDX(id)		(((sys_reg_CRm(id) - 1) << 3) | sys_reg_Op2(id))
+#define IDREG(kvm, id)		((kvm)->arch.id_regs[IDREG_IDX(id)])
+#define KVM_ARM_ID_REG_NUM	(IDREG_IDX(sys_reg(3, 0, 0, 7, 7)) + 1)
+	u64 id_regs[KVM_ARM_ID_REG_NUM];
+
 	/*
 	 * For an untrusted host VM, 'pkvm.handle' is used to lookup
 	 * the associated pKVM instance in the hypervisor.
@@ -410,6 +430,7 @@ struct kvm_host_data {
 struct kvm_host_psci_config {
 	/* PSCI version used by host. */
 	u32 version;
+	u32 smccc_version;
 
 	/* Function IDs used by host if version is v0.1. */
 	struct psci_0_1_function_ids function_ids_0_1;

arch/arm64/include/asm/kvm_hyp.h

@@ -16,12 +16,35 @@ DECLARE_PER_CPU(struct kvm_cpu_context, kvm_hyp_ctxt);
 DECLARE_PER_CPU(unsigned long, kvm_hyp_vector);
 DECLARE_PER_CPU(struct kvm_nvhe_init_params, kvm_init_params);
 
+/*
+ * Unified accessors for registers that have a different encoding
+ * between VHE and non-VHE. They must be specified without their "ELx"
+ * encoding, but with the SYS_ prefix, as defined in asm/sysreg.h.
+ */
+
+#if defined(__KVM_VHE_HYPERVISOR__)
+
+#define read_sysreg_el0(r)	read_sysreg_s(r##_EL02)
+#define write_sysreg_el0(v,r)	write_sysreg_s(v, r##_EL02)
+#define read_sysreg_el1(r)	read_sysreg_s(r##_EL12)
+#define write_sysreg_el1(v,r)	write_sysreg_s(v, r##_EL12)
+#define read_sysreg_el2(r)	read_sysreg_s(r##_EL1)
+#define write_sysreg_el2(v,r)	write_sysreg_s(v, r##_EL1)
+
+#else // !__KVM_VHE_HYPERVISOR__
+
+#if defined(__KVM_NVHE_HYPERVISOR__)
+#define VHE_ALT_KEY	ARM64_KVM_HVHE
+#else
+#define VHE_ALT_KEY	ARM64_HAS_VIRT_HOST_EXTN
+#endif
+
 #define read_sysreg_elx(r,nvh,vh)					\
 	({								\
 		u64 reg;						\
-		asm volatile(ALTERNATIVE(__mrs_s("%0", r##nvh),	\
+		asm volatile(ALTERNATIVE(__mrs_s("%0", r##nvh),		\
 					 __mrs_s("%0", r##vh),		\
-					 ARM64_HAS_VIRT_HOST_EXTN)	\
+					 VHE_ALT_KEY)			\
 			     : "=r" (reg));				\
 		reg;							\
 	})
@@ -31,16 +54,10 @@ DECLARE_PER_CPU(struct kvm_nvhe_init_params, kvm_init_params);
 		u64 __val = (u64)(v);					\
 		asm volatile(ALTERNATIVE(__msr_s(r##nvh, "%x0"),	\
 					 __msr_s(r##vh, "%x0"),		\
-					 ARM64_HAS_VIRT_HOST_EXTN)	\
+					 VHE_ALT_KEY)			\
 					 : : "rZ" (__val));		\
 	} while (0)
 
-/*
- * Unified accessors for registers that have a different encoding
- * between VHE and non-VHE. They must be specified without their "ELx"
- * encoding, but with the SYS_ prefix, as defined in asm/sysreg.h.
- */
-
 #define read_sysreg_el0(r)	read_sysreg_elx(r, _EL0, _EL02)
 #define write_sysreg_el0(v,r)	write_sysreg_elx(v, r, _EL0, _EL02)
 #define read_sysreg_el1(r)	read_sysreg_elx(r, _EL1, _EL12)
@@ -48,6 +65,8 @@ DECLARE_PER_CPU(struct kvm_nvhe_init_params, kvm_init_params);
 #define read_sysreg_el2(r)	read_sysreg_elx(r, _EL2, _EL1)
 #define write_sysreg_el2(v,r)	write_sysreg_elx(v, r, _EL2, _EL1)
 
+#endif	// __KVM_VHE_HYPERVISOR__
+
 /*
  * Without an __arch_swab32(), we fall back to ___constant_swab32(), but the
  * static inline can allow the compiler to out-of-line this. KVM always wants

arch/arm64/include/asm/kvm_mmu.h

@@ -172,6 +172,7 @@ void __init free_hyp_pgds(void);
 
 void stage2_unmap_vm(struct kvm *kvm);
 int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long type);
+void kvm_uninit_stage2_mmu(struct kvm *kvm);
 void kvm_free_stage2_pgd(struct kvm_s2_mmu *mmu);
 int kvm_phys_addr_ioremap(struct kvm *kvm, phys_addr_t guest_ipa,
 			  phys_addr_t pa, unsigned long size, bool writable);
@@ -227,7 +228,8 @@ static inline void __invalidate_icache_guest_page(void *va, size_t size)
 	if (icache_is_aliasing()) {
 		/* any kind of VIPT cache */
 		icache_inval_all_pou();
-	} else if (is_kernel_in_hyp_mode() || !icache_is_vpipt()) {
+	} else if (read_sysreg(CurrentEL) != CurrentEL_EL1 ||
+		   !icache_is_vpipt()) {
 		/* PIPT or VPIPT at EL2 (see comment in __kvm_tlb_flush_vmid_ipa) */
 		icache_inval_pou((unsigned long)va, (unsigned long)va + size);
 	}

arch/arm64/include/asm/kvm_pgtable.h

@@ -92,6 +92,24 @@ static inline bool kvm_level_supports_block_mapping(u32 level)
 	return level >= KVM_PGTABLE_MIN_BLOCK_LEVEL;
 }
 
+static inline u32 kvm_supported_block_sizes(void)
+{
+	u32 level = KVM_PGTABLE_MIN_BLOCK_LEVEL;
+	u32 r = 0;
+
+	for (; level < KVM_PGTABLE_MAX_LEVELS; level++)
+		r |= BIT(kvm_granule_shift(level));
+
+	return r;
+}
+
+static inline bool kvm_is_block_size_supported(u64 size)
+{
+	bool is_power_of_two = IS_ALIGNED(size, size);
+
+	return is_power_of_two && (size & kvm_supported_block_sizes());
+}
+
 /**
  * struct kvm_pgtable_mm_ops - Memory management callbacks.
  * @zalloc_page:		Allocate a single zeroed memory page.
@@ -104,7 +122,7 @@ static inline bool kvm_level_supports_block_mapping(u32 level)
  *				allocation is physically contiguous.
  * @free_pages_exact:		Free an exact number of memory pages previously
  *				allocated by zalloc_pages_exact.
- * @free_removed_table:		Free a removed paging structure by unlinking and
+ * @free_unlinked_table:	Free an unlinked paging structure by unlinking and
  *				dropping references.
  * @get_page:			Increment the refcount on a page.
  * @put_page:			Decrement the refcount on a page. When the
@@ -124,7 +142,7 @@ struct kvm_pgtable_mm_ops {
 	void*		(*zalloc_page)(void *arg);
 	void*		(*zalloc_pages_exact)(size_t size);
 	void		(*free_pages_exact)(void *addr, size_t size);
-	void		(*free_removed_table)(void *addr, u32 level);
+	void		(*free_unlinked_table)(void *addr, u32 level);
 	void		(*get_page)(void *addr);
 	void		(*put_page)(void *addr);
 	int		(*page_count)(void *addr);
@@ -195,6 +213,12 @@ typedef bool (*kvm_pgtable_force_pte_cb_t)(u64 addr, u64 end,
  *					with other software walkers.
  * @KVM_PGTABLE_WALK_HANDLE_FAULT:	Indicates the page-table walk was
  *					invoked from a fault handler.
+ * @KVM_PGTABLE_WALK_SKIP_BBM_TLBI:	Visit and update table entries
+ *					without Break-before-make's
+ *					TLB invalidation.
+ * @KVM_PGTABLE_WALK_SKIP_CMO:		Visit and update table entries
+ *					without Cache maintenance
+ *					operations required.
  */
 enum kvm_pgtable_walk_flags {
 	KVM_PGTABLE_WALK_LEAF			= BIT(0),
@@ -202,6 +226,8 @@ enum kvm_pgtable_walk_flags {
 	KVM_PGTABLE_WALK_TABLE_POST		= BIT(2),
 	KVM_PGTABLE_WALK_SHARED			= BIT(3),
 	KVM_PGTABLE_WALK_HANDLE_FAULT		= BIT(4),
+	KVM_PGTABLE_WALK_SKIP_BBM_TLBI		= BIT(5),
+	KVM_PGTABLE_WALK_SKIP_CMO		= BIT(6),
 };
 
 struct kvm_pgtable_visit_ctx {
@@ -441,7 +467,7 @@ int __kvm_pgtable_stage2_init(struct kvm_pgtable *pgt, struct kvm_s2_mmu *mmu,
 void kvm_pgtable_stage2_destroy(struct kvm_pgtable *pgt);
 
 /**
- * kvm_pgtable_stage2_free_removed() - Free a removed stage-2 paging structure.
+ * kvm_pgtable_stage2_free_unlinked() - Free an unlinked stage-2 paging structure.
  * @mm_ops:	Memory management callbacks.
  * @pgtable:	Unlinked stage-2 paging structure to be freed.
  * @level:	Level of the stage-2 paging structure to be freed.
@@ -449,7 +475,33 @@ void kvm_pgtable_stage2_destroy(struct kvm_pgtable *pgt);
  * The page-table is assumed to be unreachable by any hardware walkers prior to
  * freeing and therefore no TLB invalidation is performed.
  */
-void kvm_pgtable_stage2_free_removed(struct kvm_pgtable_mm_ops *mm_ops, void *pgtable, u32 level);
+void kvm_pgtable_stage2_free_unlinked(struct kvm_pgtable_mm_ops *mm_ops, void *pgtable, u32 level);
+
+/**
+ * kvm_pgtable_stage2_create_unlinked() - Create an unlinked stage-2 paging structure.
+ * @pgt:	Page-table structure initialised by kvm_pgtable_stage2_init*().
+ * @phys:	Physical address of the memory to map.
+ * @level:	Starting level of the stage-2 paging structure to be created.
+ * @prot:	Permissions and attributes for the mapping.
+ * @mc:		Cache of pre-allocated and zeroed memory from which to allocate
+ *		page-table pages.
+ * @force_pte:  Force mappings to PAGE_SIZE granularity.
+ *
+ * Returns an unlinked page-table tree.  This new page-table tree is
+ * not reachable (i.e., it is unlinked) from the root pgd and it's
+ * therefore unreachableby the hardware page-table walker. No TLB
+ * invalidation or CMOs are performed.
+ *
+ * If device attributes are not explicitly requested in @prot, then the
+ * mapping will be normal, cacheable.
+ *
+ * Return: The fully populated (unlinked) stage-2 paging structure, or
+ * an ERR_PTR(error) on failure.
+ */
+kvm_pte_t *kvm_pgtable_stage2_create_unlinked(struct kvm_pgtable *pgt,
+					      u64 phys, u32 level,
+					      enum kvm_pgtable_prot prot,
+					      void *mc, bool force_pte);
 
 /**
  * kvm_pgtable_stage2_map() - Install a mapping in a guest stage-2 page-table.
@@ -620,6 +672,25 @@ bool kvm_pgtable_stage2_is_young(struct kvm_pgtable *pgt, u64 addr);
  */
 int kvm_pgtable_stage2_flush(struct kvm_pgtable *pgt, u64 addr, u64 size);
 
+/**
+ * kvm_pgtable_stage2_split() - Split a range of huge pages into leaf PTEs pointing
+ *				to PAGE_SIZE guest pages.
+ * @pgt:	 Page-table structure initialised by kvm_pgtable_stage2_init().
+ * @addr:	 Intermediate physical address from which to split.
+ * @size:	 Size of the range.
+ * @mc:		 Cache of pre-allocated and zeroed memory from which to allocate
+ *		 page-table pages.
+ *
+ * The function tries to split any level 1 or 2 entry that overlaps
+ * with the input range (given by @addr and @size).
+ *
+ * Return: 0 on success, negative error code on failure. Note that
+ * kvm_pgtable_stage2_split() is best effort: it tries to break as many
+ * blocks in the input range as allowed by @mc_capacity.
+ */
+int kvm_pgtable_stage2_split(struct kvm_pgtable *pgt, u64 addr, u64 size,
+			     struct kvm_mmu_memory_cache *mc);
+
 /**
  * kvm_pgtable_walk() - Walk a page-table.
  * @pgt:	Page-table structure initialised by kvm_pgtable_*_init().

arch/arm64/include/asm/kvm_pkvm.h

@@ -6,7 +6,9 @@
 #ifndef __ARM64_KVM_PKVM_H__
 #define __ARM64_KVM_PKVM_H__
 
+#include <linux/arm_ffa.h>
 #include <linux/memblock.h>
+#include <linux/scatterlist.h>
 #include <asm/kvm_pgtable.h>
 
 /* Maximum number of VMs that can co-exist under pKVM. */
@@ -106,4 +108,23 @@ static inline unsigned long host_s2_pgtable_pages(void)
 	return res;
 }
 
+#define KVM_FFA_MBOX_NR_PAGES	1
+
+static inline unsigned long hyp_ffa_proxy_pages(void)
+{
+	size_t desc_max;
+
+	/*
+	 * The hypervisor FFA proxy needs enough memory to buffer a fragmented
+	 * descriptor returned from EL3 in response to a RETRIEVE_REQ call.
+	 */
+	desc_max = sizeof(struct ffa_mem_region) +
+		   sizeof(struct ffa_mem_region_attributes) +
+		   sizeof(struct ffa_composite_mem_region) +
+		   SG_MAX_SEGMENTS * sizeof(struct ffa_mem_region_addr_range);
+
+	/* Plus a page each for the hypervisor's RX and TX mailboxes. */
+	return (2 * KVM_FFA_MBOX_NR_PAGES) + DIV_ROUND_UP(desc_max, PAGE_SIZE);
+}
+
 #endif	/* __ARM64_KVM_PKVM_H__ */

arch/arm64/include/asm/sysreg.h

@@ -510,6 +510,7 @@
 			 (BIT(18)) | (BIT(22)) | (BIT(23)) | (BIT(28)) | \
 			 (BIT(29)))
 
+#define SCTLR_EL2_BT	(BIT(36))
 #ifdef CONFIG_CPU_BIG_ENDIAN
 #define ENDIAN_SET_EL2		SCTLR_ELx_EE
 #else

arch/arm64/include/asm/virt.h

@@ -110,8 +110,10 @@ static inline bool is_hyp_mode_mismatched(void)
 	return __boot_cpu_mode[0] != __boot_cpu_mode[1];
 }
 
-static inline bool is_kernel_in_hyp_mode(void)
+static __always_inline bool is_kernel_in_hyp_mode(void)
 {
+	BUILD_BUG_ON(__is_defined(__KVM_NVHE_HYPERVISOR__) ||
+		     __is_defined(__KVM_VHE_HYPERVISOR__));
 	return read_sysreg(CurrentEL) == CurrentEL_EL2;
 }
 
@@ -140,6 +142,14 @@ static __always_inline bool is_protected_kvm_enabled(void)
 		return cpus_have_final_cap(ARM64_KVM_PROTECTED_MODE);
 }
 
+static __always_inline bool has_hvhe(void)
+{
+	if (is_vhe_hyp_code())
+		return false;
+
+	return cpus_have_final_cap(ARM64_KVM_HVHE);
+}
+
 static inline bool is_hyp_nvhe(void)
 {
 	return is_hyp_mode_available() && !is_kernel_in_hyp_mode();

arch/arm64/kernel/cpu_errata.c

@@ -729,6 +729,13 @@ const struct arm64_cpu_capabilities arm64_errata[] = {
 		MIDR_FIXED(MIDR_CPU_VAR_REV(1,1), BIT(25)),
 		.cpu_enable = cpu_clear_bf16_from_user_emulation,
 	},
+#endif
+#ifdef CONFIG_AMPERE_ERRATUM_AC03_CPU_38
+	{
+		.desc = "AmpereOne erratum AC03_CPU_38",
+		.capability = ARM64_WORKAROUND_AMPERE_AC03_CPU_38,
+		ERRATA_MIDR_ALL_VERSIONS(MIDR_AMPERE1),
+	},
 #endif
 	{
 	}

arch/arm64/kernel/cpufeature.c

@@ -672,6 +672,8 @@ struct arm64_ftr_override __ro_after_init id_aa64smfr0_override;
 struct arm64_ftr_override __ro_after_init id_aa64isar1_override;
 struct arm64_ftr_override __ro_after_init id_aa64isar2_override;
 
+struct arm64_ftr_override arm64_sw_feature_override;
+
 static const struct __ftr_reg_entry {
 	u32			sys_id;
 	struct arm64_ftr_reg 	*reg;
@@ -807,7 +809,7 @@ static u64 arm64_ftr_set_value(const struct arm64_ftr_bits *ftrp, s64 reg,
 	return reg;
 }
 
-static s64 arm64_ftr_safe_value(const struct arm64_ftr_bits *ftrp, s64 new,
+s64 arm64_ftr_safe_value(const struct arm64_ftr_bits *ftrp, s64 new,
 				s64 cur)
 {
 	s64 ret = 0;
@@ -2009,6 +2011,19 @@ static bool has_nested_virt_support(const struct arm64_cpu_capabilities *cap,
 	return true;
 }
 
+static bool hvhe_possible(const struct arm64_cpu_capabilities *entry,
+			  int __unused)
+{
+	u64 val;
+
+	val = read_sysreg(id_aa64mmfr1_el1);
+	if (!cpuid_feature_extract_unsigned_field(val, ID_AA64MMFR1_EL1_VH_SHIFT))
+		return false;
+
+	val = arm64_sw_feature_override.val & arm64_sw_feature_override.mask;
+	return cpuid_feature_extract_unsigned_field(val, ARM64_SW_FEATURE_OVERRIDE_HVHE);
+}
+
 #ifdef CONFIG_ARM64_PAN
 static void cpu_enable_pan(const struct arm64_cpu_capabilities *__unused)
 {
@@ -2683,6 +2698,23 @@ static const struct arm64_cpu_capabilities arm64_features[] = {
 		.matches = has_cpuid_feature,
 		ARM64_CPUID_FIELDS(ID_AA64MMFR3_EL1, S1PIE, IMP)
 	},
+	{
+		.desc = "VHE for hypervisor only",
+		.capability = ARM64_KVM_HVHE,
+		.type = ARM64_CPUCAP_SYSTEM_FEATURE,
+		.matches = hvhe_possible,
+	},
+	{
+		.desc = "Enhanced Virtualization Traps",
+		.capability = ARM64_HAS_EVT,
+		.type = ARM64_CPUCAP_SYSTEM_FEATURE,
+		.sys_reg = SYS_ID_AA64MMFR2_EL1,
+		.sign = FTR_UNSIGNED,
+		.field_pos = ID_AA64MMFR2_EL1_EVT_SHIFT,
+		.field_width = 4,
+		.min_field_value = ID_AA64MMFR2_EL1_EVT_IMP,
+		.matches = has_cpuid_feature,
+	},
 	{},
 };
 

arch/arm64/kernel/head.S

@@ -603,6 +603,8 @@ SYM_INNER_LABEL(init_el2, SYM_L_LOCAL)
 	msr	sctlr_el1, x1
 	mov	x2, xzr
 2:
+	__init_el2_nvhe_prepare_eret
+
 	mov	w0, #BOOT_CPU_MODE_EL2
 	orr	x0, x0, x2
 	eret

arch/arm64/kernel/hyp-stub.S

@@ -82,7 +82,15 @@ SYM_CODE_START_LOCAL(__finalise_el2)
 	tbnz	x1, #0, 1f
 
 	// Needs to be VHE capable, obviously
-	check_override id_aa64mmfr1 ID_AA64MMFR1_EL1_VH_SHIFT 2f 1f x1 x2
+	check_override id_aa64mmfr1 ID_AA64MMFR1_EL1_VH_SHIFT 0f 1f x1 x2
+
+0:	// Check whether we only want the hypervisor to run VHE, not the kernel
+	adr_l	x1, arm64_sw_feature_override
+	ldr	x2, [x1, FTR_OVR_VAL_OFFSET]
+	ldr	x1, [x1, FTR_OVR_MASK_OFFSET]
+	and	x2, x2, x1
+	ubfx	x2, x2, #ARM64_SW_FEATURE_OVERRIDE_HVHE, #4
+	cbz	x2, 2f
 
 1:	mov_q	x0, HVC_STUB_ERR
 	eret

arch/arm64/kernel/idreg-override.c

@@ -139,15 +139,22 @@ static const struct ftr_set_desc smfr0 __initconst = {
 	},
 };
 
-extern struct arm64_ftr_override kaslr_feature_override;
+static bool __init hvhe_filter(u64 val)
+{
+	u64 mmfr1 = read_sysreg(id_aa64mmfr1_el1);
 
-static const struct ftr_set_desc kaslr __initconst = {
+	return (val == 1 &&
-	.name		= "kaslr",
+		lower_32_bits(__boot_status) == BOOT_CPU_MODE_EL2 &&
-#ifdef CONFIG_RANDOMIZE_BASE
+		cpuid_feature_extract_unsigned_field(mmfr1,
-	.override	= &kaslr_feature_override,
+						     ID_AA64MMFR1_EL1_VH_SHIFT));
-#endif
+}
+
+static const struct ftr_set_desc sw_features __initconst = {
+	.name		= "arm64_sw",
+	.override	= &arm64_sw_feature_override,
 	.fields		= {
-		FIELD("disabled", 0, NULL),
+		FIELD("nokaslr", ARM64_SW_FEATURE_OVERRIDE_NOKASLR, NULL),
+		FIELD("hvhe", ARM64_SW_FEATURE_OVERRIDE_HVHE, hvhe_filter),
 		{}
 	},
 };
@@ -159,7 +166,7 @@ static const struct ftr_set_desc * const regs[] __initconst = {
 	&isar1,
 	&isar2,
 	&smfr0,
-	&kaslr,
+	&sw_features,
 };
 
 static const struct {
@@ -177,7 +184,7 @@ static const struct {
 	  "id_aa64isar2.gpa3=0 id_aa64isar2.apa3=0"	   },
 	{ "arm64.nomops",		"id_aa64isar2.mops=0" },
 	{ "arm64.nomte",		"id_aa64pfr1.mte=0" },
-	{ "nokaslr",			"kaslr.disabled=1" },
+	{ "nokaslr",			"arm64_sw.nokaslr=1" },
 };
 
 static int __init parse_nokaslr(char *unused)

arch/arm64/kernel/kaslr.c

@@ -12,13 +12,13 @@
 
 u16 __initdata memstart_offset_seed;
 
-struct arm64_ftr_override kaslr_feature_override __initdata;
-
 bool __ro_after_init __kaslr_is_enabled = false;
 
 void __init kaslr_init(void)
 {
-	if (kaslr_feature_override.val & kaslr_feature_override.mask & 0xf) {
+	if (cpuid_feature_extract_unsigned_field(arm64_sw_feature_override.val &
+						 arm64_sw_feature_override.mask,
+						 ARM64_SW_FEATURE_OVERRIDE_NOKASLR)) {
 		pr_info("KASLR disabled on command line\n");
 		return;
 	}

arch/arm64/kvm/arch_timer.c

@@ -1406,7 +1406,7 @@ int __init kvm_timer_hyp_init(bool has_gic)
 					    kvm_get_running_vcpus());
 		if (err) {
 			kvm_err("kvm_arch_timer: error setting vcpu affinity\n");
-			goto out_free_irq;
+			goto out_free_vtimer_irq;
 		}
 
 		static_branch_enable(&has_gic_active_state);
@@ -1422,7 +1422,7 @@ int __init kvm_timer_hyp_init(bool has_gic)
 		if (err) {
 			kvm_err("kvm_arch_timer: can't request ptimer interrupt %d (%d)\n",
 				host_ptimer_irq, err);
-			return err;
+			goto out_free_vtimer_irq;
 		}
 
 		if (has_gic) {
@@ -1430,7 +1430,7 @@ int __init kvm_timer_hyp_init(bool has_gic)
 						    kvm_get_running_vcpus());
 			if (err) {
 				kvm_err("kvm_arch_timer: error setting vcpu affinity\n");
-				goto out_free_irq;
+				goto out_free_ptimer_irq;
 			}
 		}
 
@@ -1439,11 +1439,15 @@ int __init kvm_timer_hyp_init(bool has_gic)
 		kvm_err("kvm_arch_timer: invalid physical timer IRQ: %d\n",
 			info->physical_irq);
 		err = -ENODEV;
-		goto out_free_irq;
+		goto out_free_vtimer_irq;
 	}
 
 	return 0;
-out_free_irq:
+
+out_free_ptimer_irq:
+	if (info->physical_irq > 0)
+		free_percpu_irq(host_ptimer_irq, kvm_get_running_vcpus());
+out_free_vtimer_irq:
 	free_percpu_irq(host_vtimer_irq, kvm_get_running_vcpus());
 	return err;
 }

arch/arm64/kvm/arm.c

@@ -51,6 +51,8 @@ DECLARE_KVM_HYP_PER_CPU(unsigned long, kvm_hyp_vector);
 DEFINE_PER_CPU(unsigned long, kvm_arm_hyp_stack_page);
 DECLARE_KVM_NVHE_PER_CPU(struct kvm_nvhe_init_params, kvm_init_params);
 
+DECLARE_KVM_NVHE_PER_CPU(struct kvm_cpu_context, kvm_hyp_ctxt);
+
 static bool vgic_present;
 
 static DEFINE_PER_CPU(unsigned char, kvm_arm_hardware_enabled);
@@ -65,6 +67,7 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
 			    struct kvm_enable_cap *cap)
 {
 	int r;
+	u64 new_cap;
 
 	if (cap->flags)
 		return -EINVAL;
@@ -89,6 +92,24 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
 		r = 0;
 		set_bit(KVM_ARCH_FLAG_SYSTEM_SUSPEND_ENABLED, &kvm->arch.flags);
 		break;
+	case KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE:
+		new_cap = cap->args[0];
+
+		mutex_lock(&kvm->slots_lock);
+		/*
+		 * To keep things simple, allow changing the chunk
+		 * size only when no memory slots have been created.
+		 */
+		if (!kvm_are_all_memslots_empty(kvm)) {
+			r = -EINVAL;
+		} else if (new_cap && !kvm_is_block_size_supported(new_cap)) {
+			r = -EINVAL;
+		} else {
+			r = 0;
+			kvm->arch.mmu.split_page_chunk_size = new_cap;
+		}
+		mutex_unlock(&kvm->slots_lock);
+		break;
 	default:
 		r = -EINVAL;
 		break;
@@ -102,22 +123,6 @@ static int kvm_arm_default_max_vcpus(void)
 	return vgic_present ? kvm_vgic_get_max_vcpus() : KVM_MAX_VCPUS;
 }
 
-static void set_default_spectre(struct kvm *kvm)
-{
-	/*
-	 * The default is to expose CSV2 == 1 if the HW isn't affected.
-	 * Although this is a per-CPU feature, we make it global because
-	 * asymmetric systems are just a nuisance.
-	 *
-	 * Userspace can override this as long as it doesn't promise
-	 * the impossible.
-	 */
-	if (arm64_get_spectre_v2_state() == SPECTRE_UNAFFECTED)
-		kvm->arch.pfr0_csv2 = 1;
-	if (arm64_get_meltdown_state() == SPECTRE_UNAFFECTED)
-		kvm->arch.pfr0_csv3 = 1;
-}
-
 /**
  * kvm_arch_init_vm - initializes a VM data structure
  * @kvm:	pointer to the KVM struct
@@ -161,14 +166,9 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
 	/* The maximum number of VCPUs is limited by the host's GIC model */
 	kvm->max_vcpus = kvm_arm_default_max_vcpus();
 
-	set_default_spectre(kvm);
 	kvm_arm_init_hypercalls(kvm);
 
-	/*
+	bitmap_zero(kvm->arch.vcpu_features, KVM_VCPU_MAX_FEATURES);
-	 * Initialise the default PMUver before there is a chance to
-	 * create an actual PMU.
-	 */
-	kvm->arch.dfr0_pmuver.imp = kvm_arm_pmu_get_pmuver_limit();
 
 	return 0;
 
@@ -302,6 +302,15 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 	case KVM_CAP_ARM_PTRAUTH_GENERIC:
 		r = system_has_full_ptr_auth();
 		break;
+	case KVM_CAP_ARM_EAGER_SPLIT_CHUNK_SIZE:
+		if (kvm)
+			r = kvm->arch.mmu.split_page_chunk_size;
+		else
+			r = KVM_ARM_EAGER_SPLIT_CHUNK_SIZE_DEFAULT;
+		break;
+	case KVM_CAP_ARM_SUPPORTED_BLOCK_SIZES:
+		r = kvm_supported_block_sizes();
+		break;
 	default:
 		r = 0;
 	}
@@ -1167,58 +1176,115 @@ int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level,
 	return -EINVAL;
 }
 
-static int kvm_vcpu_set_target(struct kvm_vcpu *vcpu,
+static int kvm_vcpu_init_check_features(struct kvm_vcpu *vcpu,
-			       const struct kvm_vcpu_init *init)
+					const struct kvm_vcpu_init *init)
 {
-	unsigned int i, ret;
+	unsigned long features = init->features[0];
-	u32 phys_target = kvm_target_cpu();
+	int i;
 
-	if (init->target != phys_target)
+	if (features & ~KVM_VCPU_VALID_FEATURES)
-		return -EINVAL;
+		return -ENOENT;
 
-	/*
+	for (i = 1; i < ARRAY_SIZE(init->features); i++) {
-	 * Secondary and subsequent calls to KVM_ARM_VCPU_INIT must
+		if (init->features[i])
-	 * use the same target.
-	 */
-	if (vcpu->arch.target != -1 && vcpu->arch.target != init->target)
-		return -EINVAL;
-
-	/* -ENOENT for unknown features, -EINVAL for invalid combinations. */
-	for (i = 0; i < sizeof(init->features) * 8; i++) {
-		bool set = (init->features[i / 32] & (1 << (i % 32)));
-
-		if (set && i >= KVM_VCPU_MAX_FEATURES)
 			return -ENOENT;
-
-		/*
-		 * Secondary and subsequent calls to KVM_ARM_VCPU_INIT must
-		 * use the same feature set.
-		 */
-		if (vcpu->arch.target != -1 && i < KVM_VCPU_MAX_FEATURES &&
-		    test_bit(i, vcpu->arch.features) != set)
-			return -EINVAL;
-
-		if (set)
-			set_bit(i, vcpu->arch.features);
 	}
 
-	vcpu->arch.target = phys_target;
+	if (!test_bit(KVM_ARM_VCPU_EL1_32BIT, &features))
+		return 0;
+
+	if (!cpus_have_const_cap(ARM64_HAS_32BIT_EL1))
+		return -EINVAL;
+
+	/* MTE is incompatible with AArch32 */
+	if (kvm_has_mte(vcpu->kvm))
+		return -EINVAL;
+
+	/* NV is incompatible with AArch32 */
+	if (test_bit(KVM_ARM_VCPU_HAS_EL2, &features))
+		return -EINVAL;
+
+	return 0;
+}
+
+static bool kvm_vcpu_init_changed(struct kvm_vcpu *vcpu,
+				  const struct kvm_vcpu_init *init)
+{
+	unsigned long features = init->features[0];
+
+	return !bitmap_equal(vcpu->arch.features, &features, KVM_VCPU_MAX_FEATURES) ||
+			vcpu->arch.target != init->target;
+}
+
+static int __kvm_vcpu_set_target(struct kvm_vcpu *vcpu,
+				 const struct kvm_vcpu_init *init)
+{
+	unsigned long features = init->features[0];
+	struct kvm *kvm = vcpu->kvm;
+	int ret = -EINVAL;
+
+	mutex_lock(&kvm->arch.config_lock);
+
+	if (test_bit(KVM_ARCH_FLAG_VCPU_FEATURES_CONFIGURED, &kvm->arch.flags) &&
+	    !bitmap_equal(kvm->arch.vcpu_features, &features, KVM_VCPU_MAX_FEATURES))
+		goto out_unlock;
+
+	vcpu->arch.target = init->target;
+	bitmap_copy(vcpu->arch.features, &features, KVM_VCPU_MAX_FEATURES);
 
 	/* Now we know what it is, we can reset it. */
 	ret = kvm_reset_vcpu(vcpu);
 	if (ret) {
 		vcpu->arch.target = -1;
 		bitmap_zero(vcpu->arch.features, KVM_VCPU_MAX_FEATURES);
+		goto out_unlock;
 	}
 
+	bitmap_copy(kvm->arch.vcpu_features, &features, KVM_VCPU_MAX_FEATURES);
+	set_bit(KVM_ARCH_FLAG_VCPU_FEATURES_CONFIGURED, &kvm->arch.flags);
+
+out_unlock:
+	mutex_unlock(&kvm->arch.config_lock);
 	return ret;
 }
 
+static int kvm_vcpu_set_target(struct kvm_vcpu *vcpu,
+			       const struct kvm_vcpu_init *init)
+{
+	int ret;
+
+	if (init->target != kvm_target_cpu())
+		return -EINVAL;
+
+	ret = kvm_vcpu_init_check_features(vcpu, init);
+	if (ret)
+		return ret;
+
+	if (vcpu->arch.target == -1)
+		return __kvm_vcpu_set_target(vcpu, init);
+
+	if (kvm_vcpu_init_changed(vcpu, init))
+		return -EINVAL;
+
+	return kvm_reset_vcpu(vcpu);
+}
+
 static int kvm_arch_vcpu_ioctl_vcpu_init(struct kvm_vcpu *vcpu,
 					 struct kvm_vcpu_init *init)
 {
+	bool power_off = false;
 	int ret;
 
+	/*
+	 * Treat the power-off vCPU feature as ephemeral. Clear the bit to avoid
+	 * reflecting it in the finalized feature set, thus limiting its scope
+	 * to a single KVM_ARM_VCPU_INIT call.
+	 */
+	if (init->features[0] & BIT(KVM_ARM_VCPU_POWER_OFF)) {
+		init->features[0] &= ~BIT(KVM_ARM_VCPU_POWER_OFF);
+		power_off = true;
+	}
+
 	ret = kvm_vcpu_set_target(vcpu, init);
 	if (ret)
 		return ret;
@@ -1240,14 +1306,14 @@ static int kvm_arch_vcpu_ioctl_vcpu_init(struct kvm_vcpu *vcpu,
 	}
 
 	vcpu_reset_hcr(vcpu);
-	vcpu->arch.cptr_el2 = CPTR_EL2_DEFAULT;
+	vcpu->arch.cptr_el2 = kvm_get_reset_cptr_el2(vcpu);
 
 	/*
 	 * Handle the "start in power-off" case.
 	 */
 	spin_lock(&vcpu->arch.mp_state_lock);
 
-	if (test_bit(KVM_ARM_VCPU_POWER_OFF, vcpu->arch.features))
+	if (power_off)
 		__kvm_arm_vcpu_power_off(vcpu);
 	else
 		WRITE_ONCE(vcpu->arch.mp_state.mp_state, KVM_MP_STATE_RUNNABLE);
@@ -1666,7 +1732,13 @@ static void __init cpu_prepare_hyp_mode(int cpu, u32 hyp_va_bits)
 
 	params->mair_el2 = read_sysreg(mair_el1);
 
-	tcr = (read_sysreg(tcr_el1) & TCR_EL2_MASK) | TCR_EL2_RES1;
+	tcr = read_sysreg(tcr_el1);
+	if (cpus_have_final_cap(ARM64_KVM_HVHE)) {
+		tcr |= TCR_EPD1_MASK;
+	} else {
+		tcr &= TCR_EL2_MASK;
+		tcr |= TCR_EL2_RES1;
+	}
 	tcr &= ~TCR_T0SZ_MASK;
 	tcr |= TCR_T0SZ(hyp_va_bits);
 	params->tcr_el2 = tcr;
@@ -1676,6 +1748,8 @@ static void __init cpu_prepare_hyp_mode(int cpu, u32 hyp_va_bits)
 		params->hcr_el2 = HCR_HOST_NVHE_PROTECTED_FLAGS;
 	else
 		params->hcr_el2 = HCR_HOST_NVHE_FLAGS;
+	if (cpus_have_final_cap(ARM64_KVM_HVHE))
+		params->hcr_el2 |= HCR_E2H;
 	params->vttbr = params->vtcr = 0;
 
 	/*
@@ -1910,6 +1984,7 @@ static bool __init init_psci_relay(void)
 	}
 
 	kvm_host_psci_config.version = psci_ops.get_version();
+	kvm_host_psci_config.smccc_version = arm_smccc_get_version();
 
 	if (kvm_host_psci_config.version == PSCI_VERSION(0, 1)) {
 		kvm_host_psci_config.function_ids_0_1 = get_psci_0_1_function_ids();
@@ -2067,6 +2142,26 @@ static int __init kvm_hyp_init_protection(u32 hyp_va_bits)
 	return 0;
 }
 
+static void pkvm_hyp_init_ptrauth(void)
+{
+	struct kvm_cpu_context *hyp_ctxt;
+	int cpu;
+
+	for_each_possible_cpu(cpu) {
+		hyp_ctxt = per_cpu_ptr_nvhe_sym(kvm_hyp_ctxt, cpu);
+		hyp_ctxt->sys_regs[APIAKEYLO_EL1] = get_random_long();
+		hyp_ctxt->sys_regs[APIAKEYHI_EL1] = get_random_long();
+		hyp_ctxt->sys_regs[APIBKEYLO_EL1] = get_random_long();
+		hyp_ctxt->sys_regs[APIBKEYHI_EL1] = get_random_long();
+		hyp_ctxt->sys_regs[APDAKEYLO_EL1] = get_random_long();
+		hyp_ctxt->sys_regs[APDAKEYHI_EL1] = get_random_long();
+		hyp_ctxt->sys_regs[APDBKEYLO_EL1] = get_random_long();
+		hyp_ctxt->sys_regs[APDBKEYHI_EL1] = get_random_long();
+		hyp_ctxt->sys_regs[APGAKEYLO_EL1] = get_random_long();
+		hyp_ctxt->sys_regs[APGAKEYHI_EL1] = get_random_long();
+	}
+}
+
 /* Inits Hyp-mode on all online CPUs */
 static int __init init_hyp_mode(void)
 {
@@ -2228,6 +2323,10 @@ static int __init init_hyp_mode(void)
 	kvm_hyp_init_symbols();
 
 	if (is_protected_kvm_enabled()) {
+		if (IS_ENABLED(CONFIG_ARM64_PTR_AUTH_KERNEL) &&
+		    cpus_have_const_cap(ARM64_HAS_ADDRESS_AUTH))
+			pkvm_hyp_init_ptrauth();
+
 		init_cpu_logical_map();
 
 		if (!init_psci_relay()) {

arch/arm64/kvm/fpsimd.c

@@ -180,7 +180,7 @@ void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu)
 
 	/*
 	 * If we have VHE then the Hyp code will reset CPACR_EL1 to
-	 * CPACR_EL1_DEFAULT and we need to reenable SME.
+	 * the default value and we need to reenable SME.
 	 */
 	if (has_vhe() && system_supports_sme()) {
 		/* Also restore EL0 state seen on entry */
@@ -210,7 +210,7 @@ void kvm_arch_vcpu_put_fp(struct kvm_vcpu *vcpu)
 		/*
 		 * The FPSIMD/SVE state in the CPU has not been touched, and we
 		 * have SVE (and VHE): CPACR_EL1 (alias CPTR_EL2) has been
-		 * reset to CPACR_EL1_DEFAULT by the Hyp code, disabling SVE
+		 * reset by kvm_reset_cptr_el2() in the Hyp code, disabling SVE
 		 * for EL0.  To avoid spurious traps, restore the trap state
 		 * seen by kvm_arch_vcpu_load_fp():
 		 */

arch/arm64/kvm/hyp/include/hyp/switch.h

@@ -70,6 +70,56 @@ static inline void __activate_traps_fpsimd32(struct kvm_vcpu *vcpu)
 	}
 }
 
+static inline bool __hfgxtr_traps_required(void)
+{
+	if (cpus_have_final_cap(ARM64_SME))
+		return true;
+
+	if (cpus_have_final_cap(ARM64_WORKAROUND_AMPERE_AC03_CPU_38))
+		return true;
+
+	return false;
+}
+
+static inline void __activate_traps_hfgxtr(void)
+{
+	u64 r_clr = 0, w_clr = 0, r_set = 0, w_set = 0, tmp;
+
+	if (cpus_have_final_cap(ARM64_SME)) {
+		tmp = HFGxTR_EL2_nSMPRI_EL1_MASK | HFGxTR_EL2_nTPIDR2_EL0_MASK;
+
+		r_clr |= tmp;
+		w_clr |= tmp;
+	}
+
+	/*
+	 * Trap guest writes to TCR_EL1 to prevent it from enabling HA or HD.
+	 */
+	if (cpus_have_final_cap(ARM64_WORKAROUND_AMPERE_AC03_CPU_38))
+		w_set |= HFGxTR_EL2_TCR_EL1_MASK;
+
+	sysreg_clear_set_s(SYS_HFGRTR_EL2, r_clr, r_set);
+	sysreg_clear_set_s(SYS_HFGWTR_EL2, w_clr, w_set);
+}
+
+static inline void __deactivate_traps_hfgxtr(void)
+{
+	u64 r_clr = 0, w_clr = 0, r_set = 0, w_set = 0, tmp;
+
+	if (cpus_have_final_cap(ARM64_SME)) {
+		tmp = HFGxTR_EL2_nSMPRI_EL1_MASK | HFGxTR_EL2_nTPIDR2_EL0_MASK;
+
+		r_set |= tmp;
+		w_set |= tmp;
+	}
+
+	if (cpus_have_final_cap(ARM64_WORKAROUND_AMPERE_AC03_CPU_38))
+		w_clr |= HFGxTR_EL2_TCR_EL1_MASK;
+
+	sysreg_clear_set_s(SYS_HFGRTR_EL2, r_clr, r_set);
+	sysreg_clear_set_s(SYS_HFGWTR_EL2, w_clr, w_set);
+}
+
 static inline void __activate_traps_common(struct kvm_vcpu *vcpu)
 {
 	/* Trap on AArch32 cp15 c15 (impdef sysregs) accesses (EL1 or EL0) */
@@ -95,16 +145,8 @@ static inline void __activate_traps_common(struct kvm_vcpu *vcpu)
 	vcpu->arch.mdcr_el2_host = read_sysreg(mdcr_el2);
 	write_sysreg(vcpu->arch.mdcr_el2, mdcr_el2);
 
-	if (cpus_have_final_cap(ARM64_SME)) {
+	if (__hfgxtr_traps_required())
-		sysreg_clear_set_s(SYS_HFGRTR_EL2,
+		__activate_traps_hfgxtr();
-				   HFGxTR_EL2_nSMPRI_EL1_MASK |
-				   HFGxTR_EL2_nTPIDR2_EL0_MASK,
-				   0);
-		sysreg_clear_set_s(SYS_HFGWTR_EL2,
-				   HFGxTR_EL2_nSMPRI_EL1_MASK |
-				   HFGxTR_EL2_nTPIDR2_EL0_MASK,
-				   0);
-	}
 }
 
 static inline void __deactivate_traps_common(struct kvm_vcpu *vcpu)
@@ -120,14 +162,8 @@ static inline void __deactivate_traps_common(struct kvm_vcpu *vcpu)
 		vcpu_clear_flag(vcpu, PMUSERENR_ON_CPU);
 	}
 
-	if (cpus_have_final_cap(ARM64_SME)) {
+	if (__hfgxtr_traps_required())
-		sysreg_clear_set_s(SYS_HFGRTR_EL2, 0,
+		__deactivate_traps_hfgxtr();
-				   HFGxTR_EL2_nSMPRI_EL1_MASK |
-				   HFGxTR_EL2_nTPIDR2_EL0_MASK);
-		sysreg_clear_set_s(SYS_HFGWTR_EL2, 0,
-				   HFGxTR_EL2_nSMPRI_EL1_MASK |
-				   HFGxTR_EL2_nTPIDR2_EL0_MASK);
-	}
 }
 
 static inline void ___activate_traps(struct kvm_vcpu *vcpu)
@@ -209,7 +245,7 @@ static bool kvm_hyp_handle_fpsimd(struct kvm_vcpu *vcpu, u64 *exit_code)
 	/* Valid trap.  Switch the context: */
 
 	/* First disable enough traps to allow us to update the registers */
-	if (has_vhe()) {
+	if (has_vhe() || has_hvhe()) {
 		reg = CPACR_EL1_FPEN_EL0EN | CPACR_EL1_FPEN_EL1EN;
 		if (sve_guest)
 			reg |= CPACR_EL1_ZEN_EL0EN | CPACR_EL1_ZEN_EL1EN;
@@ -401,12 +437,39 @@ static bool kvm_hyp_handle_cntpct(struct kvm_vcpu *vcpu)
 	return true;
 }
 
+static bool handle_ampere1_tcr(struct kvm_vcpu *vcpu)
+{
+	u32 sysreg = esr_sys64_to_sysreg(kvm_vcpu_get_esr(vcpu));
+	int rt = kvm_vcpu_sys_get_rt(vcpu);
+	u64 val = vcpu_get_reg(vcpu, rt);
+
+	if (sysreg != SYS_TCR_EL1)
+		return false;
+
+	/*
+	 * Affected parts do not advertise support for hardware Access Flag /
+	 * Dirty state management in ID_AA64MMFR1_EL1.HAFDBS, but the underlying
+	 * control bits are still functional. The architecture requires these be
+	 * RES0 on systems that do not implement FEAT_HAFDBS.
+	 *
+	 * Uphold the requirements of the architecture by masking guest writes
+	 * to TCR_EL1.{HA,HD} here.
+	 */
+	val &= ~(TCR_HD | TCR_HA);
+	write_sysreg_el1(val, SYS_TCR);
+	return true;
+}
+
 static bool kvm_hyp_handle_sysreg(struct kvm_vcpu *vcpu, u64 *exit_code)
 {
 	if (cpus_have_final_cap(ARM64_WORKAROUND_CAVIUM_TX2_219_TVM) &&
 	    handle_tx2_tvm(vcpu))
 		return true;
 
+	if (cpus_have_final_cap(ARM64_WORKAROUND_AMPERE_AC03_CPU_38) &&
+	    handle_ampere1_tcr(vcpu))
+		return true;
+
 	if (static_branch_unlikely(&vgic_v3_cpuif_trap) &&
 	    __vgic_v3_perform_cpuif_access(vcpu) == 1)
 		return true;

arch/arm64/kvm/hyp/include/nvhe/ffa.h

@@ -0,0 +1,17 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2022 - Google LLC
+ * Author: Andrew Walbran <qwandor@google.com>
+ */
+#ifndef __KVM_HYP_FFA_H
+#define __KVM_HYP_FFA_H
+
+#include <asm/kvm_host.h>
+
+#define FFA_MIN_FUNC_NUM 0x60
+#define FFA_MAX_FUNC_NUM 0x7F
+
+int hyp_ffa_init(void *pages);
+bool kvm_host_ffa_handler(struct kvm_cpu_context *host_ctxt);
+
+#endif /* __KVM_HYP_FFA_H */

arch/arm64/kvm/hyp/include/nvhe/mem_protect.h

@@ -57,6 +57,7 @@ extern struct host_mmu host_mmu;
 enum pkvm_component_id {
 	PKVM_ID_HOST,
 	PKVM_ID_HYP,
+	PKVM_ID_FFA,
 };
 
 extern unsigned long hyp_nr_cpus;
@@ -66,6 +67,8 @@ int __pkvm_host_share_hyp(u64 pfn);
 int __pkvm_host_unshare_hyp(u64 pfn);
 int __pkvm_host_donate_hyp(u64 pfn, u64 nr_pages);
 int __pkvm_hyp_donate_host(u64 pfn, u64 nr_pages);
+int __pkvm_host_share_ffa(u64 pfn, u64 nr_pages);
+int __pkvm_host_unshare_ffa(u64 pfn, u64 nr_pages);
 
 bool addr_is_memory(phys_addr_t phys);
 int host_stage2_idmap_locked(phys_addr_t addr, u64 size, enum kvm_pgtable_prot prot);

arch/arm64/kvm/hyp/nvhe/Makefile

@@ -22,7 +22,7 @@ lib-objs := $(addprefix ../../../lib/, $(lib-objs))
 
 hyp-obj-y := timer-sr.o sysreg-sr.o debug-sr.o switch.o tlb.o hyp-init.o host.o \
 	 hyp-main.o hyp-smp.o psci-relay.o early_alloc.o page_alloc.o \
-	 cache.o setup.o mm.o mem_protect.o sys_regs.o pkvm.o stacktrace.o
+	 cache.o setup.o mm.o mem_protect.o sys_regs.o pkvm.o stacktrace.o ffa.o
 hyp-obj-y += ../vgic-v3-sr.o ../aarch32.o ../vgic-v2-cpuif-proxy.o ../entry.o \
 	 ../fpsimd.o ../hyp-entry.o ../exception.o ../pgtable.o
 hyp-obj-$(CONFIG_DEBUG_LIST) += list_debug.o

arch/arm64/kvm/hyp/nvhe/ffa.c

@@ -0,0 +1,762 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * FF-A v1.0 proxy to filter out invalid memory-sharing SMC calls issued by
+ * the host. FF-A is a slightly more palatable abbreviation of "Arm Firmware
+ * Framework for Arm A-profile", which is specified by Arm in document
+ * number DEN0077.
+ *
+ * Copyright (C) 2022 - Google LLC
+ * Author: Andrew Walbran <qwandor@google.com>
+ *
+ * This driver hooks into the SMC trapping logic for the host and intercepts
+ * all calls falling within the FF-A range. Each call is either:
+ *
+ *	- Forwarded on unmodified to the SPMD at EL3
+ *	- Rejected as "unsupported"
+ *	- Accompanied by a host stage-2 page-table check/update and reissued
+ *
+ * Consequently, any attempts by the host to make guest memory pages
+ * accessible to the secure world using FF-A will be detected either here
+ * (in the case that the memory is already owned by the guest) or during
+ * donation to the guest (in the case that the memory was previously shared
+ * with the secure world).
+ *
+ * To allow the rolling-back of page-table updates and FF-A calls in the
+ * event of failure, operations involving the RXTX buffers are locked for
+ * the duration and are therefore serialised.
+ */
+
+#include <linux/arm-smccc.h>
+#include <linux/arm_ffa.h>
+#include <asm/kvm_pkvm.h>
+
+#include <nvhe/ffa.h>
+#include <nvhe/mem_protect.h>
+#include <nvhe/memory.h>
+#include <nvhe/trap_handler.h>
+#include <nvhe/spinlock.h>
+
+/*
+ * "ID value 0 must be returned at the Non-secure physical FF-A instance"
+ * We share this ID with the host.
+ */
+#define HOST_FFA_ID	0
+
+/*
+ * A buffer to hold the maximum descriptor size we can see from the host,
+ * which is required when the SPMD returns a fragmented FFA_MEM_RETRIEVE_RESP
+ * when resolving the handle on the reclaim path.
+ */
+struct kvm_ffa_descriptor_buffer {
+	void	*buf;
+	size_t	len;
+};
+
+static struct kvm_ffa_descriptor_buffer ffa_desc_buf;
+
+struct kvm_ffa_buffers {
+	hyp_spinlock_t lock;
+	void *tx;
+	void *rx;
+};
+
+/*
+ * Note that we don't currently lock these buffers explicitly, instead
+ * relying on the locking of the host FFA buffers as we only have one
+ * client.
+ */
+static struct kvm_ffa_buffers hyp_buffers;
+static struct kvm_ffa_buffers host_buffers;
+
+static void ffa_to_smccc_error(struct arm_smccc_res *res, u64 ffa_errno)
+{
+	*res = (struct arm_smccc_res) {
+		.a0	= FFA_ERROR,
+		.a2	= ffa_errno,
+	};
+}
+
+static void ffa_to_smccc_res_prop(struct arm_smccc_res *res, int ret, u64 prop)
+{
+	if (ret == FFA_RET_SUCCESS) {
+		*res = (struct arm_smccc_res) { .a0 = FFA_SUCCESS,
+						.a2 = prop };
+	} else {
+		ffa_to_smccc_error(res, ret);
+	}
+}
+
+static void ffa_to_smccc_res(struct arm_smccc_res *res, int ret)
+{
+	ffa_to_smccc_res_prop(res, ret, 0);
+}
+
+static void ffa_set_retval(struct kvm_cpu_context *ctxt,
+			   struct arm_smccc_res *res)
+{
+	cpu_reg(ctxt, 0) = res->a0;
+	cpu_reg(ctxt, 1) = res->a1;
+	cpu_reg(ctxt, 2) = res->a2;
+	cpu_reg(ctxt, 3) = res->a3;
+}
+
+static bool is_ffa_call(u64 func_id)
+{
+	return ARM_SMCCC_IS_FAST_CALL(func_id) &&
+	       ARM_SMCCC_OWNER_NUM(func_id) == ARM_SMCCC_OWNER_STANDARD &&
+	       ARM_SMCCC_FUNC_NUM(func_id) >= FFA_MIN_FUNC_NUM &&
+	       ARM_SMCCC_FUNC_NUM(func_id) <= FFA_MAX_FUNC_NUM;
+}
+
+static int ffa_map_hyp_buffers(u64 ffa_page_count)
+{
+	struct arm_smccc_res res;
+
+	arm_smccc_1_1_smc(FFA_FN64_RXTX_MAP,
+			  hyp_virt_to_phys(hyp_buffers.tx),
+			  hyp_virt_to_phys(hyp_buffers.rx),
+			  ffa_page_count,
+			  0, 0, 0, 0,
+			  &res);
+
+	return res.a0 == FFA_SUCCESS ? FFA_RET_SUCCESS : res.a2;
+}
+
+static int ffa_unmap_hyp_buffers(void)
+{
+	struct arm_smccc_res res;
+
+	arm_smccc_1_1_smc(FFA_RXTX_UNMAP,
+			  HOST_FFA_ID,
+			  0, 0, 0, 0, 0, 0,
+			  &res);
+
+	return res.a0 == FFA_SUCCESS ? FFA_RET_SUCCESS : res.a2;
+}
+
+static void ffa_mem_frag_tx(struct arm_smccc_res *res, u32 handle_lo,
+			     u32 handle_hi, u32 fraglen, u32 endpoint_id)
+{
+	arm_smccc_1_1_smc(FFA_MEM_FRAG_TX,
+			  handle_lo, handle_hi, fraglen, endpoint_id,
+			  0, 0, 0,
+			  res);
+}
+
+static void ffa_mem_frag_rx(struct arm_smccc_res *res, u32 handle_lo,
+			     u32 handle_hi, u32 fragoff)
+{
+	arm_smccc_1_1_smc(FFA_MEM_FRAG_RX,
+			  handle_lo, handle_hi, fragoff, HOST_FFA_ID,
+			  0, 0, 0,
+			  res);
+}
+
+static void ffa_mem_xfer(struct arm_smccc_res *res, u64 func_id, u32 len,
+			  u32 fraglen)
+{
+	arm_smccc_1_1_smc(func_id, len, fraglen,
+			  0, 0, 0, 0, 0,
+			  res);
+}
+
+static void ffa_mem_reclaim(struct arm_smccc_res *res, u32 handle_lo,
+			     u32 handle_hi, u32 flags)
+{
+	arm_smccc_1_1_smc(FFA_MEM_RECLAIM,
+			  handle_lo, handle_hi, flags,
+			  0, 0, 0, 0,
+			  res);
+}
+
+static void ffa_retrieve_req(struct arm_smccc_res *res, u32 len)
+{
+	arm_smccc_1_1_smc(FFA_FN64_MEM_RETRIEVE_REQ,
+			  len, len,
+			  0, 0, 0, 0, 0,
+			  res);
+}
+
+static void do_ffa_rxtx_map(struct arm_smccc_res *res,
+			    struct kvm_cpu_context *ctxt)
+{
+	DECLARE_REG(phys_addr_t, tx, ctxt, 1);
+	DECLARE_REG(phys_addr_t, rx, ctxt, 2);
+	DECLARE_REG(u32, npages, ctxt, 3);
+	int ret = 0;
+	void *rx_virt, *tx_virt;
+
+	if (npages != (KVM_FFA_MBOX_NR_PAGES * PAGE_SIZE) / FFA_PAGE_SIZE) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto out;
+	}
+
+	if (!PAGE_ALIGNED(tx) || !PAGE_ALIGNED(rx)) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto out;
+	}
+
+	hyp_spin_lock(&host_buffers.lock);
+	if (host_buffers.tx) {
+		ret = FFA_RET_DENIED;
+		goto out_unlock;
+	}
+
+	/*
+	 * Map our hypervisor buffers into the SPMD before mapping and
+	 * pinning the host buffers in our own address space.
+	 */
+	ret = ffa_map_hyp_buffers(npages);
+	if (ret)
+		goto out_unlock;
+
+	ret = __pkvm_host_share_hyp(hyp_phys_to_pfn(tx));
+	if (ret) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto err_unmap;
+	}
+
+	ret = __pkvm_host_share_hyp(hyp_phys_to_pfn(rx));
+	if (ret) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto err_unshare_tx;
+	}
+
+	tx_virt = hyp_phys_to_virt(tx);
+	ret = hyp_pin_shared_mem(tx_virt, tx_virt + 1);
+	if (ret) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto err_unshare_rx;
+	}
+
+	rx_virt = hyp_phys_to_virt(rx);
+	ret = hyp_pin_shared_mem(rx_virt, rx_virt + 1);
+	if (ret) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto err_unpin_tx;
+	}
+
+	host_buffers.tx = tx_virt;
+	host_buffers.rx = rx_virt;
+
+out_unlock:
+	hyp_spin_unlock(&host_buffers.lock);
+out:
+	ffa_to_smccc_res(res, ret);
+	return;
+
+err_unpin_tx:
+	hyp_unpin_shared_mem(tx_virt, tx_virt + 1);
+err_unshare_rx:
+	__pkvm_host_unshare_hyp(hyp_phys_to_pfn(rx));
+err_unshare_tx:
+	__pkvm_host_unshare_hyp(hyp_phys_to_pfn(tx));
+err_unmap:
+	ffa_unmap_hyp_buffers();
+	goto out_unlock;
+}
+
+static void do_ffa_rxtx_unmap(struct arm_smccc_res *res,
+			      struct kvm_cpu_context *ctxt)
+{
+	DECLARE_REG(u32, id, ctxt, 1);
+	int ret = 0;
+
+	if (id != HOST_FFA_ID) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto out;
+	}
+
+	hyp_spin_lock(&host_buffers.lock);
+	if (!host_buffers.tx) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto out_unlock;
+	}
+
+	hyp_unpin_shared_mem(host_buffers.tx, host_buffers.tx + 1);
+	WARN_ON(__pkvm_host_unshare_hyp(hyp_virt_to_pfn(host_buffers.tx)));
+	host_buffers.tx = NULL;
+
+	hyp_unpin_shared_mem(host_buffers.rx, host_buffers.rx + 1);
+	WARN_ON(__pkvm_host_unshare_hyp(hyp_virt_to_pfn(host_buffers.rx)));
+	host_buffers.rx = NULL;
+
+	ffa_unmap_hyp_buffers();
+
+out_unlock:
+	hyp_spin_unlock(&host_buffers.lock);
+out:
+	ffa_to_smccc_res(res, ret);
+}
+
+static u32 __ffa_host_share_ranges(struct ffa_mem_region_addr_range *ranges,
+				   u32 nranges)
+{
+	u32 i;
+
+	for (i = 0; i < nranges; ++i) {
+		struct ffa_mem_region_addr_range *range = &ranges[i];
+		u64 sz = (u64)range->pg_cnt * FFA_PAGE_SIZE;
+		u64 pfn = hyp_phys_to_pfn(range->address);
+
+		if (!PAGE_ALIGNED(sz))
+			break;
+
+		if (__pkvm_host_share_ffa(pfn, sz / PAGE_SIZE))
+			break;
+	}
+
+	return i;
+}
+
+static u32 __ffa_host_unshare_ranges(struct ffa_mem_region_addr_range *ranges,
+				     u32 nranges)
+{
+	u32 i;
+
+	for (i = 0; i < nranges; ++i) {
+		struct ffa_mem_region_addr_range *range = &ranges[i];
+		u64 sz = (u64)range->pg_cnt * FFA_PAGE_SIZE;
+		u64 pfn = hyp_phys_to_pfn(range->address);
+
+		if (!PAGE_ALIGNED(sz))
+			break;
+
+		if (__pkvm_host_unshare_ffa(pfn, sz / PAGE_SIZE))
+			break;
+	}
+
+	return i;
+}
+
+static int ffa_host_share_ranges(struct ffa_mem_region_addr_range *ranges,
+				 u32 nranges)
+{
+	u32 nshared = __ffa_host_share_ranges(ranges, nranges);
+	int ret = 0;
+
+	if (nshared != nranges) {
+		WARN_ON(__ffa_host_unshare_ranges(ranges, nshared) != nshared);
+		ret = FFA_RET_DENIED;
+	}
+
+	return ret;
+}
+
+static int ffa_host_unshare_ranges(struct ffa_mem_region_addr_range *ranges,
+				   u32 nranges)
+{
+	u32 nunshared = __ffa_host_unshare_ranges(ranges, nranges);
+	int ret = 0;
+
+	if (nunshared != nranges) {
+		WARN_ON(__ffa_host_share_ranges(ranges, nunshared) != nunshared);
+		ret = FFA_RET_DENIED;
+	}
+
+	return ret;
+}
+
+static void do_ffa_mem_frag_tx(struct arm_smccc_res *res,
+			       struct kvm_cpu_context *ctxt)
+{
+	DECLARE_REG(u32, handle_lo, ctxt, 1);
+	DECLARE_REG(u32, handle_hi, ctxt, 2);
+	DECLARE_REG(u32, fraglen, ctxt, 3);
+	DECLARE_REG(u32, endpoint_id, ctxt, 4);
+	struct ffa_mem_region_addr_range *buf;
+	int ret = FFA_RET_INVALID_PARAMETERS;
+	u32 nr_ranges;
+
+	if (fraglen > KVM_FFA_MBOX_NR_PAGES * PAGE_SIZE)
+		goto out;
+
+	if (fraglen % sizeof(*buf))
+		goto out;
+
+	hyp_spin_lock(&host_buffers.lock);
+	if (!host_buffers.tx)
+		goto out_unlock;
+
+	buf = hyp_buffers.tx;
+	memcpy(buf, host_buffers.tx, fraglen);
+	nr_ranges = fraglen / sizeof(*buf);
+
+	ret = ffa_host_share_ranges(buf, nr_ranges);
+	if (ret) {
+		/*
+		 * We're effectively aborting the transaction, so we need
+		 * to restore the global state back to what it was prior to
+		 * transmission of the first fragment.
+		 */
+		ffa_mem_reclaim(res, handle_lo, handle_hi, 0);
+		WARN_ON(res->a0 != FFA_SUCCESS);
+		goto out_unlock;
+	}
+
+	ffa_mem_frag_tx(res, handle_lo, handle_hi, fraglen, endpoint_id);
+	if (res->a0 != FFA_SUCCESS && res->a0 != FFA_MEM_FRAG_RX)
+		WARN_ON(ffa_host_unshare_ranges(buf, nr_ranges));
+
+out_unlock:
+	hyp_spin_unlock(&host_buffers.lock);
+out:
+	if (ret)
+		ffa_to_smccc_res(res, ret);
+
+	/*
+	 * If for any reason this did not succeed, we're in trouble as we have
+	 * now lost the content of the previous fragments and we can't rollback
+	 * the host stage-2 changes. The pages previously marked as shared will
+	 * remain stuck in that state forever, hence preventing the host from
+	 * sharing/donating them again and may possibly lead to subsequent
+	 * failures, but this will not compromise confidentiality.
+	 */
+	return;
+}
+
+static __always_inline void do_ffa_mem_xfer(const u64 func_id,
+					    struct arm_smccc_res *res,
+					    struct kvm_cpu_context *ctxt)
+{
+	DECLARE_REG(u32, len, ctxt, 1);
+	DECLARE_REG(u32, fraglen, ctxt, 2);
+	DECLARE_REG(u64, addr_mbz, ctxt, 3);
+	DECLARE_REG(u32, npages_mbz, ctxt, 4);
+	struct ffa_composite_mem_region *reg;
+	struct ffa_mem_region *buf;
+	u32 offset, nr_ranges;
+	int ret = 0;
+
+	BUILD_BUG_ON(func_id != FFA_FN64_MEM_SHARE &&
+		     func_id != FFA_FN64_MEM_LEND);
+
+	if (addr_mbz || npages_mbz || fraglen > len ||
+	    fraglen > KVM_FFA_MBOX_NR_PAGES * PAGE_SIZE) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto out;
+	}
+
+	if (fraglen < sizeof(struct ffa_mem_region) +
+		      sizeof(struct ffa_mem_region_attributes)) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto out;
+	}
+
+	hyp_spin_lock(&host_buffers.lock);
+	if (!host_buffers.tx) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto out_unlock;
+	}
+
+	buf = hyp_buffers.tx;
+	memcpy(buf, host_buffers.tx, fraglen);
+
+	offset = buf->ep_mem_access[0].composite_off;
+	if (!offset || buf->ep_count != 1 || buf->sender_id != HOST_FFA_ID) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto out_unlock;
+	}
+
+	if (fraglen < offset + sizeof(struct ffa_composite_mem_region)) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto out_unlock;
+	}
+
+	reg = (void *)buf + offset;
+	nr_ranges = ((void *)buf + fraglen) - (void *)reg->constituents;
+	if (nr_ranges % sizeof(reg->constituents[0])) {
+		ret = FFA_RET_INVALID_PARAMETERS;
+		goto out_unlock;
+	}
+
+	nr_ranges /= sizeof(reg->constituents[0]);
+	ret = ffa_host_share_ranges(reg->constituents, nr_ranges);
+	if (ret)
+		goto out_unlock;
+
+	ffa_mem_xfer(res, func_id, len, fraglen);
+	if (fraglen != len) {
+		if (res->a0 != FFA_MEM_FRAG_RX)
+			goto err_unshare;
+
+		if (res->a3 != fraglen)
+			goto err_unshare;
+	} else if (res->a0 != FFA_SUCCESS) {
+		goto err_unshare;
+	}
+
+out_unlock:
+	hyp_spin_unlock(&host_buffers.lock);
+out:
+	if (ret)
+		ffa_to_smccc_res(res, ret);
+	return;
+
+err_unshare:
+	WARN_ON(ffa_host_unshare_ranges(reg->constituents, nr_ranges));
+	goto out_unlock;
+}
+
+static void do_ffa_mem_reclaim(struct arm_smccc_res *res,
+			       struct kvm_cpu_context *ctxt)
+{
+	DECLARE_REG(u32, handle_lo, ctxt, 1);
+	DECLARE_REG(u32, handle_hi, ctxt, 2);
+	DECLARE_REG(u32, flags, ctxt, 3);
+	struct ffa_composite_mem_region *reg;
+	u32 offset, len, fraglen, fragoff;
+	struct ffa_mem_region *buf;
+	int ret = 0;
+	u64 handle;
+
+	handle = PACK_HANDLE(handle_lo, handle_hi);
+
+	hyp_spin_lock(&host_buffers.lock);
+
+	buf = hyp_buffers.tx;
+	*buf = (struct ffa_mem_region) {
+		.sender_id	= HOST_FFA_ID,
+		.handle		= handle,
+	};
+
+	ffa_retrieve_req(res, sizeof(*buf));
+	buf = hyp_buffers.rx;
+	if (res->a0 != FFA_MEM_RETRIEVE_RESP)
+		goto out_unlock;
+
+	len = res->a1;
+	fraglen = res->a2;
+
+	offset = buf->ep_mem_access[0].composite_off;
+	/*
+	 * We can trust the SPMD to get this right, but let's at least
+	 * check that we end up with something that doesn't look _completely_
+	 * bogus.
+	 */
+	if (WARN_ON(offset > len ||
+		    fraglen > KVM_FFA_MBOX_NR_PAGES * PAGE_SIZE)) {
+		ret = FFA_RET_ABORTED;
+		goto out_unlock;
+	}
+
+	if (len > ffa_desc_buf.len) {
+		ret = FFA_RET_NO_MEMORY;
+		goto out_unlock;
+	}
+
+	buf = ffa_desc_buf.buf;
+	memcpy(buf, hyp_buffers.rx, fraglen);
+
+	for (fragoff = fraglen; fragoff < len; fragoff += fraglen) {
+		ffa_mem_frag_rx(res, handle_lo, handle_hi, fragoff);
+		if (res->a0 != FFA_MEM_FRAG_TX) {
+			ret = FFA_RET_INVALID_PARAMETERS;
+			goto out_unlock;
+		}
+
+		fraglen = res->a3;
+		memcpy((void *)buf + fragoff, hyp_buffers.rx, fraglen);
+	}
+
+	ffa_mem_reclaim(res, handle_lo, handle_hi, flags);
+	if (res->a0 != FFA_SUCCESS)
+		goto out_unlock;
+
+	reg = (void *)buf + offset;
+	/* If the SPMD was happy, then we should be too. */
+	WARN_ON(ffa_host_unshare_ranges(reg->constituents,
+					reg->addr_range_cnt));
+out_unlock:
+	hyp_spin_unlock(&host_buffers.lock);
+
+	if (ret)
+		ffa_to_smccc_res(res, ret);
+}
+
+/*
+ * Is a given FFA function supported, either by forwarding on directly
+ * or by handling at EL2?
+ */
+static bool ffa_call_supported(u64 func_id)
+{
+	switch (func_id) {
+	/* Unsupported memory management calls */
+	case FFA_FN64_MEM_RETRIEVE_REQ:
+	case FFA_MEM_RETRIEVE_RESP:
+	case FFA_MEM_RELINQUISH:
+	case FFA_MEM_OP_PAUSE:
+	case FFA_MEM_OP_RESUME:
+	case FFA_MEM_FRAG_RX:
+	case FFA_FN64_MEM_DONATE:
+	/* Indirect message passing via RX/TX buffers */
+	case FFA_MSG_SEND:
+	case FFA_MSG_POLL:
+	case FFA_MSG_WAIT:
+	/* 32-bit variants of 64-bit calls */
+	case FFA_MSG_SEND_DIRECT_REQ:
+	case FFA_MSG_SEND_DIRECT_RESP:
+	case FFA_RXTX_MAP:
+	case FFA_MEM_DONATE:
+	case FFA_MEM_RETRIEVE_REQ:
+		return false;
+	}
+
+	return true;
+}
+
+static bool do_ffa_features(struct arm_smccc_res *res,
+			    struct kvm_cpu_context *ctxt)
+{
+	DECLARE_REG(u32, id, ctxt, 1);
+	u64 prop = 0;
+	int ret = 0;
+
+	if (!ffa_call_supported(id)) {
+		ret = FFA_RET_NOT_SUPPORTED;
+		goto out_handled;
+	}
+
+	switch (id) {
+	case FFA_MEM_SHARE:
+	case FFA_FN64_MEM_SHARE:
+	case FFA_MEM_LEND:
+	case FFA_FN64_MEM_LEND:
+		ret = FFA_RET_SUCCESS;
+		prop = 0; /* No support for dynamic buffers */
+		goto out_handled;
+	default:
+		return false;
+	}
+
+out_handled:
+	ffa_to_smccc_res_prop(res, ret, prop);
+	return true;
+}
+
+bool kvm_host_ffa_handler(struct kvm_cpu_context *host_ctxt)
+{
+	DECLARE_REG(u64, func_id, host_ctxt, 0);
+	struct arm_smccc_res res;
+
+	/*
+	 * There's no way we can tell what a non-standard SMC call might
+	 * be up to. Ideally, we would terminate these here and return
+	 * an error to the host, but sadly devices make use of custom
+	 * firmware calls for things like power management, debugging,
+	 * RNG access and crash reporting.
+	 *
+	 * Given that the architecture requires us to trust EL3 anyway,
+	 * we forward unrecognised calls on under the assumption that
+	 * the firmware doesn't expose a mechanism to access arbitrary
+	 * non-secure memory. Short of a per-device table of SMCs, this
+	 * is the best we can do.
+	 */
+	if (!is_ffa_call(func_id))
+		return false;
+
+	switch (func_id) {
+	case FFA_FEATURES:
+		if (!do_ffa_features(&res, host_ctxt))
+			return false;
+		goto out_handled;
+	/* Memory management */
+	case FFA_FN64_RXTX_MAP:
+		do_ffa_rxtx_map(&res, host_ctxt);
+		goto out_handled;
+	case FFA_RXTX_UNMAP:
+		do_ffa_rxtx_unmap(&res, host_ctxt);
+		goto out_handled;
+	case FFA_MEM_SHARE:
+	case FFA_FN64_MEM_SHARE:
+		do_ffa_mem_xfer(FFA_FN64_MEM_SHARE, &res, host_ctxt);
+		goto out_handled;
+	case FFA_MEM_RECLAIM:
+		do_ffa_mem_reclaim(&res, host_ctxt);
+		goto out_handled;
+	case FFA_MEM_LEND:
+	case FFA_FN64_MEM_LEND:
+		do_ffa_mem_xfer(FFA_FN64_MEM_LEND, &res, host_ctxt);
+		goto out_handled;
+	case FFA_MEM_FRAG_TX:
+		do_ffa_mem_frag_tx(&res, host_ctxt);
+		goto out_handled;
+	}
+
+	if (ffa_call_supported(func_id))
+		return false; /* Pass through */
+
+	ffa_to_smccc_error(&res, FFA_RET_NOT_SUPPORTED);
+out_handled:
+	ffa_set_retval(host_ctxt, &res);
+	return true;
+}
+
+int hyp_ffa_init(void *pages)
+{
+	struct arm_smccc_res res;
+	size_t min_rxtx_sz;
+	void *tx, *rx;
+
+	if (kvm_host_psci_config.smccc_version < ARM_SMCCC_VERSION_1_2)
+		return 0;
+
+	arm_smccc_1_1_smc(FFA_VERSION, FFA_VERSION_1_0, 0, 0, 0, 0, 0, 0, &res);
+	if (res.a0 == FFA_RET_NOT_SUPPORTED)
+		return 0;
+
+	if (res.a0 != FFA_VERSION_1_0)
+		return -EOPNOTSUPP;
+
+	arm_smccc_1_1_smc(FFA_ID_GET, 0, 0, 0, 0, 0, 0, 0, &res);
+	if (res.a0 != FFA_SUCCESS)
+		return -EOPNOTSUPP;
+
+	if (res.a2 != HOST_FFA_ID)
+		return -EINVAL;
+
+	arm_smccc_1_1_smc(FFA_FEATURES, FFA_FN64_RXTX_MAP,
+			  0, 0, 0, 0, 0, 0, &res);
+	if (res.a0 != FFA_SUCCESS)
+		return -EOPNOTSUPP;
+
+	switch (res.a2) {
+	case FFA_FEAT_RXTX_MIN_SZ_4K:
+		min_rxtx_sz = SZ_4K;
+		break;
+	case FFA_FEAT_RXTX_MIN_SZ_16K:
+		min_rxtx_sz = SZ_16K;
+		break;
+	case FFA_FEAT_RXTX_MIN_SZ_64K:
+		min_rxtx_sz = SZ_64K;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (min_rxtx_sz > PAGE_SIZE)
+		return -EOPNOTSUPP;
+
+	tx = pages;
+	pages += KVM_FFA_MBOX_NR_PAGES * PAGE_SIZE;
+	rx = pages;
+	pages += KVM_FFA_MBOX_NR_PAGES * PAGE_SIZE;
+
+	ffa_desc_buf = (struct kvm_ffa_descriptor_buffer) {
+		.buf	= pages,
+		.len	= PAGE_SIZE *
+			  (hyp_ffa_proxy_pages() - (2 * KVM_FFA_MBOX_NR_PAGES)),
+	};
+
+	hyp_buffers = (struct kvm_ffa_buffers) {
+		.lock	= __HYP_SPIN_LOCK_UNLOCKED,
+		.tx	= tx,
+		.rx	= rx,
+	};
+
+	host_buffers = (struct kvm_ffa_buffers) {
+		.lock	= __HYP_SPIN_LOCK_UNLOCKED,
+	};
+
+	return 0;
+}

arch/arm64/kvm/hyp/nvhe/host.S

@@ -10,6 +10,7 @@
 #include <asm/kvm_arm.h>
 #include <asm/kvm_asm.h>
 #include <asm/kvm_mmu.h>
+#include <asm/kvm_ptrauth.h>
 
 	.text
 
@@ -37,10 +38,43 @@ SYM_FUNC_START(__host_exit)
 
 	/* Save the host context pointer in x29 across the function call */
 	mov	x29, x0
+
+#ifdef CONFIG_ARM64_PTR_AUTH_KERNEL
+alternative_if_not ARM64_HAS_ADDRESS_AUTH
+b __skip_pauth_save
+alternative_else_nop_endif
+
+alternative_if ARM64_KVM_PROTECTED_MODE
+	/* Save kernel ptrauth keys. */
+	add x18, x29, #CPU_APIAKEYLO_EL1
+	ptrauth_save_state x18, x19, x20
+
+	/* Use hyp keys. */
+	adr_this_cpu x18, kvm_hyp_ctxt, x19
+	add x18, x18, #CPU_APIAKEYLO_EL1
+	ptrauth_restore_state x18, x19, x20
+	isb
+alternative_else_nop_endif
+__skip_pauth_save:
+#endif /* CONFIG_ARM64_PTR_AUTH_KERNEL */
+
 	bl	handle_trap
 
-	/* Restore host regs x0-x17 */
 __host_enter_restore_full:
+	/* Restore kernel keys. */
+#ifdef CONFIG_ARM64_PTR_AUTH_KERNEL
+alternative_if_not ARM64_HAS_ADDRESS_AUTH
+b __skip_pauth_restore
+alternative_else_nop_endif
+
+alternative_if ARM64_KVM_PROTECTED_MODE
+	add x18, x29, #CPU_APIAKEYLO_EL1
+	ptrauth_restore_state x18, x19, x20
+alternative_else_nop_endif
+__skip_pauth_restore:
+#endif /* CONFIG_ARM64_PTR_AUTH_KERNEL */
+
+	/* Restore host regs x0-x17 */
 	ldp	x0, x1,   [x29, #CPU_XREG_OFFSET(0)]
 	ldp	x2, x3,   [x29, #CPU_XREG_OFFSET(2)]
 	ldp	x4, x5,   [x29, #CPU_XREG_OFFSET(4)]

arch/arm64/kvm/hyp/nvhe/hyp-init.S

@@ -83,9 +83,6 @@ SYM_CODE_END(__kvm_hyp_init)
  * x0: struct kvm_nvhe_init_params PA
  */
 SYM_CODE_START_LOCAL(___kvm_hyp_init)
-	ldr	x1, [x0, #NVHE_INIT_TPIDR_EL2]
-	msr	tpidr_el2, x1
-
 	ldr	x1, [x0, #NVHE_INIT_STACK_HYP_VA]
 	mov	sp, x1
 
@@ -95,6 +92,22 @@ SYM_CODE_START_LOCAL(___kvm_hyp_init)
 	ldr	x1, [x0, #NVHE_INIT_HCR_EL2]
 	msr	hcr_el2, x1
 
+	mov	x2, #HCR_E2H
+	and	x2, x1, x2
+	cbz	x2, 1f
+
+	// hVHE: Replay the EL2 setup to account for the E2H bit
+	// TPIDR_EL2 is used to preserve x0 across the macro maze...
+	isb
+	msr	tpidr_el2, x0
+	init_el2_state
+	finalise_el2_state
+	mrs	x0, tpidr_el2
+
+1:
+	ldr	x1, [x0, #NVHE_INIT_TPIDR_EL2]
+	msr	tpidr_el2, x1
+
 	ldr	x1, [x0, #NVHE_INIT_VTTBR]
 	msr	vttbr_el2, x1
 
@@ -128,6 +141,13 @@ alternative_if ARM64_HAS_ADDRESS_AUTH
 		     SCTLR_ELx_ENDA | SCTLR_ELx_ENDB)
 	orr	x0, x0, x1
 alternative_else_nop_endif
+
+#ifdef CONFIG_ARM64_BTI_KERNEL
+alternative_if ARM64_BTI
+	orr	x0, x0, #SCTLR_EL2_BT
+alternative_else_nop_endif
+#endif /* CONFIG_ARM64_BTI_KERNEL */
+
 	msr	sctlr_el2, x0
 	isb
 
@@ -184,6 +204,7 @@ SYM_CODE_START_LOCAL(__kvm_hyp_init_cpu)
 	/* Initialize EL2 CPU state to sane values. */
 	init_el2_state				// Clobbers x0..x2
 	finalise_el2_state
+	__init_el2_nvhe_prepare_eret
 
 	/* Enable MMU, set vectors and stack. */
 	mov	x0, x28
@@ -196,6 +217,11 @@ SYM_CODE_START_LOCAL(__kvm_hyp_init_cpu)
 SYM_CODE_END(__kvm_hyp_init_cpu)
 
 SYM_CODE_START(__kvm_handle_stub_hvc)
+	/*
+	 * __kvm_handle_stub_hvc called from __host_hvc through branch instruction(br) so
+	 * we need bti j at beginning.
+	 */
+	bti j
 	cmp	x0, #HVC_SOFT_RESTART
 	b.ne	1f
 

arch/arm64/kvm/hyp/nvhe/hyp-main.c

@@ -13,6 +13,7 @@
 #include <asm/kvm_hyp.h>
 #include <asm/kvm_mmu.h>
 
+#include <nvhe/ffa.h>
 #include <nvhe/mem_protect.h>
 #include <nvhe/mm.h>
 #include <nvhe/pkvm.h>
@@ -125,6 +126,15 @@ static void handle___kvm_tlb_flush_vmid_ipa(struct kvm_cpu_context *host_ctxt)
 	__kvm_tlb_flush_vmid_ipa(kern_hyp_va(mmu), ipa, level);
 }
 
+static void handle___kvm_tlb_flush_vmid_ipa_nsh(struct kvm_cpu_context *host_ctxt)
+{
+	DECLARE_REG(struct kvm_s2_mmu *, mmu, host_ctxt, 1);
+	DECLARE_REG(phys_addr_t, ipa, host_ctxt, 2);
+	DECLARE_REG(int, level, host_ctxt, 3);
+
+	__kvm_tlb_flush_vmid_ipa_nsh(kern_hyp_va(mmu), ipa, level);
+}
+
 static void handle___kvm_tlb_flush_vmid(struct kvm_cpu_context *host_ctxt)
 {
 	DECLARE_REG(struct kvm_s2_mmu *, mmu, host_ctxt, 1);
@@ -315,6 +325,7 @@ static const hcall_t host_hcall[] = {
 	HANDLE_FUNC(__kvm_vcpu_run),
 	HANDLE_FUNC(__kvm_flush_vm_context),
 	HANDLE_FUNC(__kvm_tlb_flush_vmid_ipa),
+	HANDLE_FUNC(__kvm_tlb_flush_vmid_ipa_nsh),
 	HANDLE_FUNC(__kvm_tlb_flush_vmid),
 	HANDLE_FUNC(__kvm_flush_cpu_context),
 	HANDLE_FUNC(__kvm_timer_set_cntvoff),
@@ -373,6 +384,8 @@ static void handle_host_smc(struct kvm_cpu_context *host_ctxt)
 	bool handled;
 
 	handled = kvm_host_psci_handler(host_ctxt);
+	if (!handled)
+		handled = kvm_host_ffa_handler(host_ctxt);
 	if (!handled)
 		default_host_smc_handler(host_ctxt);
 
@@ -392,7 +405,11 @@ void handle_trap(struct kvm_cpu_context *host_ctxt)
 		handle_host_smc(host_ctxt);
 		break;
 	case ESR_ELx_EC_SVE:
-		sysreg_clear_set(cptr_el2, CPTR_EL2_TZ, 0);
+		if (has_hvhe())
+			sysreg_clear_set(cpacr_el1, 0, (CPACR_EL1_ZEN_EL1EN |
+							CPACR_EL1_ZEN_EL0EN));
+		else
+			sysreg_clear_set(cptr_el2, CPTR_EL2_TZ, 0);
 		isb();
 		sve_cond_update_zcr_vq(ZCR_ELx_LEN_MASK, SYS_ZCR_EL2);
 		break;

arch/arm64/kvm/hyp/nvhe/mem_protect.c

@@ -91,9 +91,9 @@ static void host_s2_put_page(void *addr)
 	hyp_put_page(&host_s2_pool, addr);
 }
 
-static void host_s2_free_removed_table(void *addr, u32 level)
+static void host_s2_free_unlinked_table(void *addr, u32 level)
 {
-	kvm_pgtable_stage2_free_removed(&host_mmu.mm_ops, addr, level);
+	kvm_pgtable_stage2_free_unlinked(&host_mmu.mm_ops, addr, level);
 }
 
 static int prepare_s2_pool(void *pgt_pool_base)
@@ -110,7 +110,7 @@ static int prepare_s2_pool(void *pgt_pool_base)
 	host_mmu.mm_ops = (struct kvm_pgtable_mm_ops) {
 		.zalloc_pages_exact = host_s2_zalloc_pages_exact,
 		.zalloc_page = host_s2_zalloc_page,
-		.free_removed_table = host_s2_free_removed_table,
+		.free_unlinked_table = host_s2_free_unlinked_table,
 		.phys_to_virt = hyp_phys_to_virt,
 		.virt_to_phys = hyp_virt_to_phys,
 		.page_count = hyp_page_count,
@@ -842,6 +842,13 @@ static int check_share(struct pkvm_mem_share *share)
 	case PKVM_ID_HYP:
 		ret = hyp_ack_share(completer_addr, tx, share->completer_prot);
 		break;
+	case PKVM_ID_FFA:
+		/*
+		 * We only check the host; the secure side will check the other
+		 * end when we forward the FFA call.
+		 */
+		ret = 0;
+		break;
 	default:
 		ret = -EINVAL;
 	}
@@ -870,6 +877,13 @@ static int __do_share(struct pkvm_mem_share *share)
 	case PKVM_ID_HYP:
 		ret = hyp_complete_share(completer_addr, tx, share->completer_prot);
 		break;
+	case PKVM_ID_FFA:
+		/*
+		 * We're not responsible for any secure page-tables, so there's
+		 * nothing to do here.
+		 */
+		ret = 0;
+		break;
 	default:
 		ret = -EINVAL;
 	}
@@ -918,6 +932,10 @@ static int check_unshare(struct pkvm_mem_share *share)
 	case PKVM_ID_HYP:
 		ret = hyp_ack_unshare(completer_addr, tx);
 		break;
+	case PKVM_ID_FFA:
+		/* See check_share() */
+		ret = 0;
+		break;
 	default:
 		ret = -EINVAL;
 	}
@@ -946,6 +964,10 @@ static int __do_unshare(struct pkvm_mem_share *share)
 	case PKVM_ID_HYP:
 		ret = hyp_complete_unshare(completer_addr, tx);
 		break;
+	case PKVM_ID_FFA:
+		/* See __do_share() */
+		ret = 0;
+		break;
 	default:
 		ret = -EINVAL;
 	}
@@ -1235,3 +1257,49 @@ void hyp_unpin_shared_mem(void *from, void *to)
 	hyp_unlock_component();
 	host_unlock_component();
 }
+
+int __pkvm_host_share_ffa(u64 pfn, u64 nr_pages)
+{
+	int ret;
+	struct pkvm_mem_share share = {
+		.tx	= {
+			.nr_pages	= nr_pages,
+			.initiator	= {
+				.id	= PKVM_ID_HOST,
+				.addr	= hyp_pfn_to_phys(pfn),
+			},
+			.completer	= {
+				.id	= PKVM_ID_FFA,
+			},
+		},
+	};
+
+	host_lock_component();
+	ret = do_share(&share);
+	host_unlock_component();
+
+	return ret;
+}
+
+int __pkvm_host_unshare_ffa(u64 pfn, u64 nr_pages)
+{
+	int ret;
+	struct pkvm_mem_share share = {
+		.tx	= {
+			.nr_pages	= nr_pages,
+			.initiator	= {
+				.id	= PKVM_ID_HOST,
+				.addr	= hyp_pfn_to_phys(pfn),
+			},
+			.completer	= {
+				.id	= PKVM_ID_FFA,
+			},
+		},
+	};
+
+	host_lock_component();
+	ret = do_unshare(&share);
+	host_unlock_component();
+
+	return ret;
+}

arch/arm64/kvm/hyp/nvhe/pkvm.c

@@ -27,6 +27,7 @@ static void pvm_init_traps_aa64pfr0(struct kvm_vcpu *vcpu)
 	u64 hcr_set = HCR_RW;
 	u64 hcr_clear = 0;
 	u64 cptr_set = 0;
+	u64 cptr_clear = 0;
 
 	/* Protected KVM does not support AArch32 guests. */
 	BUILD_BUG_ON(FIELD_GET(ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_EL0),
@@ -43,6 +44,9 @@ static void pvm_init_traps_aa64pfr0(struct kvm_vcpu *vcpu)
 	BUILD_BUG_ON(!FIELD_GET(ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_AdvSIMD),
 				PVM_ID_AA64PFR0_ALLOW));
 
+	if (has_hvhe())
+		hcr_set |= HCR_E2H;
+
 	/* Trap RAS unless all current versions are supported */
 	if (FIELD_GET(ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_RAS), feature_ids) <
 	    ID_AA64PFR0_EL1_RAS_V1P1) {
@@ -57,12 +61,17 @@ static void pvm_init_traps_aa64pfr0(struct kvm_vcpu *vcpu)
 	}
 
 	/* Trap SVE */
-	if (!FIELD_GET(ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_SVE), feature_ids))
+	if (!FIELD_GET(ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_SVE), feature_ids)) {
-		cptr_set |= CPTR_EL2_TZ;
+		if (has_hvhe())
+			cptr_clear |= CPACR_EL1_ZEN_EL0EN | CPACR_EL1_ZEN_EL1EN;
+		else
+			cptr_set |= CPTR_EL2_TZ;
+	}
 
 	vcpu->arch.hcr_el2 |= hcr_set;
 	vcpu->arch.hcr_el2 &= ~hcr_clear;
 	vcpu->arch.cptr_el2 |= cptr_set;
+	vcpu->arch.cptr_el2 &= ~cptr_clear;
 }
 
 /*
@@ -120,8 +129,12 @@ static void pvm_init_traps_aa64dfr0(struct kvm_vcpu *vcpu)
 		mdcr_set |= MDCR_EL2_TTRF;
 
 	/* Trap Trace */
-	if (!FIELD_GET(ARM64_FEATURE_MASK(ID_AA64DFR0_EL1_TraceVer), feature_ids))
+	if (!FIELD_GET(ARM64_FEATURE_MASK(ID_AA64DFR0_EL1_TraceVer), feature_ids)) {
-		cptr_set |= CPTR_EL2_TTA;
+		if (has_hvhe())
+			cptr_set |= CPACR_EL1_TTA;
+		else
+			cptr_set |= CPTR_EL2_TTA;
+	}
 
 	vcpu->arch.mdcr_el2 |= mdcr_set;
 	vcpu->arch.mdcr_el2 &= ~mdcr_clear;
@@ -176,8 +189,10 @@ static void pvm_init_trap_regs(struct kvm_vcpu *vcpu)
 	/* Clear res0 and set res1 bits to trap potential new features. */
 	vcpu->arch.hcr_el2 &= ~(HCR_RES0);
 	vcpu->arch.mdcr_el2 &= ~(MDCR_EL2_RES0);
-	vcpu->arch.cptr_el2 |= CPTR_NVHE_EL2_RES1;
+	if (!has_hvhe()) {
-	vcpu->arch.cptr_el2 &= ~(CPTR_NVHE_EL2_RES0);
+		vcpu->arch.cptr_el2 |= CPTR_NVHE_EL2_RES1;
+		vcpu->arch.cptr_el2 &= ~(CPTR_NVHE_EL2_RES0);
+	}
 }
 
 /*

arch/arm64/kvm/hyp/nvhe/setup.c

@@ -11,6 +11,7 @@
 #include <asm/kvm_pkvm.h>
 
 #include <nvhe/early_alloc.h>
+#include <nvhe/ffa.h>
 #include <nvhe/fixed_config.h>
 #include <nvhe/gfp.h>
 #include <nvhe/memory.h>
@@ -28,6 +29,7 @@ static void *vmemmap_base;
 static void *vm_table_base;
 static void *hyp_pgt_base;
 static void *host_s2_pgt_base;
+static void *ffa_proxy_pages;
 static struct kvm_pgtable_mm_ops pkvm_pgtable_mm_ops;
 static struct hyp_pool hpool;
 
@@ -57,6 +59,11 @@ static int divide_memory_pool(void *virt, unsigned long size)
 	if (!host_s2_pgt_base)
 		return -ENOMEM;
 
+	nr_pages = hyp_ffa_proxy_pages();
+	ffa_proxy_pages = hyp_early_alloc_contig(nr_pages);
+	if (!ffa_proxy_pages)
+		return -ENOMEM;
+
 	return 0;
 }
 
@@ -314,6 +321,10 @@ void __noreturn __pkvm_init_finalise(void)
 	if (ret)
 		goto out;
 
+	ret = hyp_ffa_init(ffa_proxy_pages);
+	if (ret)
+		goto out;
+
 	pkvm_hyp_vm_table_init(vm_table_base);
 out:
 	/*

arch/arm64/kvm/hyp/nvhe/switch.c

@@ -44,13 +44,24 @@ static void __activate_traps(struct kvm_vcpu *vcpu)
 	__activate_traps_common(vcpu);
 
 	val = vcpu->arch.cptr_el2;
-	val |= CPTR_EL2_TTA | CPTR_EL2_TAM;
+	val |= CPTR_EL2_TAM;	/* Same bit irrespective of E2H */
+	val |= has_hvhe() ? CPACR_EL1_TTA : CPTR_EL2_TTA;
+	if (cpus_have_final_cap(ARM64_SME)) {
+		if (has_hvhe())
+			val &= ~(CPACR_EL1_SMEN_EL1EN | CPACR_EL1_SMEN_EL0EN);
+		else
+			val |= CPTR_EL2_TSM;
+	}
+
 	if (!guest_owns_fp_regs(vcpu)) {
-		val |= CPTR_EL2_TFP | CPTR_EL2_TZ;
+		if (has_hvhe())
+			val &= ~(CPACR_EL1_FPEN_EL0EN | CPACR_EL1_FPEN_EL1EN |
+				 CPACR_EL1_ZEN_EL0EN | CPACR_EL1_ZEN_EL1EN);
+		else
+			val |= CPTR_EL2_TFP | CPTR_EL2_TZ;
+
 		__activate_traps_fpsimd32(vcpu);
 	}
-	if (cpus_have_final_cap(ARM64_SME))
-		val |= CPTR_EL2_TSM;
 
 	write_sysreg(val, cptr_el2);
 	write_sysreg(__this_cpu_read(kvm_hyp_vector), vbar_el2);
@@ -73,7 +84,6 @@ static void __activate_traps(struct kvm_vcpu *vcpu)
 static void __deactivate_traps(struct kvm_vcpu *vcpu)
 {
 	extern char __kvm_hyp_host_vector[];
-	u64 cptr;
 
 	___deactivate_traps(vcpu);
 
@@ -98,13 +108,7 @@ static void __deactivate_traps(struct kvm_vcpu *vcpu)
 
 	write_sysreg(this_cpu_ptr(&kvm_init_params)->hcr_el2, hcr_el2);
 
-	cptr = CPTR_EL2_DEFAULT;
+	kvm_reset_cptr_el2(vcpu);
-	if (vcpu_has_sve(vcpu) && (vcpu->arch.fp_state == FP_STATE_GUEST_OWNED))
-		cptr |= CPTR_EL2_TZ;
-	if (cpus_have_final_cap(ARM64_SME))
-		cptr &= ~CPTR_EL2_TSM;
-
-	write_sysreg(cptr, cptr_el2);
 	write_sysreg(__kvm_hyp_host_vector, vbar_el2);
 }
 

arch/arm64/kvm/hyp/nvhe/timer-sr.c

@@ -17,21 +17,24 @@ void __kvm_timer_set_cntvoff(u64 cntvoff)
 }
 
 /*
- * Should only be called on non-VHE systems.
+ * Should only be called on non-VHE or hVHE setups.
  * VHE systems use EL2 timers and configure EL1 timers in kvm_timer_init_vhe().
  */
 void __timer_disable_traps(struct kvm_vcpu *vcpu)
 {
-	u64 val;
+	u64 val, shift = 0;
+
+	if (has_hvhe())
+		shift = 10;
 
 	/* Allow physical timer/counter access for the host */
 	val = read_sysreg(cnthctl_el2);
-	val |= CNTHCTL_EL1PCTEN | CNTHCTL_EL1PCEN;
+	val |= (CNTHCTL_EL1PCTEN | CNTHCTL_EL1PCEN) << shift;
 	write_sysreg(val, cnthctl_el2);
 }
 
 /*
- * Should only be called on non-VHE systems.
+ * Should only be called on non-VHE or hVHE setups.
  * VHE systems use EL2 timers and configure EL1 timers in kvm_timer_init_vhe().
  */
 void __timer_enable_traps(struct kvm_vcpu *vcpu)
@@ -50,5 +53,10 @@ void __timer_enable_traps(struct kvm_vcpu *vcpu)
 	else
 		clr |= CNTHCTL_EL1PCTEN;
 
+	if (has_hvhe()) {
+		clr <<= 10;
+		set <<= 10;
+	}
+
 	sysreg_clear_set(cnthctl_el2, clr, set);
 }

arch/arm64/kvm/hyp/nvhe/tlb.c

@@ -130,6 +130,58 @@ void __kvm_tlb_flush_vmid_ipa(struct kvm_s2_mmu *mmu,
 	__tlb_switch_to_host(&cxt);
 }
 
+void __kvm_tlb_flush_vmid_ipa_nsh(struct kvm_s2_mmu *mmu,
+				  phys_addr_t ipa, int level)
+{
+	struct tlb_inv_context cxt;
+
+	/* Switch to requested VMID */
+	__tlb_switch_to_guest(mmu, &cxt, true);
+
+	/*
+	 * We could do so much better if we had the VA as well.
+	 * Instead, we invalidate Stage-2 for this IPA, and the
+	 * whole of Stage-1. Weep...
+	 */
+	ipa >>= 12;
+	__tlbi_level(ipas2e1, ipa, level);
+
+	/*
+	 * We have to ensure completion of the invalidation at Stage-2,
+	 * since a table walk on another CPU could refill a TLB with a
+	 * complete (S1 + S2) walk based on the old Stage-2 mapping if
+	 * the Stage-1 invalidation happened first.
+	 */
+	dsb(nsh);
+	__tlbi(vmalle1);
+	dsb(nsh);
+	isb();
+
+	/*
+	 * If the host is running at EL1 and we have a VPIPT I-cache,
+	 * then we must perform I-cache maintenance at EL2 in order for
+	 * it to have an effect on the guest. Since the guest cannot hit
+	 * I-cache lines allocated with a different VMID, we don't need
+	 * to worry about junk out of guest reset (we nuke the I-cache on
+	 * VMID rollover), but we do need to be careful when remapping
+	 * executable pages for the same guest. This can happen when KSM
+	 * takes a CoW fault on an executable page, copies the page into
+	 * a page that was previously mapped in the guest and then needs
+	 * to invalidate the guest view of the I-cache for that page
+	 * from EL1. To solve this, we invalidate the entire I-cache when
+	 * unmapping a page from a guest if we have a VPIPT I-cache but
+	 * the host is running at EL1. As above, we could do better if
+	 * we had the VA.
+	 *
+	 * The moral of this story is: if you have a VPIPT I-cache, then
+	 * you should be running with VHE enabled.
+	 */
+	if (icache_is_vpipt())
+		icache_inval_all_pou();
+
+	__tlb_switch_to_host(&cxt);
+}
+
 void __kvm_tlb_flush_vmid(struct kvm_s2_mmu *mmu)
 {
 	struct tlb_inv_context cxt;

arch/arm64/kvm/hyp/pgtable.c

@@ -21,8 +21,10 @@
 
 #define KVM_PTE_LEAF_ATTR_LO_S1_ATTRIDX	GENMASK(4, 2)
 #define KVM_PTE_LEAF_ATTR_LO_S1_AP	GENMASK(7, 6)
-#define KVM_PTE_LEAF_ATTR_LO_S1_AP_RO	3
+#define KVM_PTE_LEAF_ATTR_LO_S1_AP_RO		\
-#define KVM_PTE_LEAF_ATTR_LO_S1_AP_RW	1
+	({ cpus_have_final_cap(ARM64_KVM_HVHE) ? 2 : 3; })
+#define KVM_PTE_LEAF_ATTR_LO_S1_AP_RW		\
+	({ cpus_have_final_cap(ARM64_KVM_HVHE) ? 0 : 1; })
 #define KVM_PTE_LEAF_ATTR_LO_S1_SH	GENMASK(9, 8)
 #define KVM_PTE_LEAF_ATTR_LO_S1_SH_IS	3
 #define KVM_PTE_LEAF_ATTR_LO_S1_AF	BIT(10)
@@ -34,7 +36,7 @@
 #define KVM_PTE_LEAF_ATTR_LO_S2_SH_IS	3
 #define KVM_PTE_LEAF_ATTR_LO_S2_AF	BIT(10)
 
-#define KVM_PTE_LEAF_ATTR_HI		GENMASK(63, 51)
+#define KVM_PTE_LEAF_ATTR_HI		GENMASK(63, 50)
 
 #define KVM_PTE_LEAF_ATTR_HI_SW		GENMASK(58, 55)
 
@@ -42,6 +44,8 @@
 
 #define KVM_PTE_LEAF_ATTR_HI_S2_XN	BIT(54)
 
+#define KVM_PTE_LEAF_ATTR_HI_S1_GP	BIT(50)
+
 #define KVM_PTE_LEAF_ATTR_S2_PERMS	(KVM_PTE_LEAF_ATTR_LO_S2_S2AP_R | \
 					 KVM_PTE_LEAF_ATTR_LO_S2_S2AP_W | \
 					 KVM_PTE_LEAF_ATTR_HI_S2_XN)
@@ -63,6 +67,16 @@ struct kvm_pgtable_walk_data {
 	const u64			end;
 };
 
+static bool kvm_pgtable_walk_skip_bbm_tlbi(const struct kvm_pgtable_visit_ctx *ctx)
+{
+	return unlikely(ctx->flags & KVM_PGTABLE_WALK_SKIP_BBM_TLBI);
+}
+
+static bool kvm_pgtable_walk_skip_cmo(const struct kvm_pgtable_visit_ctx *ctx)
+{
+	return unlikely(ctx->flags & KVM_PGTABLE_WALK_SKIP_CMO);
+}
+
 static bool kvm_phys_is_valid(u64 phys)
 {
 	return phys < BIT(id_aa64mmfr0_parange_to_phys_shift(ID_AA64MMFR0_EL1_PARANGE_MAX));
@@ -386,6 +400,9 @@ static int hyp_set_prot_attr(enum kvm_pgtable_prot prot, kvm_pte_t *ptep)
 
 		if (device)
 			return -EINVAL;
+
+		if (IS_ENABLED(CONFIG_ARM64_BTI_KERNEL) && system_supports_bti())
+			attr |= KVM_PTE_LEAF_ATTR_HI_S1_GP;
 	} else {
 		attr |= KVM_PTE_LEAF_ATTR_HI_S1_XN;
 	}
@@ -623,10 +640,18 @@ u64 kvm_get_vtcr(u64 mmfr0, u64 mmfr1, u32 phys_shift)
 #ifdef CONFIG_ARM64_HW_AFDBM
 	/*
 	 * Enable the Hardware Access Flag management, unconditionally
-	 * on all CPUs. The features is RES0 on CPUs without the support
+	 * on all CPUs. In systems that have asymmetric support for the feature
-	 * and must be ignored by the CPUs.
+	 * this allows KVM to leverage hardware support on the subset of cores
+	 * that implement the feature.
+	 *
+	 * The architecture requires VTCR_EL2.HA to be RES0 (thus ignored by
+	 * hardware) on implementations that do not advertise support for the
+	 * feature. As such, setting HA unconditionally is safe, unless you
+	 * happen to be running on a design that has unadvertised support for
+	 * HAFDBS. Here be dragons.
 	 */
-	vtcr |= VTCR_EL2_HA;
+	if (!cpus_have_final_cap(ARM64_WORKAROUND_AMPERE_AC03_CPU_38))
+		vtcr |= VTCR_EL2_HA;
 #endif /* CONFIG_ARM64_HW_AFDBM */
 
 	/* Set the vmid bits */
@@ -755,14 +780,17 @@ static bool stage2_try_break_pte(const struct kvm_pgtable_visit_ctx *ctx,
 	if (!stage2_try_set_pte(ctx, KVM_INVALID_PTE_LOCKED))
 		return false;
 
-	/*
+	if (!kvm_pgtable_walk_skip_bbm_tlbi(ctx)) {
-	 * Perform the appropriate TLB invalidation based on the evicted pte
+		/*
-	 * value (if any).
+		 * Perform the appropriate TLB invalidation based on the
-	 */
+		 * evicted pte value (if any).
-	if (kvm_pte_table(ctx->old, ctx->level))
+		 */
-		kvm_call_hyp(__kvm_tlb_flush_vmid, mmu);
+		if (kvm_pte_table(ctx->old, ctx->level))
-	else if (kvm_pte_valid(ctx->old))
+			kvm_call_hyp(__kvm_tlb_flush_vmid, mmu);
-		kvm_call_hyp(__kvm_tlb_flush_vmid_ipa, mmu, ctx->addr, ctx->level);
+		else if (kvm_pte_valid(ctx->old))
+			kvm_call_hyp(__kvm_tlb_flush_vmid_ipa, mmu,
+				     ctx->addr, ctx->level);
+	}
 
 	if (stage2_pte_is_counted(ctx->old))
 		mm_ops->put_page(ctx->ptep);
@@ -869,11 +897,13 @@ static int stage2_map_walker_try_leaf(const struct kvm_pgtable_visit_ctx *ctx,
 		return -EAGAIN;
 
 	/* Perform CMOs before installation of the guest stage-2 PTE */
-	if (mm_ops->dcache_clean_inval_poc && stage2_pte_cacheable(pgt, new))
+	if (!kvm_pgtable_walk_skip_cmo(ctx) && mm_ops->dcache_clean_inval_poc &&
+	    stage2_pte_cacheable(pgt, new))
 		mm_ops->dcache_clean_inval_poc(kvm_pte_follow(new, mm_ops),
-						granule);
+					       granule);
 
-	if (mm_ops->icache_inval_pou && stage2_pte_executable(new))
+	if (!kvm_pgtable_walk_skip_cmo(ctx) && mm_ops->icache_inval_pou &&
+	    stage2_pte_executable(new))
 		mm_ops->icache_inval_pou(kvm_pte_follow(new, mm_ops), granule);
 
 	stage2_make_pte(ctx, new);
@@ -895,7 +925,7 @@ static int stage2_map_walk_table_pre(const struct kvm_pgtable_visit_ctx *ctx,
 	if (ret)
 		return ret;
 
-	mm_ops->free_removed_table(childp, ctx->level);
+	mm_ops->free_unlinked_table(childp, ctx->level);
 	return 0;
 }
 
@@ -940,7 +970,7 @@ static int stage2_map_walk_leaf(const struct kvm_pgtable_visit_ctx *ctx,
  * The TABLE_PRE callback runs for table entries on the way down, looking
  * for table entries which we could conceivably replace with a block entry
  * for this mapping. If it finds one it replaces the entry and calls
- * kvm_pgtable_mm_ops::free_removed_table() to tear down the detached table.
+ * kvm_pgtable_mm_ops::free_unlinked_table() to tear down the detached table.
  *
  * Otherwise, the LEAF callback performs the mapping at the existing leaves
  * instead.
@@ -1209,7 +1239,7 @@ int kvm_pgtable_stage2_relax_perms(struct kvm_pgtable *pgt, u64 addr,
 				       KVM_PGTABLE_WALK_HANDLE_FAULT |
 				       KVM_PGTABLE_WALK_SHARED);
 	if (!ret)
-		kvm_call_hyp(__kvm_tlb_flush_vmid_ipa, pgt->mmu, addr, level);
+		kvm_call_hyp(__kvm_tlb_flush_vmid_ipa_nsh, pgt->mmu, addr, level);
 	return ret;
 }
 
@@ -1242,6 +1272,162 @@ int kvm_pgtable_stage2_flush(struct kvm_pgtable *pgt, u64 addr, u64 size)
 	return kvm_pgtable_walk(pgt, addr, size, &walker);
 }
 
+kvm_pte_t *kvm_pgtable_stage2_create_unlinked(struct kvm_pgtable *pgt,
+					      u64 phys, u32 level,
+					      enum kvm_pgtable_prot prot,
+					      void *mc, bool force_pte)
+{
+	struct stage2_map_data map_data = {
+		.phys		= phys,
+		.mmu		= pgt->mmu,
+		.memcache	= mc,
+		.force_pte	= force_pte,
+	};
+	struct kvm_pgtable_walker walker = {
+		.cb		= stage2_map_walker,
+		.flags		= KVM_PGTABLE_WALK_LEAF |
+				  KVM_PGTABLE_WALK_SKIP_BBM_TLBI |
+				  KVM_PGTABLE_WALK_SKIP_CMO,
+		.arg		= &map_data,
+	};
+	/*
+	 * The input address (.addr) is irrelevant for walking an
+	 * unlinked table. Construct an ambiguous IA range to map
+	 * kvm_granule_size(level) worth of memory.
+	 */
+	struct kvm_pgtable_walk_data data = {
+		.walker	= &walker,
+		.addr	= 0,
+		.end	= kvm_granule_size(level),
+	};
+	struct kvm_pgtable_mm_ops *mm_ops = pgt->mm_ops;
+	kvm_pte_t *pgtable;
+	int ret;
+
+	if (!IS_ALIGNED(phys, kvm_granule_size(level)))
+		return ERR_PTR(-EINVAL);
+
+	ret = stage2_set_prot_attr(pgt, prot, &map_data.attr);
+	if (ret)
+		return ERR_PTR(ret);
+
+	pgtable = mm_ops->zalloc_page(mc);
+	if (!pgtable)
+		return ERR_PTR(-ENOMEM);
+
+	ret = __kvm_pgtable_walk(&data, mm_ops, (kvm_pteref_t)pgtable,
+				 level + 1);
+	if (ret) {
+		kvm_pgtable_stage2_free_unlinked(mm_ops, pgtable, level);
+		mm_ops->put_page(pgtable);
+		return ERR_PTR(ret);
+	}
+
+	return pgtable;
+}
+
+/*
+ * Get the number of page-tables needed to replace a block with a
+ * fully populated tree up to the PTE entries. Note that @level is
+ * interpreted as in "level @level entry".
+ */
+static int stage2_block_get_nr_page_tables(u32 level)
+{
+	switch (level) {
+	case 1:
+		return PTRS_PER_PTE + 1;
+	case 2:
+		return 1;
+	case 3:
+		return 0;
+	default:
+		WARN_ON_ONCE(level < KVM_PGTABLE_MIN_BLOCK_LEVEL ||
+			     level >= KVM_PGTABLE_MAX_LEVELS);
+		return -EINVAL;
+	};
+}
+
+static int stage2_split_walker(const struct kvm_pgtable_visit_ctx *ctx,
+			       enum kvm_pgtable_walk_flags visit)
+{
+	struct kvm_pgtable_mm_ops *mm_ops = ctx->mm_ops;
+	struct kvm_mmu_memory_cache *mc = ctx->arg;
+	struct kvm_s2_mmu *mmu;
+	kvm_pte_t pte = ctx->old, new, *childp;
+	enum kvm_pgtable_prot prot;
+	u32 level = ctx->level;
+	bool force_pte;
+	int nr_pages;
+	u64 phys;
+
+	/* No huge-pages exist at the last level */
+	if (level == KVM_PGTABLE_MAX_LEVELS - 1)
+		return 0;
+
+	/* We only split valid block mappings */
+	if (!kvm_pte_valid(pte))
+		return 0;
+
+	nr_pages = stage2_block_get_nr_page_tables(level);
+	if (nr_pages < 0)
+		return nr_pages;
+
+	if (mc->nobjs >= nr_pages) {
+		/* Build a tree mapped down to the PTE granularity. */
+		force_pte = true;
+	} else {
+		/*
+		 * Don't force PTEs, so create_unlinked() below does
+		 * not populate the tree up to the PTE level. The
+		 * consequence is that the call will require a single
+		 * page of level 2 entries at level 1, or a single
+		 * page of PTEs at level 2. If we are at level 1, the
+		 * PTEs will be created recursively.
+		 */
+		force_pte = false;
+		nr_pages = 1;
+	}
+
+	if (mc->nobjs < nr_pages)
+		return -ENOMEM;
+
+	mmu = container_of(mc, struct kvm_s2_mmu, split_page_cache);
+	phys = kvm_pte_to_phys(pte);
+	prot = kvm_pgtable_stage2_pte_prot(pte);
+
+	childp = kvm_pgtable_stage2_create_unlinked(mmu->pgt, phys,
+						    level, prot, mc, force_pte);
+	if (IS_ERR(childp))
+		return PTR_ERR(childp);
+
+	if (!stage2_try_break_pte(ctx, mmu)) {
+		kvm_pgtable_stage2_free_unlinked(mm_ops, childp, level);
+		mm_ops->put_page(childp);
+		return -EAGAIN;
+	}
+
+	/*
+	 * Note, the contents of the page table are guaranteed to be made
+	 * visible before the new PTE is assigned because stage2_make_pte()
+	 * writes the PTE using smp_store_release().
+	 */
+	new = kvm_init_table_pte(childp, mm_ops);
+	stage2_make_pte(ctx, new);
+	dsb(ishst);
+	return 0;
+}
+
+int kvm_pgtable_stage2_split(struct kvm_pgtable *pgt, u64 addr, u64 size,
+			     struct kvm_mmu_memory_cache *mc)
+{
+	struct kvm_pgtable_walker walker = {
+		.cb	= stage2_split_walker,
+		.flags	= KVM_PGTABLE_WALK_LEAF,
+		.arg	= mc,
+	};
+
+	return kvm_pgtable_walk(pgt, addr, size, &walker);
+}
 
 int __kvm_pgtable_stage2_init(struct kvm_pgtable *pgt, struct kvm_s2_mmu *mmu,
 			      struct kvm_pgtable_mm_ops *mm_ops,
@@ -1311,7 +1497,7 @@ void kvm_pgtable_stage2_destroy(struct kvm_pgtable *pgt)
 	pgt->pgd = NULL;
 }
 
-void kvm_pgtable_stage2_free_removed(struct kvm_pgtable_mm_ops *mm_ops, void *pgtable, u32 level)
+void kvm_pgtable_stage2_free_unlinked(struct kvm_pgtable_mm_ops *mm_ops, void *pgtable, u32 level)
 {
 	kvm_pteref_t ptep = (kvm_pteref_t)pgtable;
 	struct kvm_pgtable_walker walker = {

arch/arm64/kvm/hyp/vhe/switch.c

@@ -84,7 +84,7 @@ static void __deactivate_traps(struct kvm_vcpu *vcpu)
 	 */
 	asm(ALTERNATIVE("nop", "isb", ARM64_WORKAROUND_SPECULATIVE_AT));
 
-	write_sysreg(CPACR_EL1_DEFAULT, cpacr_el1);
+	kvm_reset_cptr_el2(vcpu);
 
 	if (!arm64_kernel_unmapped_at_el0())
 		host_vectors = __this_cpu_read(this_cpu_vector);

arch/arm64/kvm/hyp/vhe/tlb.c

@@ -111,6 +111,38 @@ void __kvm_tlb_flush_vmid_ipa(struct kvm_s2_mmu *mmu,
 	__tlb_switch_to_host(&cxt);
 }
 
+void __kvm_tlb_flush_vmid_ipa_nsh(struct kvm_s2_mmu *mmu,
+				  phys_addr_t ipa, int level)
+{
+	struct tlb_inv_context cxt;
+
+	dsb(nshst);
+
+	/* Switch to requested VMID */
+	__tlb_switch_to_guest(mmu, &cxt);
+
+	/*
+	 * We could do so much better if we had the VA as well.
+	 * Instead, we invalidate Stage-2 for this IPA, and the
+	 * whole of Stage-1. Weep...
+	 */
+	ipa >>= 12;
+	__tlbi_level(ipas2e1, ipa, level);
+
+	/*
+	 * We have to ensure completion of the invalidation at Stage-2,
+	 * since a table walk on another CPU could refill a TLB with a
+	 * complete (S1 + S2) walk based on the old Stage-2 mapping if
+	 * the Stage-1 invalidation happened first.
+	 */
+	dsb(nsh);
+	__tlbi(vmalle1);
+	dsb(nsh);
+	isb();
+
+	__tlb_switch_to_host(&cxt);
+}
+
 void __kvm_tlb_flush_vmid(struct kvm_s2_mmu *mmu)
 {
 	struct tlb_inv_context cxt;

arch/arm64/kvm/mmu.c

@@ -31,14 +31,21 @@ static phys_addr_t __ro_after_init hyp_idmap_vector;
 
 static unsigned long __ro_after_init io_map_base;
 
-static phys_addr_t stage2_range_addr_end(phys_addr_t addr, phys_addr_t end)
+static phys_addr_t __stage2_range_addr_end(phys_addr_t addr, phys_addr_t end,
+					   phys_addr_t size)
 {
-	phys_addr_t size = kvm_granule_size(KVM_PGTABLE_MIN_BLOCK_LEVEL);
 	phys_addr_t boundary = ALIGN_DOWN(addr + size, size);
 
 	return (boundary - 1 < end - 1) ? boundary : end;
 }
 
+static phys_addr_t stage2_range_addr_end(phys_addr_t addr, phys_addr_t end)
+{
+	phys_addr_t size = kvm_granule_size(KVM_PGTABLE_MIN_BLOCK_LEVEL);
+
+	return __stage2_range_addr_end(addr, end, size);
+}
+
 /*
  * Release kvm_mmu_lock periodically if the memory region is large. Otherwise,
  * we may see kernel panics with CONFIG_DETECT_HUNG_TASK,
@@ -75,6 +82,79 @@ static int stage2_apply_range(struct kvm_s2_mmu *mmu, phys_addr_t addr,
 #define stage2_apply_range_resched(mmu, addr, end, fn)			\
 	stage2_apply_range(mmu, addr, end, fn, true)
 
+/*
+ * Get the maximum number of page-tables pages needed to split a range
+ * of blocks into PAGE_SIZE PTEs. It assumes the range is already
+ * mapped at level 2, or at level 1 if allowed.
+ */
+static int kvm_mmu_split_nr_page_tables(u64 range)
+{
+	int n = 0;
+
+	if (KVM_PGTABLE_MIN_BLOCK_LEVEL < 2)
+		n += DIV_ROUND_UP(range, PUD_SIZE);
+	n += DIV_ROUND_UP(range, PMD_SIZE);
+	return n;
+}
+
+static bool need_split_memcache_topup_or_resched(struct kvm *kvm)
+{
+	struct kvm_mmu_memory_cache *cache;
+	u64 chunk_size, min;
+
+	if (need_resched() || rwlock_needbreak(&kvm->mmu_lock))
+		return true;
+
+	chunk_size = kvm->arch.mmu.split_page_chunk_size;
+	min = kvm_mmu_split_nr_page_tables(chunk_size);
+	cache = &kvm->arch.mmu.split_page_cache;
+	return kvm_mmu_memory_cache_nr_free_objects(cache) < min;
+}
+
+static int kvm_mmu_split_huge_pages(struct kvm *kvm, phys_addr_t addr,
+				    phys_addr_t end)
+{
+	struct kvm_mmu_memory_cache *cache;
+	struct kvm_pgtable *pgt;
+	int ret, cache_capacity;
+	u64 next, chunk_size;
+
+	lockdep_assert_held_write(&kvm->mmu_lock);
+
+	chunk_size = kvm->arch.mmu.split_page_chunk_size;
+	cache_capacity = kvm_mmu_split_nr_page_tables(chunk_size);
+
+	if (chunk_size == 0)
+		return 0;
+
+	cache = &kvm->arch.mmu.split_page_cache;
+
+	do {
+		if (need_split_memcache_topup_or_resched(kvm)) {
+			write_unlock(&kvm->mmu_lock);
+			cond_resched();
+			/* Eager page splitting is best-effort. */
+			ret = __kvm_mmu_topup_memory_cache(cache,
+							   cache_capacity,
+							   cache_capacity);
+			write_lock(&kvm->mmu_lock);
+			if (ret)
+				break;
+		}
+
+		pgt = kvm->arch.mmu.pgt;
+		if (!pgt)
+			return -EINVAL;
+
+		next = __stage2_range_addr_end(addr, end, chunk_size);
+		ret = kvm_pgtable_stage2_split(pgt, addr, next - addr, cache);
+		if (ret)
+			break;
+	} while (addr = next, addr != end);
+
+	return ret;
+}
+
 static bool memslot_is_logging(struct kvm_memory_slot *memslot)
 {
 	return memslot->dirty_bitmap && !(memslot->flags & KVM_MEM_READONLY);
@@ -131,21 +211,21 @@ static void kvm_s2_free_pages_exact(void *virt, size_t size)
 
 static struct kvm_pgtable_mm_ops kvm_s2_mm_ops;
 
-static void stage2_free_removed_table_rcu_cb(struct rcu_head *head)
+static void stage2_free_unlinked_table_rcu_cb(struct rcu_head *head)
 {
 	struct page *page = container_of(head, struct page, rcu_head);
 	void *pgtable = page_to_virt(page);
 	u32 level = page_private(page);
 
-	kvm_pgtable_stage2_free_removed(&kvm_s2_mm_ops, pgtable, level);
+	kvm_pgtable_stage2_free_unlinked(&kvm_s2_mm_ops, pgtable, level);
 }
 
-static void stage2_free_removed_table(void *addr, u32 level)
+static void stage2_free_unlinked_table(void *addr, u32 level)
 {
 	struct page *page = virt_to_page(addr);
 
 	set_page_private(page, (unsigned long)level);
-	call_rcu(&page->rcu_head, stage2_free_removed_table_rcu_cb);
+	call_rcu(&page->rcu_head, stage2_free_unlinked_table_rcu_cb);
 }
 
 static void kvm_host_get_page(void *addr)
@@ -701,7 +781,7 @@ static struct kvm_pgtable_mm_ops kvm_s2_mm_ops = {
 	.zalloc_page		= stage2_memcache_zalloc_page,
 	.zalloc_pages_exact	= kvm_s2_zalloc_pages_exact,
 	.free_pages_exact	= kvm_s2_free_pages_exact,
-	.free_removed_table	= stage2_free_removed_table,
+	.free_unlinked_table	= stage2_free_unlinked_table,
 	.get_page		= kvm_host_get_page,
 	.put_page		= kvm_s2_put_page,
 	.page_count		= kvm_host_page_count,
@@ -775,6 +855,10 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long t
 	for_each_possible_cpu(cpu)
 		*per_cpu_ptr(mmu->last_vcpu_ran, cpu) = -1;
 
+	 /* The eager page splitting is disabled by default */
+	mmu->split_page_chunk_size = KVM_ARM_EAGER_SPLIT_CHUNK_SIZE_DEFAULT;
+	mmu->split_page_cache.gfp_zero = __GFP_ZERO;
+
 	mmu->pgt = pgt;
 	mmu->pgd_phys = __pa(pgt->pgd);
 	return 0;
@@ -786,6 +870,12 @@ out_free_pgtable:
 	return err;
 }
 
+void kvm_uninit_stage2_mmu(struct kvm *kvm)
+{
+	kvm_free_stage2_pgd(&kvm->arch.mmu);
+	kvm_mmu_free_memory_cache(&kvm->arch.mmu.split_page_cache);
+}
+
 static void stage2_unmap_memslot(struct kvm *kvm,
 				 struct kvm_memory_slot *memslot)
 {
@@ -989,17 +1079,45 @@ static void kvm_mmu_wp_memory_region(struct kvm *kvm, int slot)
 }
 
 /**
- * kvm_mmu_write_protect_pt_masked() - write protect dirty pages
+ * kvm_mmu_split_memory_region() - split the stage 2 blocks into PAGE_SIZE
+ *				   pages for memory slot
+ * @kvm:	The KVM pointer
+ * @slot:	The memory slot to split
+ *
+ * Acquires kvm->mmu_lock. Called with kvm->slots_lock mutex acquired,
+ * serializing operations for VM memory regions.
+ */
+static void kvm_mmu_split_memory_region(struct kvm *kvm, int slot)
+{
+	struct kvm_memslots *slots;
+	struct kvm_memory_slot *memslot;
+	phys_addr_t start, end;
+
+	lockdep_assert_held(&kvm->slots_lock);
+
+	slots = kvm_memslots(kvm);
+	memslot = id_to_memslot(slots, slot);
+
+	start = memslot->base_gfn << PAGE_SHIFT;
+	end = (memslot->base_gfn + memslot->npages) << PAGE_SHIFT;
+
+	write_lock(&kvm->mmu_lock);
+	kvm_mmu_split_huge_pages(kvm, start, end);
+	write_unlock(&kvm->mmu_lock);
+}
+
+/*
+ * kvm_arch_mmu_enable_log_dirty_pt_masked() - enable dirty logging for selected pages.
  * @kvm:	The KVM pointer
  * @slot:	The memory slot associated with mask
  * @gfn_offset:	The gfn offset in memory slot
- * @mask:	The mask of dirty pages at offset 'gfn_offset' in this memory
+ * @mask:	The mask of pages at offset 'gfn_offset' in this memory
- *		slot to be write protected
+ *		slot to enable dirty logging on
  *
- * Walks bits set in mask write protects the associated pte's. Caller must
+ * Writes protect selected pages to enable dirty logging, and then
- * acquire kvm_mmu_lock.
+ * splits them to PAGE_SIZE. Caller must acquire kvm->mmu_lock.
  */
-static void kvm_mmu_write_protect_pt_masked(struct kvm *kvm,
+void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
 		struct kvm_memory_slot *slot,
 		gfn_t gfn_offset, unsigned long mask)
 {
@@ -1007,21 +1125,20 @@ static void kvm_mmu_write_protect_pt_masked(struct kvm *kvm,
 	phys_addr_t start = (base_gfn +  __ffs(mask)) << PAGE_SHIFT;
 	phys_addr_t end = (base_gfn + __fls(mask) + 1) << PAGE_SHIFT;
 
-	stage2_wp_range(&kvm->arch.mmu, start, end);
+	lockdep_assert_held_write(&kvm->mmu_lock);
-}
 
-/*
+	stage2_wp_range(&kvm->arch.mmu, start, end);
- * kvm_arch_mmu_enable_log_dirty_pt_masked - enable dirty logging for selected
+
- * dirty pages.
+	/*
- *
+	 * Eager-splitting is done when manual-protect is set.  We
- * It calls kvm_mmu_write_protect_pt_masked to write protect selected pages to
+	 * also check for initially-all-set because we can avoid
- * enable dirty logging for them.
+	 * eager-splitting if initially-all-set is false.
- */
+	 * Initially-all-set equal false implies that huge-pages were
-void kvm_arch_mmu_enable_log_dirty_pt_masked(struct kvm *kvm,
+	 * already split when enabling dirty logging: no need to do it
-		struct kvm_memory_slot *slot,
+	 * again.
-		gfn_t gfn_offset, unsigned long mask)
+	 */
-{
+	if (kvm_dirty_log_manual_protect_and_init_set(kvm))
-	kvm_mmu_write_protect_pt_masked(kvm, slot, gfn_offset, mask);
+		kvm_mmu_split_huge_pages(kvm, start, end);
 }
 
 static void kvm_send_hwpoison_signal(unsigned long address, short lsb)
@@ -1790,20 +1907,42 @@ void kvm_arch_commit_memory_region(struct kvm *kvm,
 				   const struct kvm_memory_slot *new,
 				   enum kvm_mr_change change)
 {
+	bool log_dirty_pages = new && new->flags & KVM_MEM_LOG_DIRTY_PAGES;
+
 	/*
 	 * At this point memslot has been committed and there is an
 	 * allocated dirty_bitmap[], dirty pages will be tracked while the
 	 * memory slot is write protected.
 	 */
-	if (change != KVM_MR_DELETE && new->flags & KVM_MEM_LOG_DIRTY_PAGES) {
+	if (log_dirty_pages) {
+
+		if (change == KVM_MR_DELETE)
+			return;
+
 		/*
-		 * If we're with initial-all-set, we don't need to write
+		 * Huge and normal pages are write-protected and split
-		 * protect any pages because they're all reported as dirty.
+		 * on either of these two cases:
-		 * Huge pages and normal pages will be write protect gradually.
+		 *
+		 * 1. with initial-all-set: gradually with CLEAR ioctls,
 		 */
-		if (!kvm_dirty_log_manual_protect_and_init_set(kvm)) {
+		if (kvm_dirty_log_manual_protect_and_init_set(kvm))
-			kvm_mmu_wp_memory_region(kvm, new->id);
+			return;
-		}
+		/*
+		 * or
+		 * 2. without initial-all-set: all in one shot when
+		 *    enabling dirty logging.
+		 */
+		kvm_mmu_wp_memory_region(kvm, new->id);
+		kvm_mmu_split_memory_region(kvm, new->id);
+	} else {
+		/*
+		 * Free any leftovers from the eager page splitting cache. Do
+		 * this when deleting, moving, disabling dirty logging, or
+		 * creating the memslot (a nop). Doing it for deletes makes
+		 * sure we don't leak memory, and there's no need to keep the
+		 * cache around for any of the other cases.
+		 */
+		kvm_mmu_free_memory_cache(&kvm->arch.mmu.split_page_cache);
 	}
 }
 
@@ -1877,7 +2016,7 @@ void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen)
 
 void kvm_arch_flush_shadow_all(struct kvm *kvm)
 {
-	kvm_free_stage2_pgd(&kvm->arch.mmu);
+	kvm_uninit_stage2_mmu(kvm);
 }
 
 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,

arch/arm64/kvm/pkvm.c

@@ -78,6 +78,7 @@ void __init kvm_hyp_reserve(void)
 	hyp_mem_pages += host_s2_pgtable_pages();
 	hyp_mem_pages += hyp_vm_table_pages();
 	hyp_mem_pages += hyp_vmemmap_pages(STRUCT_HYP_PAGE_SIZE);
+	hyp_mem_pages += hyp_ffa_proxy_pages();
 
 	/*
 	 * Try to allocate a PMD-aligned region to reduce TLB pressure once

arch/arm64/kvm/reset.c

@@ -186,57 +186,6 @@ static int kvm_vcpu_enable_ptrauth(struct kvm_vcpu *vcpu)
 	return 0;
 }
 
-/**
- * kvm_set_vm_width() - set the register width for the guest
- * @vcpu: Pointer to the vcpu being configured
- *
- * Set both KVM_ARCH_FLAG_EL1_32BIT and KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED
- * in the VM flags based on the vcpu's requested register width, the HW
- * capabilities and other options (such as MTE).
- * When REG_WIDTH_CONFIGURED is already set, the vcpu settings must be
- * consistent with the value of the FLAG_EL1_32BIT bit in the flags.
- *
- * Return: 0 on success, negative error code on failure.
- */
-static int kvm_set_vm_width(struct kvm_vcpu *vcpu)
-{
-	struct kvm *kvm = vcpu->kvm;
-	bool is32bit;
-
-	is32bit = vcpu_has_feature(vcpu, KVM_ARM_VCPU_EL1_32BIT);
-
-	lockdep_assert_held(&kvm->arch.config_lock);
-
-	if (test_bit(KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED, &kvm->arch.flags)) {
-		/*
-		 * The guest's register width is already configured.
-		 * Make sure that the vcpu is consistent with it.
-		 */
-		if (is32bit == test_bit(KVM_ARCH_FLAG_EL1_32BIT, &kvm->arch.flags))
-			return 0;
-
-		return -EINVAL;
-	}
-
-	if (!cpus_have_const_cap(ARM64_HAS_32BIT_EL1) && is32bit)
-		return -EINVAL;
-
-	/* MTE is incompatible with AArch32 */
-	if (kvm_has_mte(kvm) && is32bit)
-		return -EINVAL;
-
-	/* NV is incompatible with AArch32 */
-	if (vcpu_has_nv(vcpu) && is32bit)
-		return -EINVAL;
-
-	if (is32bit)
-		set_bit(KVM_ARCH_FLAG_EL1_32BIT, &kvm->arch.flags);
-
-	set_bit(KVM_ARCH_FLAG_REG_WIDTH_CONFIGURED, &kvm->arch.flags);
-
-	return 0;
-}
-
 /**
  * kvm_reset_vcpu - sets core registers and sys_regs to reset value
  * @vcpu: The VCPU pointer
@@ -262,13 +211,6 @@ int kvm_reset_vcpu(struct kvm_vcpu *vcpu)
 	bool loaded;
 	u32 pstate;
 
-	mutex_lock(&vcpu->kvm->arch.config_lock);
-	ret = kvm_set_vm_width(vcpu);
-	mutex_unlock(&vcpu->kvm->arch.config_lock);
-
-	if (ret)
-		return ret;
-
 	spin_lock(&vcpu->arch.mp_state_lock);
 	reset_state = vcpu->arch.reset_state;
 	vcpu->arch.reset_state.reset = false;

arch/arm64/kvm/sys_regs.c

@@ -42,6 +42,8 @@
  */
 
 static u64 sys_reg_to_index(const struct sys_reg_desc *reg);
+static int set_id_reg(struct kvm_vcpu *vcpu, const struct sys_reg_desc *rd,
+		      u64 val);
 
 static bool read_from_write_only(struct kvm_vcpu *vcpu,
 				 struct sys_reg_params *params,
@@ -553,10 +555,11 @@ static int get_bvr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *rd,
 	return 0;
 }
 
-static void reset_bvr(struct kvm_vcpu *vcpu,
+static u64 reset_bvr(struct kvm_vcpu *vcpu,
 		      const struct sys_reg_desc *rd)
 {
 	vcpu->arch.vcpu_debug_state.dbg_bvr[rd->CRm] = rd->val;
+	return rd->val;
 }
 
 static bool trap_bcr(struct kvm_vcpu *vcpu,
@@ -589,10 +592,11 @@ static int get_bcr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *rd,
 	return 0;
 }
 
-static void reset_bcr(struct kvm_vcpu *vcpu,
+static u64 reset_bcr(struct kvm_vcpu *vcpu,
 		      const struct sys_reg_desc *rd)
 {
 	vcpu->arch.vcpu_debug_state.dbg_bcr[rd->CRm] = rd->val;
+	return rd->val;
 }
 
 static bool trap_wvr(struct kvm_vcpu *vcpu,
@@ -626,10 +630,11 @@ static int get_wvr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *rd,
 	return 0;
 }
 
-static void reset_wvr(struct kvm_vcpu *vcpu,
+static u64 reset_wvr(struct kvm_vcpu *vcpu,
 		      const struct sys_reg_desc *rd)
 {
 	vcpu->arch.vcpu_debug_state.dbg_wvr[rd->CRm] = rd->val;
+	return rd->val;
 }
 
 static bool trap_wcr(struct kvm_vcpu *vcpu,
@@ -662,25 +667,28 @@ static int get_wcr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *rd,
 	return 0;
 }
 
-static void reset_wcr(struct kvm_vcpu *vcpu,
+static u64 reset_wcr(struct kvm_vcpu *vcpu,
 		      const struct sys_reg_desc *rd)
 {
 	vcpu->arch.vcpu_debug_state.dbg_wcr[rd->CRm] = rd->val;
+	return rd->val;
 }
 
-static void reset_amair_el1(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
+static u64 reset_amair_el1(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 {
 	u64 amair = read_sysreg(amair_el1);
 	vcpu_write_sys_reg(vcpu, amair, AMAIR_EL1);
+	return amair;
 }
 
-static void reset_actlr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
+static u64 reset_actlr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 {
 	u64 actlr = read_sysreg(actlr_el1);
 	vcpu_write_sys_reg(vcpu, actlr, ACTLR_EL1);
+	return actlr;
 }
 
-static void reset_mpidr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
+static u64 reset_mpidr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 {
 	u64 mpidr;
 
@@ -694,7 +702,10 @@ static void reset_mpidr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 	mpidr = (vcpu->vcpu_id & 0x0f) << MPIDR_LEVEL_SHIFT(0);
 	mpidr |= ((vcpu->vcpu_id >> 4) & 0xff) << MPIDR_LEVEL_SHIFT(1);
 	mpidr |= ((vcpu->vcpu_id >> 12) & 0xff) << MPIDR_LEVEL_SHIFT(2);
-	vcpu_write_sys_reg(vcpu, (1ULL << 31) | mpidr, MPIDR_EL1);
+	mpidr |= (1ULL << 31);
+	vcpu_write_sys_reg(vcpu, mpidr, MPIDR_EL1);
+
+	return mpidr;
 }
 
 static unsigned int pmu_visibility(const struct kvm_vcpu *vcpu,
@@ -706,13 +717,13 @@ static unsigned int pmu_visibility(const struct kvm_vcpu *vcpu,
 	return REG_HIDDEN;
 }
 
-static void reset_pmu_reg(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
+static u64 reset_pmu_reg(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 {
 	u64 n, mask = BIT(ARMV8_PMU_CYCLE_IDX);
 
 	/* No PMU available, any PMU reg may UNDEF... */
 	if (!kvm_arm_support_pmu_v3())
-		return;
+		return 0;
 
 	n = read_sysreg(pmcr_el0) >> ARMV8_PMU_PMCR_N_SHIFT;
 	n &= ARMV8_PMU_PMCR_N_MASK;
@@ -721,33 +732,41 @@ static void reset_pmu_reg(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 
 	reset_unknown(vcpu, r);
 	__vcpu_sys_reg(vcpu, r->reg) &= mask;
+
+	return __vcpu_sys_reg(vcpu, r->reg);
 }
 
-static void reset_pmevcntr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
+static u64 reset_pmevcntr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 {
 	reset_unknown(vcpu, r);
 	__vcpu_sys_reg(vcpu, r->reg) &= GENMASK(31, 0);
+
+	return __vcpu_sys_reg(vcpu, r->reg);
 }
 
-static void reset_pmevtyper(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
+static u64 reset_pmevtyper(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 {
 	reset_unknown(vcpu, r);
 	__vcpu_sys_reg(vcpu, r->reg) &= ARMV8_PMU_EVTYPE_MASK;
+
+	return __vcpu_sys_reg(vcpu, r->reg);
 }
 
-static void reset_pmselr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
+static u64 reset_pmselr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 {
 	reset_unknown(vcpu, r);
 	__vcpu_sys_reg(vcpu, r->reg) &= ARMV8_PMU_COUNTER_MASK;
+
+	return __vcpu_sys_reg(vcpu, r->reg);
 }
 
-static void reset_pmcr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
+static u64 reset_pmcr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 {
 	u64 pmcr;
 
 	/* No PMU available, PMCR_EL0 may UNDEF... */
 	if (!kvm_arm_support_pmu_v3())
-		return;
+		return 0;
 
 	/* Only preserve PMCR_EL0.N, and reset the rest to 0 */
 	pmcr = read_sysreg(pmcr_el0) & (ARMV8_PMU_PMCR_N_MASK << ARMV8_PMU_PMCR_N_SHIFT);
@@ -755,6 +774,8 @@ static void reset_pmcr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 		pmcr |= ARMV8_PMU_PMCR_LC;
 
 	__vcpu_sys_reg(vcpu, r->reg) = pmcr;
+
+	return __vcpu_sys_reg(vcpu, r->reg);
 }
 
 static bool check_pmu_access_disabled(struct kvm_vcpu *vcpu, u64 flags)
@@ -1187,25 +1208,89 @@ static bool access_arch_timer(struct kvm_vcpu *vcpu,
 	return true;
 }
 
-static u8 vcpu_pmuver(const struct kvm_vcpu *vcpu)
+static s64 kvm_arm64_ftr_safe_value(u32 id, const struct arm64_ftr_bits *ftrp,
+				    s64 new, s64 cur)
 {
-	if (kvm_vcpu_has_pmu(vcpu))
+	struct arm64_ftr_bits kvm_ftr = *ftrp;
-		return vcpu->kvm->arch.dfr0_pmuver.imp;
 
-	return vcpu->kvm->arch.dfr0_pmuver.unimp;
+	/* Some features have different safe value type in KVM than host features */
+	switch (id) {
+	case SYS_ID_AA64DFR0_EL1:
+		if (kvm_ftr.shift == ID_AA64DFR0_EL1_PMUVer_SHIFT)
+			kvm_ftr.type = FTR_LOWER_SAFE;
+		break;
+	case SYS_ID_DFR0_EL1:
+		if (kvm_ftr.shift == ID_DFR0_EL1_PerfMon_SHIFT)
+			kvm_ftr.type = FTR_LOWER_SAFE;
+		break;
+	}
+
+	return arm64_ftr_safe_value(&kvm_ftr, new, cur);
 }
 
-static u8 perfmon_to_pmuver(u8 perfmon)
+/**
+ * arm64_check_features() - Check if a feature register value constitutes
+ * a subset of features indicated by the idreg's KVM sanitised limit.
+ *
+ * This function will check if each feature field of @val is the "safe" value
+ * against idreg's KVM sanitised limit return from reset() callback.
+ * If a field value in @val is the same as the one in limit, it is always
+ * considered the safe value regardless For register fields that are not in
+ * writable, only the value in limit is considered the safe value.
+ *
+ * Return: 0 if all the fields are safe. Otherwise, return negative errno.
+ */
+static int arm64_check_features(struct kvm_vcpu *vcpu,
+				const struct sys_reg_desc *rd,
+				u64 val)
 {
-	switch (perfmon) {
+	const struct arm64_ftr_reg *ftr_reg;
-	case ID_DFR0_EL1_PerfMon_PMUv3:
+	const struct arm64_ftr_bits *ftrp = NULL;
-		return ID_AA64DFR0_EL1_PMUVer_IMP;
+	u32 id = reg_to_encoding(rd);
-	case ID_DFR0_EL1_PerfMon_IMPDEF:
+	u64 writable_mask = rd->val;
-		return ID_AA64DFR0_EL1_PMUVer_IMP_DEF;
+	u64 limit = rd->reset(vcpu, rd);
-	default:
+	u64 mask = 0;
-		/* Anything ARMv8.1+ and NI have the same value. For now. */
+
-		return perfmon;
+	/*
+	 * Hidden and unallocated ID registers may not have a corresponding
+	 * struct arm64_ftr_reg. Of course, if the register is RAZ we know the
+	 * only safe value is 0.
+	 */
+	if (sysreg_visible_as_raz(vcpu, rd))
+		return val ? -E2BIG : 0;
+
+	ftr_reg = get_arm64_ftr_reg(id);
+	if (!ftr_reg)
+		return -EINVAL;
+
+	ftrp = ftr_reg->ftr_bits;
+
+	for (; ftrp && ftrp->width; ftrp++) {
+		s64 f_val, f_lim, safe_val;
+		u64 ftr_mask;
+
+		ftr_mask = arm64_ftr_mask(ftrp);
+		if ((ftr_mask & writable_mask) != ftr_mask)
+			continue;
+
+		f_val = arm64_ftr_value(ftrp, val);
+		f_lim = arm64_ftr_value(ftrp, limit);
+		mask |= ftr_mask;
+
+		if (f_val == f_lim)
+			safe_val = f_val;
+		else
+			safe_val = kvm_arm64_ftr_safe_value(id, ftrp, f_val, f_lim);
+
+		if (safe_val != f_val)
+			return -E2BIG;
 	}
+
+	/* For fields that are not writable, values in limit are the safe values. */
+	if ((val & ~mask) != (limit & ~mask))
+		return -E2BIG;
+
+	return 0;
 }
 
 static u8 pmuver_to_perfmon(u8 pmuver)
@@ -1222,7 +1307,8 @@ static u8 pmuver_to_perfmon(u8 pmuver)
 }
 
 /* Read a sanitised cpufeature ID register by sys_reg_desc */
-static u64 read_id_reg(const struct kvm_vcpu *vcpu, struct sys_reg_desc const *r)
+static u64 __kvm_read_sanitised_id_reg(const struct kvm_vcpu *vcpu,
+				       const struct sys_reg_desc *r)
 {
 	u32 id = reg_to_encoding(r);
 	u64 val;
@@ -1233,19 +1319,6 @@ static u64 read_id_reg(const struct kvm_vcpu *vcpu, struct sys_reg_desc const *r
 	val = read_sanitised_ftr_reg(id);
 
 	switch (id) {
-	case SYS_ID_AA64PFR0_EL1:
-		if (!vcpu_has_sve(vcpu))
-			val &= ~ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_SVE);
-		val &= ~ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_AMU);
-		val &= ~ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_CSV2);
-		val |= FIELD_PREP(ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_CSV2), (u64)vcpu->kvm->arch.pfr0_csv2);
-		val &= ~ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_CSV3);
-		val |= FIELD_PREP(ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_CSV3), (u64)vcpu->kvm->arch.pfr0_csv3);
-		if (kvm_vgic_global_state.type == VGIC_V3) {
-			val &= ~ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_GIC);
-			val |= FIELD_PREP(ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_GIC), 1);
-		}
-		break;
 	case SYS_ID_AA64PFR1_EL1:
 		if (!kvm_has_mte(vcpu->kvm))
 			val &= ~ARM64_FEATURE_MASK(ID_AA64PFR1_EL1_MTE);
@@ -1267,22 +1340,6 @@ static u64 read_id_reg(const struct kvm_vcpu *vcpu, struct sys_reg_desc const *r
 			val &= ~ARM64_FEATURE_MASK(ID_AA64ISAR2_EL1_WFxT);
 		val &= ~ARM64_FEATURE_MASK(ID_AA64ISAR2_EL1_MOPS);
 		break;
-	case SYS_ID_AA64DFR0_EL1:
-		/* Limit debug to ARMv8.0 */
-		val &= ~ARM64_FEATURE_MASK(ID_AA64DFR0_EL1_DebugVer);
-		val |= FIELD_PREP(ARM64_FEATURE_MASK(ID_AA64DFR0_EL1_DebugVer), 6);
-		/* Set PMUver to the required version */
-		val &= ~ARM64_FEATURE_MASK(ID_AA64DFR0_EL1_PMUVer);
-		val |= FIELD_PREP(ARM64_FEATURE_MASK(ID_AA64DFR0_EL1_PMUVer),
-				  vcpu_pmuver(vcpu));
-		/* Hide SPE from guests */
-		val &= ~ARM64_FEATURE_MASK(ID_AA64DFR0_EL1_PMSVer);
-		break;
-	case SYS_ID_DFR0_EL1:
-		val &= ~ARM64_FEATURE_MASK(ID_DFR0_EL1_PerfMon);
-		val |= FIELD_PREP(ARM64_FEATURE_MASK(ID_DFR0_EL1_PerfMon),
-				  pmuver_to_perfmon(vcpu_pmuver(vcpu)));
-		break;
 	case SYS_ID_AA64MMFR2_EL1:
 		val &= ~ID_AA64MMFR2_EL1_CCIDX_MASK;
 		break;
@@ -1294,6 +1351,28 @@ static u64 read_id_reg(const struct kvm_vcpu *vcpu, struct sys_reg_desc const *r
 	return val;
 }
 
+static u64 kvm_read_sanitised_id_reg(struct kvm_vcpu *vcpu,
+				     const struct sys_reg_desc *r)
+{
+	return __kvm_read_sanitised_id_reg(vcpu, r);
+}
+
+static u64 read_id_reg(const struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
+{
+	return IDREG(vcpu->kvm, reg_to_encoding(r));
+}
+
+/*
+ * Return true if the register's (Op0, Op1, CRn, CRm, Op2) is
+ * (3, 0, 0, crm, op2), where 1<=crm<8, 0<=op2<8.
+ */
+static inline bool is_id_reg(u32 id)
+{
+	return (sys_reg_Op0(id) == 3 && sys_reg_Op1(id) == 0 &&
+		sys_reg_CRn(id) == 0 && sys_reg_CRm(id) >= 1 &&
+		sys_reg_CRm(id) < 8);
+}
+
 static unsigned int id_visibility(const struct kvm_vcpu *vcpu,
 				  const struct sys_reg_desc *r)
 {
@@ -1355,88 +1434,113 @@ static unsigned int sve_visibility(const struct kvm_vcpu *vcpu,
 	return REG_HIDDEN;
 }
 
-static int set_id_aa64pfr0_el1(struct kvm_vcpu *vcpu,
+static u64 read_sanitised_id_aa64pfr0_el1(struct kvm_vcpu *vcpu,
-			       const struct sys_reg_desc *rd,
+					  const struct sys_reg_desc *rd)
-			       u64 val)
 {
-	u8 csv2, csv3;
+	u64 val = read_sanitised_ftr_reg(SYS_ID_AA64PFR0_EL1);
+
+	if (!vcpu_has_sve(vcpu))
+		val &= ~ID_AA64PFR0_EL1_SVE_MASK;
 
 	/*
-	 * Allow AA64PFR0_EL1.CSV2 to be set from userspace as long as
+	 * The default is to expose CSV2 == 1 if the HW isn't affected.
-	 * it doesn't promise more than what is actually provided (the
+	 * Although this is a per-CPU feature, we make it global because
-	 * guest could otherwise be covered in ectoplasmic residue).
+	 * asymmetric systems are just a nuisance.
+	 *
+	 * Userspace can override this as long as it doesn't promise
+	 * the impossible.
 	 */
-	csv2 = cpuid_feature_extract_unsigned_field(val, ID_AA64PFR0_EL1_CSV2_SHIFT);
+	if (arm64_get_spectre_v2_state() == SPECTRE_UNAFFECTED) {
-	if (csv2 > 1 ||
+		val &= ~ID_AA64PFR0_EL1_CSV2_MASK;
-	    (csv2 && arm64_get_spectre_v2_state() != SPECTRE_UNAFFECTED))
+		val |= SYS_FIELD_PREP_ENUM(ID_AA64PFR0_EL1, CSV2, IMP);
-		return -EINVAL;
+	}
+	if (arm64_get_meltdown_state() == SPECTRE_UNAFFECTED) {
+		val &= ~ID_AA64PFR0_EL1_CSV3_MASK;
+		val |= SYS_FIELD_PREP_ENUM(ID_AA64PFR0_EL1, CSV3, IMP);
+	}
 
-	/* Same thing for CSV3 */
+	if (kvm_vgic_global_state.type == VGIC_V3) {
-	csv3 = cpuid_feature_extract_unsigned_field(val, ID_AA64PFR0_EL1_CSV3_SHIFT);
+		val &= ~ID_AA64PFR0_EL1_GIC_MASK;
-	if (csv3 > 1 ||
+		val |= SYS_FIELD_PREP_ENUM(ID_AA64PFR0_EL1, GIC, IMP);
-	    (csv3 && arm64_get_meltdown_state() != SPECTRE_UNAFFECTED))
+	}
-		return -EINVAL;
 
-	/* We can only differ with CSV[23], and anything else is an error */
+	val &= ~ID_AA64PFR0_EL1_AMU_MASK;
-	val ^= read_id_reg(vcpu, rd);
-	val &= ~(ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_CSV2) |
-		 ARM64_FEATURE_MASK(ID_AA64PFR0_EL1_CSV3));
-	if (val)
-		return -EINVAL;
 
-	vcpu->kvm->arch.pfr0_csv2 = csv2;
+	return val;
-	vcpu->kvm->arch.pfr0_csv3 = csv3;
+}
 
-	return 0;
+static u64 read_sanitised_id_aa64dfr0_el1(struct kvm_vcpu *vcpu,
+					  const struct sys_reg_desc *rd)
+{
+	u64 val = read_sanitised_ftr_reg(SYS_ID_AA64DFR0_EL1);
+
+	/* Limit debug to ARMv8.0 */
+	val &= ~ID_AA64DFR0_EL1_DebugVer_MASK;
+	val |= SYS_FIELD_PREP_ENUM(ID_AA64DFR0_EL1, DebugVer, IMP);
+
+	/*
+	 * Only initialize the PMU version if the vCPU was configured with one.
+	 */
+	val &= ~ID_AA64DFR0_EL1_PMUVer_MASK;
+	if (kvm_vcpu_has_pmu(vcpu))
+		val |= SYS_FIELD_PREP(ID_AA64DFR0_EL1, PMUVer,
+				      kvm_arm_pmu_get_pmuver_limit());
+
+	/* Hide SPE from guests */
+	val &= ~ID_AA64DFR0_EL1_PMSVer_MASK;
+
+	return val;
 }
 
 static int set_id_aa64dfr0_el1(struct kvm_vcpu *vcpu,
 			       const struct sys_reg_desc *rd,
 			       u64 val)
 {
-	u8 pmuver, host_pmuver;
+	u8 pmuver = SYS_FIELD_GET(ID_AA64DFR0_EL1, PMUVer, val);
-	bool valid_pmu;
-
-	host_pmuver = kvm_arm_pmu_get_pmuver_limit();
 
 	/*
-	 * Allow AA64DFR0_EL1.PMUver to be set from userspace as long
+	 * Prior to commit 3d0dba5764b9 ("KVM: arm64: PMU: Move the
-	 * as it doesn't promise more than what the HW gives us. We
+	 * ID_AA64DFR0_EL1.PMUver limit to VM creation"), KVM erroneously
-	 * allow an IMPDEF PMU though, only if no PMU is supported
+	 * exposed an IMP_DEF PMU to userspace and the guest on systems w/
-	 * (KVM backward compatibility handling).
+	 * non-architectural PMUs. Of course, PMUv3 is the only game in town for
+	 * PMU virtualization, so the IMP_DEF value was rather user-hostile.
+	 *
+	 * At minimum, we're on the hook to allow values that were given to
+	 * userspace by KVM. Cover our tracks here and replace the IMP_DEF value
+	 * with a more sensible NI. The value of an ID register changing under
+	 * the nose of the guest is unfortunate, but is certainly no more
+	 * surprising than an ill-guided PMU driver poking at impdef system
+	 * registers that end in an UNDEF...
 	 */
-	pmuver = FIELD_GET(ARM64_FEATURE_MASK(ID_AA64DFR0_EL1_PMUVer), val);
+	if (pmuver == ID_AA64DFR0_EL1_PMUVer_IMP_DEF)
-	if ((pmuver != ID_AA64DFR0_EL1_PMUVer_IMP_DEF && pmuver > host_pmuver))
+		val &= ~ID_AA64DFR0_EL1_PMUVer_MASK;
-		return -EINVAL;
 
-	valid_pmu = (pmuver != 0 && pmuver != ID_AA64DFR0_EL1_PMUVer_IMP_DEF);
+	return set_id_reg(vcpu, rd, val);
+}
 
-	/* Make sure view register and PMU support do match */
+static u64 read_sanitised_id_dfr0_el1(struct kvm_vcpu *vcpu,
-	if (kvm_vcpu_has_pmu(vcpu) != valid_pmu)
+				      const struct sys_reg_desc *rd)
-		return -EINVAL;
+{
+	u8 perfmon = pmuver_to_perfmon(kvm_arm_pmu_get_pmuver_limit());
+	u64 val = read_sanitised_ftr_reg(SYS_ID_DFR0_EL1);
 
-	/* We can only differ with PMUver, and anything else is an error */
+	val &= ~ID_DFR0_EL1_PerfMon_MASK;
-	val ^= read_id_reg(vcpu, rd);
+	if (kvm_vcpu_has_pmu(vcpu))
-	val &= ~ARM64_FEATURE_MASK(ID_AA64DFR0_EL1_PMUVer);
+		val |= SYS_FIELD_PREP(ID_DFR0_EL1, PerfMon, perfmon);
-	if (val)
-		return -EINVAL;
 
-	if (valid_pmu)
+	return val;
-		vcpu->kvm->arch.dfr0_pmuver.imp = pmuver;
-	else
-		vcpu->kvm->arch.dfr0_pmuver.unimp = pmuver;
-
-	return 0;
 }
 
 static int set_id_dfr0_el1(struct kvm_vcpu *vcpu,
 			   const struct sys_reg_desc *rd,
 			   u64 val)
 {
-	u8 perfmon, host_perfmon;
+	u8 perfmon = SYS_FIELD_GET(ID_DFR0_EL1, PerfMon, val);
-	bool valid_pmu;
 
-	host_perfmon = pmuver_to_perfmon(kvm_arm_pmu_get_pmuver_limit());
+	if (perfmon == ID_DFR0_EL1_PerfMon_IMPDEF) {
+		val &= ~ID_DFR0_EL1_PerfMon_MASK;
+		perfmon = 0;
+	}
 
 	/*
 	 * Allow DFR0_EL1.PerfMon to be set from userspace as long as
@@ -1444,29 +1548,10 @@ static int set_id_dfr0_el1(struct kvm_vcpu *vcpu,
 	 * AArch64 side (as everything is emulated with that), and
 	 * that this is a PMUv3.
 	 */
-	perfmon = FIELD_GET(ARM64_FEATURE_MASK(ID_DFR0_EL1_PerfMon), val);
+	if (perfmon != 0 && perfmon < ID_DFR0_EL1_PerfMon_PMUv3)
-	if ((perfmon != ID_DFR0_EL1_PerfMon_IMPDEF && perfmon > host_perfmon) ||
-	    (perfmon != 0 && perfmon < ID_DFR0_EL1_PerfMon_PMUv3))
 		return -EINVAL;
 
-	valid_pmu = (perfmon != 0 && perfmon != ID_DFR0_EL1_PerfMon_IMPDEF);
+	return set_id_reg(vcpu, rd, val);
-
-	/* Make sure view register and PMU support do match */
-	if (kvm_vcpu_has_pmu(vcpu) != valid_pmu)
-		return -EINVAL;
-
-	/* We can only differ with PerfMon, and anything else is an error */
-	val ^= read_id_reg(vcpu, rd);
-	val &= ~ARM64_FEATURE_MASK(ID_DFR0_EL1_PerfMon);
-	if (val)
-		return -EINVAL;
-
-	if (valid_pmu)
-		vcpu->kvm->arch.dfr0_pmuver.imp = perfmon_to_pmuver(perfmon);
-	else
-		vcpu->kvm->arch.dfr0_pmuver.unimp = perfmon_to_pmuver(perfmon);
-
-	return 0;
 }
 
 /*
@@ -1479,18 +1564,60 @@ static int set_id_dfr0_el1(struct kvm_vcpu *vcpu,
 static int get_id_reg(struct kvm_vcpu *vcpu, const struct sys_reg_desc *rd,
 		      u64 *val)
 {
+	/*
+	 * Avoid locking if the VM has already started, as the ID registers are
+	 * guaranteed to be invariant at that point.
+	 */
+	if (kvm_vm_has_ran_once(vcpu->kvm)) {
+		*val = read_id_reg(vcpu, rd);
+		return 0;
+	}
+
+	mutex_lock(&vcpu->kvm->arch.config_lock);
 	*val = read_id_reg(vcpu, rd);
+	mutex_unlock(&vcpu->kvm->arch.config_lock);
+
 	return 0;
 }
 
 static int set_id_reg(struct kvm_vcpu *vcpu, const struct sys_reg_desc *rd,
 		      u64 val)
 {
-	/* This is what we mean by invariant: you can't change it. */
+	u32 id = reg_to_encoding(rd);
-	if (val != read_id_reg(vcpu, rd))
+	int ret;
-		return -EINVAL;
 
-	return 0;
+	mutex_lock(&vcpu->kvm->arch.config_lock);
+
+	/*
+	 * Once the VM has started the ID registers are immutable. Reject any
+	 * write that does not match the final register value.
+	 */
+	if (kvm_vm_has_ran_once(vcpu->kvm)) {
+		if (val != read_id_reg(vcpu, rd))
+			ret = -EBUSY;
+		else
+			ret = 0;
+
+		mutex_unlock(&vcpu->kvm->arch.config_lock);
+		return ret;
+	}
+
+	ret = arm64_check_features(vcpu, rd, val);
+	if (!ret)
+		IDREG(vcpu->kvm, id) = val;
+
+	mutex_unlock(&vcpu->kvm->arch.config_lock);
+
+	/*
+	 * arm64_check_features() returns -E2BIG to indicate the register's
+	 * feature set is a superset of the maximally-allowed register value.
+	 * While it would be nice to precisely describe this to userspace, the
+	 * existing UAPI for KVM_SET_ONE_REG has it that invalid register
+	 * writes return -EINVAL.
+	 */
+	if (ret == -E2BIG)
+		ret = -EINVAL;
+	return ret;
 }
 
 static int get_raz_reg(struct kvm_vcpu *vcpu, const struct sys_reg_desc *rd,
@@ -1530,7 +1657,7 @@ static bool access_clidr(struct kvm_vcpu *vcpu, struct sys_reg_params *p,
  * Fabricate a CLIDR_EL1 value instead of using the real value, which can vary
  * by the physical CPU which the vcpu currently resides in.
  */
-static void reset_clidr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
+static u64 reset_clidr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 {
 	u64 ctr_el0 = read_sanitised_ftr_reg(SYS_CTR_EL0);
 	u64 clidr;
@@ -1578,6 +1705,8 @@ static void reset_clidr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 		clidr |= 2 << CLIDR_TTYPE_SHIFT(loc);
 
 	__vcpu_sys_reg(vcpu, r->reg) = clidr;
+
+	return __vcpu_sys_reg(vcpu, r->reg);
 }
 
 static int set_clidr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *rd,
@@ -1677,6 +1806,17 @@ static unsigned int elx2_visibility(const struct kvm_vcpu *vcpu,
 	.visibility = elx2_visibility,		\
 }
 
+/*
+ * Since reset() callback and field val are not used for idregs, they will be
+ * used for specific purposes for idregs.
+ * The reset() would return KVM sanitised register value. The value would be the
+ * same as the host kernel sanitised value if there is no KVM sanitisation.
+ * The val would be used as a mask indicating writable fields for the idreg.
+ * Only bits with 1 are writable from userspace. This mask might not be
+ * necessary in the future whenever all ID registers are enabled as writable
+ * from userspace.
+ */
+
 /* sys_reg_desc initialiser for known cpufeature ID registers */
 #define ID_SANITISED(name) {			\
 	SYS_DESC(SYS_##name),			\
@@ -1684,6 +1824,8 @@ static unsigned int elx2_visibility(const struct kvm_vcpu *vcpu,
 	.get_user = get_id_reg,			\
 	.set_user = set_id_reg,			\
 	.visibility = id_visibility,		\
+	.reset = kvm_read_sanitised_id_reg,	\
+	.val = 0,				\
 }
 
 /* sys_reg_desc initialiser for known cpufeature ID registers */
@@ -1693,6 +1835,8 @@ static unsigned int elx2_visibility(const struct kvm_vcpu *vcpu,
 	.get_user = get_id_reg,			\
 	.set_user = set_id_reg,			\
 	.visibility = aa32_id_visibility,	\
+	.reset = kvm_read_sanitised_id_reg,	\
+	.val = 0,				\
 }
 
 /*
@@ -1705,7 +1849,9 @@ static unsigned int elx2_visibility(const struct kvm_vcpu *vcpu,
 	.access = access_id_reg,			\
 	.get_user = get_id_reg,				\
 	.set_user = set_id_reg,				\
-	.visibility = raz_visibility			\
+	.visibility = raz_visibility,			\
+	.reset = kvm_read_sanitised_id_reg,		\
+	.val = 0,					\
 }
 
 /*
@@ -1719,6 +1865,8 @@ static unsigned int elx2_visibility(const struct kvm_vcpu *vcpu,
 	.get_user = get_id_reg,			\
 	.set_user = set_id_reg,			\
 	.visibility = raz_visibility,		\
+	.reset = kvm_read_sanitised_id_reg,	\
+	.val = 0,				\
 }
 
 static bool access_sp_el1(struct kvm_vcpu *vcpu,
@@ -1826,9 +1974,13 @@ static const struct sys_reg_desc sys_reg_descs[] = {
 	/* CRm=1 */
 	AA32_ID_SANITISED(ID_PFR0_EL1),
 	AA32_ID_SANITISED(ID_PFR1_EL1),
-	{ SYS_DESC(SYS_ID_DFR0_EL1), .access = access_id_reg,
+	{ SYS_DESC(SYS_ID_DFR0_EL1),
-	  .get_user = get_id_reg, .set_user = set_id_dfr0_el1,
+	  .access = access_id_reg,
-	  .visibility = aa32_id_visibility, },
+	  .get_user = get_id_reg,
+	  .set_user = set_id_dfr0_el1,
+	  .visibility = aa32_id_visibility,
+	  .reset = read_sanitised_id_dfr0_el1,
+	  .val = ID_DFR0_EL1_PerfMon_MASK, },
 	ID_HIDDEN(ID_AFR0_EL1),
 	AA32_ID_SANITISED(ID_MMFR0_EL1),
 	AA32_ID_SANITISED(ID_MMFR1_EL1),
@@ -1857,8 +2009,12 @@ static const struct sys_reg_desc sys_reg_descs[] = {
 
 	/* AArch64 ID registers */
 	/* CRm=4 */
-	{ SYS_DESC(SYS_ID_AA64PFR0_EL1), .access = access_id_reg,
+	{ SYS_DESC(SYS_ID_AA64PFR0_EL1),
-	  .get_user = get_id_reg, .set_user = set_id_aa64pfr0_el1, },
+	  .access = access_id_reg,
+	  .get_user = get_id_reg,
+	  .set_user = set_id_reg,
+	  .reset = read_sanitised_id_aa64pfr0_el1,
+	  .val = ID_AA64PFR0_EL1_CSV2_MASK | ID_AA64PFR0_EL1_CSV3_MASK, },
 	ID_SANITISED(ID_AA64PFR1_EL1),
 	ID_UNALLOCATED(4,2),
 	ID_UNALLOCATED(4,3),
@@ -1868,8 +2024,12 @@ static const struct sys_reg_desc sys_reg_descs[] = {
 	ID_UNALLOCATED(4,7),
 
 	/* CRm=5 */
-	{ SYS_DESC(SYS_ID_AA64DFR0_EL1), .access = access_id_reg,
+	{ SYS_DESC(SYS_ID_AA64DFR0_EL1),
-	  .get_user = get_id_reg, .set_user = set_id_aa64dfr0_el1, },
+	  .access = access_id_reg,
+	  .get_user = get_id_reg,
+	  .set_user = set_id_aa64dfr0_el1,
+	  .reset = read_sanitised_id_aa64dfr0_el1,
+	  .val = ID_AA64DFR0_EL1_PMUVer_MASK, },
 	ID_SANITISED(ID_AA64DFR1_EL1),
 	ID_UNALLOCATED(5,2),
 	ID_UNALLOCATED(5,3),
@@ -2203,7 +2363,7 @@ static const struct sys_reg_desc sys_reg_descs[] = {
 	EL2_REG(ACTLR_EL2, access_rw, reset_val, 0),
 	EL2_REG(HCR_EL2, access_rw, reset_val, 0),
 	EL2_REG(MDCR_EL2, access_rw, reset_val, 0),
-	EL2_REG(CPTR_EL2, access_rw, reset_val, CPTR_EL2_DEFAULT ),
+	EL2_REG(CPTR_EL2, access_rw, reset_val, CPTR_NVHE_EL2_RES1),
 	EL2_REG(HSTR_EL2, access_rw, reset_val, 0),
 	EL2_REG(HACR_EL2, access_rw, reset_val, 0),
 
@@ -2260,6 +2420,8 @@ static const struct sys_reg_desc sys_reg_descs[] = {
 	EL2_REG(SP_EL2, NULL, reset_unknown, 0),
 };
 
+static const struct sys_reg_desc *first_idreg;
+
 static bool trap_dbgdidr(struct kvm_vcpu *vcpu,
 			struct sys_reg_params *p,
 			const struct sys_reg_desc *r)
@@ -2950,6 +3112,28 @@ static bool emulate_sys_reg(struct kvm_vcpu *vcpu,
 	return false;
 }
 
+static void kvm_reset_id_regs(struct kvm_vcpu *vcpu)
+{
+	const struct sys_reg_desc *idreg = first_idreg;
+	u32 id = reg_to_encoding(idreg);
+	struct kvm *kvm = vcpu->kvm;
+
+	if (test_bit(KVM_ARCH_FLAG_ID_REGS_INITIALIZED, &kvm->arch.flags))
+		return;
+
+	lockdep_assert_held(&kvm->arch.config_lock);
+
+	/* Initialize all idregs */
+	while (is_id_reg(id)) {
+		IDREG(kvm, id) = idreg->reset(vcpu, idreg);
+
+		idreg++;
+		id = reg_to_encoding(idreg);
+	}
+
+	set_bit(KVM_ARCH_FLAG_ID_REGS_INITIALIZED, &kvm->arch.flags);
+}
+
 /**
  * kvm_reset_sys_regs - sets system registers to reset value
  * @vcpu: The VCPU pointer
@@ -2961,9 +3145,17 @@ void kvm_reset_sys_regs(struct kvm_vcpu *vcpu)
 {
 	unsigned long i;
 
-	for (i = 0; i < ARRAY_SIZE(sys_reg_descs); i++)
+	kvm_reset_id_regs(vcpu);
-		if (sys_reg_descs[i].reset)
+
-			sys_reg_descs[i].reset(vcpu, &sys_reg_descs[i]);
+	for (i = 0; i < ARRAY_SIZE(sys_reg_descs); i++) {
+		const struct sys_reg_desc *r = &sys_reg_descs[i];
+
+		if (is_id_reg(reg_to_encoding(r)))
+			continue;
+
+		if (r->reset)
+			r->reset(vcpu, r);
+	}
 }
 
 /**
@@ -3064,19 +3256,21 @@ id_to_sys_reg_desc(struct kvm_vcpu *vcpu, u64 id,
  */
 
 #define FUNCTION_INVARIANT(reg)						\
-	static void get_##reg(struct kvm_vcpu *v,			\
+	static u64 get_##reg(struct kvm_vcpu *v,			\
 			      const struct sys_reg_desc *r)		\
 	{								\
 		((struct sys_reg_desc *)r)->val = read_sysreg(reg);	\
+		return ((struct sys_reg_desc *)r)->val;			\
 	}
 
 FUNCTION_INVARIANT(midr_el1)
 FUNCTION_INVARIANT(revidr_el1)
 FUNCTION_INVARIANT(aidr_el1)
 
-static void get_ctr_el0(struct kvm_vcpu *v, const struct sys_reg_desc *r)
+static u64 get_ctr_el0(struct kvm_vcpu *v, const struct sys_reg_desc *r)
 {
 	((struct sys_reg_desc *)r)->val = read_sanitised_ftr_reg(SYS_CTR_EL0);
+	return ((struct sys_reg_desc *)r)->val;
 }
 
 /* ->val is filled in by kvm_sys_reg_table_init() */
@@ -3368,6 +3562,7 @@ int kvm_arm_copy_sys_reg_indices(struct kvm_vcpu *vcpu, u64 __user *uindices)
 
 int __init kvm_sys_reg_table_init(void)
 {
+	struct sys_reg_params params;
 	bool valid = true;
 	unsigned int i;
 
@@ -3386,5 +3581,11 @@ int __init kvm_sys_reg_table_init(void)
 	for (i = 0; i < ARRAY_SIZE(invariant_sys_regs); i++)
 		invariant_sys_regs[i].reset(NULL, &invariant_sys_regs[i]);
 
+	/* Find the first idreg (SYS_ID_PFR0_EL1) in sys_reg_descs. */
+	params = encoding_to_params(SYS_ID_PFR0_EL1);
+	first_idreg = find_reg(&params, sys_reg_descs, ARRAY_SIZE(sys_reg_descs));
+	if (!first_idreg)
+		return -EINVAL;
+
 	return 0;
 }

arch/arm64/kvm/sys_regs.h

@@ -27,6 +27,13 @@ struct sys_reg_params {
 	bool	is_write;
 };
 
+#define encoding_to_params(reg)						\
+	((struct sys_reg_params){ .Op0 = sys_reg_Op0(reg),		\
+				  .Op1 = sys_reg_Op1(reg),		\
+				  .CRn = sys_reg_CRn(reg),		\
+				  .CRm = sys_reg_CRm(reg),		\
+				  .Op2 = sys_reg_Op2(reg) })
+
 #define esr_sys64_to_params(esr)                                               \
 	((struct sys_reg_params){ .Op0 = ((esr) >> 20) & 3,                    \
 				  .Op1 = ((esr) >> 14) & 0x7,                  \
@@ -64,13 +71,16 @@ struct sys_reg_desc {
 		       struct sys_reg_params *,
 		       const struct sys_reg_desc *);
 
-	/* Initialization for vcpu. */
+	/*
-	void (*reset)(struct kvm_vcpu *, const struct sys_reg_desc *);
+	 * Initialization for vcpu. Return initialized value, or KVM
+	 * sanitized value for ID registers.
+	 */
+	u64 (*reset)(struct kvm_vcpu *, const struct sys_reg_desc *);
 
 	/* Index into sys_reg[], or 0 if we don't need to save it. */
 	int reg;
 
-	/* Value (usually reset value) */
+	/* Value (usually reset value), or write mask for idregs */
 	u64 val;
 
 	/* Custom get/set_user functions, fallback to generic if NULL */
@@ -123,19 +133,21 @@ static inline bool read_zero(struct kvm_vcpu *vcpu,
 }
 
 /* Reset functions */
-static inline void reset_unknown(struct kvm_vcpu *vcpu,
+static inline u64 reset_unknown(struct kvm_vcpu *vcpu,
 				 const struct sys_reg_desc *r)
 {
 	BUG_ON(!r->reg);
 	BUG_ON(r->reg >= NR_SYS_REGS);
 	__vcpu_sys_reg(vcpu, r->reg) = 0x1de7ec7edbadc0deULL;
+	return __vcpu_sys_reg(vcpu, r->reg);
 }
 
-static inline void reset_val(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
+static inline u64 reset_val(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 {
 	BUG_ON(!r->reg);
 	BUG_ON(r->reg >= NR_SYS_REGS);
 	__vcpu_sys_reg(vcpu, r->reg) = r->val;
+	return __vcpu_sys_reg(vcpu, r->reg);
 }
 
 static inline unsigned int sysreg_visibility(const struct kvm_vcpu *vcpu,

arch/arm64/tools/cpucaps

@@ -25,6 +25,7 @@ HAS_E0PD
 HAS_ECV
 HAS_ECV_CNTPOFF
 HAS_EPAN
+HAS_EVT
 HAS_GENERIC_AUTH
 HAS_GENERIC_AUTH_ARCH_QARMA3
 HAS_GENERIC_AUTH_ARCH_QARMA5
@@ -51,6 +52,7 @@ HAS_TLB_RANGE
 HAS_VIRT_HOST_EXTN
 HAS_WFXT
 HW_DBM
+KVM_HVHE
 KVM_PROTECTED_MODE
 MISMATCHED_CACHE_TYPE
 MTE
@@ -81,6 +83,7 @@ WORKAROUND_2077057
 WORKAROUND_2457168
 WORKAROUND_2645198
 WORKAROUND_2658417
+WORKAROUND_AMPERE_AC03_CPU_38
 WORKAROUND_TRBE_OVERWRITE_FILL_MODE
 WORKAROUND_TSB_FLUSH_FAILURE
 WORKAROUND_TRBE_WRITE_OUT_OF_RANGE

arch/riscv/include/asm/csr.h

@@ -90,7 +90,9 @@
 #define EXC_INST_ACCESS		1
 #define EXC_INST_ILLEGAL	2
 #define EXC_BREAKPOINT		3
+#define EXC_LOAD_MISALIGNED	4
 #define EXC_LOAD_ACCESS		5
+#define EXC_STORE_MISALIGNED	6
 #define EXC_STORE_ACCESS	7
 #define EXC_SYSCALL		8
 #define EXC_HYPERVISOR_SYSCALL	9

arch/riscv/include/asm/kvm_aia.h

@@ -20,6 +20,33 @@ struct kvm_aia {
 
 	/* In-kernel irqchip initialized */
 	bool		initialized;
+
+	/* Virtualization mode (Emulation, HW Accelerated, or Auto) */
+	u32		mode;
+
+	/* Number of MSIs */
+	u32		nr_ids;
+
+	/* Number of wired IRQs */
+	u32		nr_sources;
+
+	/* Number of group bits in IMSIC address */
+	u32		nr_group_bits;
+
+	/* Position of group bits in IMSIC address */
+	u32		nr_group_shift;
+
+	/* Number of hart bits in IMSIC address */
+	u32		nr_hart_bits;
+
+	/* Number of guest bits in IMSIC address */
+	u32		nr_guest_bits;
+
+	/* Guest physical address of APLIC */
+	gpa_t		aplic_addr;
+
+	/* Internal state of APLIC */
+	void		*aplic_state;
 };
 
 struct kvm_vcpu_aia_csr {
@@ -38,25 +65,53 @@ struct kvm_vcpu_aia {
 
 	/* CPU AIA CSR context upon Guest VCPU reset */
 	struct kvm_vcpu_aia_csr guest_reset_csr;
+
+	/* Guest physical address of IMSIC for this VCPU */
+	gpa_t		imsic_addr;
+
+	/* HART index of IMSIC extacted from guest physical address */
+	u32		hart_index;
+
+	/* Internal state of IMSIC for this VCPU */
+	void		*imsic_state;
 };
 
+#define KVM_RISCV_AIA_UNDEF_ADDR	(-1)
+
 #define kvm_riscv_aia_initialized(k)	((k)->arch.aia.initialized)
 
 #define irqchip_in_kernel(k)		((k)->arch.aia.in_kernel)
 
+extern unsigned int kvm_riscv_aia_nr_hgei;
+extern unsigned int kvm_riscv_aia_max_ids;
 DECLARE_STATIC_KEY_FALSE(kvm_riscv_aia_available);
 #define kvm_riscv_aia_available() \
 	static_branch_unlikely(&kvm_riscv_aia_available)
 
+extern struct kvm_device_ops kvm_riscv_aia_device_ops;
+
+void kvm_riscv_vcpu_aia_imsic_release(struct kvm_vcpu *vcpu);
+int kvm_riscv_vcpu_aia_imsic_update(struct kvm_vcpu *vcpu);
+
 #define KVM_RISCV_AIA_IMSIC_TOPEI	(ISELECT_MASK + 1)
-static inline int kvm_riscv_vcpu_aia_imsic_rmw(struct kvm_vcpu *vcpu,
+int kvm_riscv_vcpu_aia_imsic_rmw(struct kvm_vcpu *vcpu, unsigned long isel,
-					       unsigned long isel,
+				 unsigned long *val, unsigned long new_val,
-					       unsigned long *val,
+				 unsigned long wr_mask);
-					       unsigned long new_val,
+int kvm_riscv_aia_imsic_rw_attr(struct kvm *kvm, unsigned long type,
-					       unsigned long wr_mask)
+				bool write, unsigned long *val);
-{
+int kvm_riscv_aia_imsic_has_attr(struct kvm *kvm, unsigned long type);
-	return 0;
+void kvm_riscv_vcpu_aia_imsic_reset(struct kvm_vcpu *vcpu);
-}
+int kvm_riscv_vcpu_aia_imsic_inject(struct kvm_vcpu *vcpu,
+				    u32 guest_index, u32 offset, u32 iid);
+int kvm_riscv_vcpu_aia_imsic_init(struct kvm_vcpu *vcpu);
+void kvm_riscv_vcpu_aia_imsic_cleanup(struct kvm_vcpu *vcpu);
+
+int kvm_riscv_aia_aplic_set_attr(struct kvm *kvm, unsigned long type, u32 v);
+int kvm_riscv_aia_aplic_get_attr(struct kvm *kvm, unsigned long type, u32 *v);
+int kvm_riscv_aia_aplic_has_attr(struct kvm *kvm, unsigned long type);
+int kvm_riscv_aia_aplic_inject(struct kvm *kvm, u32 source, bool level);
+int kvm_riscv_aia_aplic_init(struct kvm *kvm);
+void kvm_riscv_aia_aplic_cleanup(struct kvm *kvm);
 
 #ifdef CONFIG_32BIT
 void kvm_riscv_vcpu_aia_flush_interrupts(struct kvm_vcpu *vcpu);
@@ -93,31 +148,23 @@ int kvm_riscv_vcpu_aia_rmw_ireg(struct kvm_vcpu *vcpu, unsigned int csr_num,
 { .base = CSR_SIREG,      .count = 1, .func = kvm_riscv_vcpu_aia_rmw_ireg }, \
 { .base = CSR_STOPEI,     .count = 1, .func = kvm_riscv_vcpu_aia_rmw_topei },
 
-static inline int kvm_riscv_vcpu_aia_update(struct kvm_vcpu *vcpu)
+int kvm_riscv_vcpu_aia_update(struct kvm_vcpu *vcpu);
-{
+void kvm_riscv_vcpu_aia_reset(struct kvm_vcpu *vcpu);
-	return 1;
+int kvm_riscv_vcpu_aia_init(struct kvm_vcpu *vcpu);
-}
+void kvm_riscv_vcpu_aia_deinit(struct kvm_vcpu *vcpu);
 
-static inline void kvm_riscv_vcpu_aia_reset(struct kvm_vcpu *vcpu)
+int kvm_riscv_aia_inject_msi_by_id(struct kvm *kvm, u32 hart_index,
-{
+				   u32 guest_index, u32 iid);
-}
+int kvm_riscv_aia_inject_msi(struct kvm *kvm, struct kvm_msi *msi);
+int kvm_riscv_aia_inject_irq(struct kvm *kvm, unsigned int irq, bool level);
 
-static inline int kvm_riscv_vcpu_aia_init(struct kvm_vcpu *vcpu)
+void kvm_riscv_aia_init_vm(struct kvm *kvm);
-{
+void kvm_riscv_aia_destroy_vm(struct kvm *kvm);
-	return 0;
-}
 
-static inline void kvm_riscv_vcpu_aia_deinit(struct kvm_vcpu *vcpu)
+int kvm_riscv_aia_alloc_hgei(int cpu, struct kvm_vcpu *owner,
-{
+			     void __iomem **hgei_va, phys_addr_t *hgei_pa);
-}
+void kvm_riscv_aia_free_hgei(int cpu, int hgei);
-
+void kvm_riscv_aia_wakeon_hgei(struct kvm_vcpu *owner, bool enable);
-static inline void kvm_riscv_aia_init_vm(struct kvm *kvm)
-{
-}
-
-static inline void kvm_riscv_aia_destroy_vm(struct kvm *kvm)
-{
-}
 
 void kvm_riscv_aia_enable(void);
 void kvm_riscv_aia_disable(void);

arch/riscv/include/asm/kvm_aia_aplic.h

@@ -0,0 +1,58 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (C) 2022 Ventana Micro Systems Inc.
+ */
+#ifndef __KVM_RISCV_AIA_IMSIC_H
+#define __KVM_RISCV_AIA_IMSIC_H
+
+#include <linux/bitops.h>
+
+#define APLIC_MAX_IDC			BIT(14)
+#define APLIC_MAX_SOURCE		1024
+
+#define APLIC_DOMAINCFG			0x0000
+#define APLIC_DOMAINCFG_RDONLY		0x80000000
+#define APLIC_DOMAINCFG_IE		BIT(8)
+#define APLIC_DOMAINCFG_DM		BIT(2)
+#define APLIC_DOMAINCFG_BE		BIT(0)
+
+#define APLIC_SOURCECFG_BASE		0x0004
+#define APLIC_SOURCECFG_D		BIT(10)
+#define APLIC_SOURCECFG_CHILDIDX_MASK	0x000003ff
+#define APLIC_SOURCECFG_SM_MASK	0x00000007
+#define APLIC_SOURCECFG_SM_INACTIVE	0x0
+#define APLIC_SOURCECFG_SM_DETACH	0x1
+#define APLIC_SOURCECFG_SM_EDGE_RISE	0x4
+#define APLIC_SOURCECFG_SM_EDGE_FALL	0x5
+#define APLIC_SOURCECFG_SM_LEVEL_HIGH	0x6
+#define APLIC_SOURCECFG_SM_LEVEL_LOW	0x7
+
+#define APLIC_IRQBITS_PER_REG		32
+
+#define APLIC_SETIP_BASE		0x1c00
+#define APLIC_SETIPNUM			0x1cdc
+
+#define APLIC_CLRIP_BASE		0x1d00
+#define APLIC_CLRIPNUM			0x1ddc
+
+#define APLIC_SETIE_BASE		0x1e00
+#define APLIC_SETIENUM			0x1edc
+
+#define APLIC_CLRIE_BASE		0x1f00
+#define APLIC_CLRIENUM			0x1fdc
+
+#define APLIC_SETIPNUM_LE		0x2000
+#define APLIC_SETIPNUM_BE		0x2004
+
+#define APLIC_GENMSI			0x3000
+
+#define APLIC_TARGET_BASE		0x3004
+#define APLIC_TARGET_HART_IDX_SHIFT	18
+#define APLIC_TARGET_HART_IDX_MASK	0x3fff
+#define APLIC_TARGET_GUEST_IDX_SHIFT	12
+#define APLIC_TARGET_GUEST_IDX_MASK	0x3f
+#define APLIC_TARGET_IPRIO_MASK	0xff
+#define APLIC_TARGET_EIID_MASK	0x7ff
+
+#endif

arch/riscv/include/asm/kvm_aia_imsic.h

@@ -0,0 +1,38 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (C) 2022 Ventana Micro Systems Inc.
+ */
+#ifndef __KVM_RISCV_AIA_IMSIC_H
+#define __KVM_RISCV_AIA_IMSIC_H
+
+#include <linux/types.h>
+#include <asm/csr.h>
+
+#define IMSIC_MMIO_PAGE_SHIFT		12
+#define IMSIC_MMIO_PAGE_SZ		(1UL << IMSIC_MMIO_PAGE_SHIFT)
+#define IMSIC_MMIO_PAGE_LE		0x00
+#define IMSIC_MMIO_PAGE_BE		0x04
+
+#define IMSIC_MIN_ID			63
+#define IMSIC_MAX_ID			2048
+
+#define IMSIC_EIDELIVERY		0x70
+
+#define IMSIC_EITHRESHOLD		0x72
+
+#define IMSIC_EIP0			0x80
+#define IMSIC_EIP63			0xbf
+#define IMSIC_EIPx_BITS			32
+
+#define IMSIC_EIE0			0xc0
+#define IMSIC_EIE63			0xff
+#define IMSIC_EIEx_BITS			32
+
+#define IMSIC_FIRST			IMSIC_EIDELIVERY
+#define IMSIC_LAST			IMSIC_EIE63
+
+#define IMSIC_MMIO_SETIPNUM_LE		0x00
+#define IMSIC_MMIO_SETIPNUM_BE		0x04
+
+#endif

arch/riscv/include/asm/kvm_host.h

@@ -28,6 +28,8 @@
 
 #define KVM_VCPU_MAX_FEATURES		0
 
+#define KVM_IRQCHIP_NUM_PINS		1024
+
 #define KVM_REQ_SLEEP \
 	KVM_ARCH_REQ_FLAGS(0, KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
 #define KVM_REQ_VCPU_RESET		KVM_ARCH_REQ(1)
@@ -320,6 +322,8 @@ int kvm_riscv_gstage_vmid_init(struct kvm *kvm);
 bool kvm_riscv_gstage_vmid_ver_changed(struct kvm_vmid *vmid);
 void kvm_riscv_gstage_vmid_update(struct kvm_vcpu *vcpu);
 
+int kvm_riscv_setup_default_irq_routing(struct kvm *kvm, u32 lines);
+
 void __kvm_riscv_unpriv_trap(void);
 
 unsigned long kvm_riscv_vcpu_unpriv_read(struct kvm_vcpu *vcpu,

arch/riscv/include/asm/kvm_vcpu_sbi.h

@@ -14,9 +14,15 @@
 #define KVM_SBI_VERSION_MAJOR 1
 #define KVM_SBI_VERSION_MINOR 0
 
+enum kvm_riscv_sbi_ext_status {
+	KVM_RISCV_SBI_EXT_UNINITIALIZED,
+	KVM_RISCV_SBI_EXT_AVAILABLE,
+	KVM_RISCV_SBI_EXT_UNAVAILABLE,
+};
+
 struct kvm_vcpu_sbi_context {
 	int return_handled;
-	bool extension_disabled[KVM_RISCV_SBI_EXT_MAX];
+	enum kvm_riscv_sbi_ext_status ext_status[KVM_RISCV_SBI_EXT_MAX];
 };
 
 struct kvm_vcpu_sbi_return {
@@ -66,4 +72,7 @@ extern const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_hsm;
 extern const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_experimental;
 extern const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_vendor;
 
+#ifdef CONFIG_RISCV_PMU_SBI
+extern const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_pmu;
+#endif
 #endif /* __RISCV_KVM_VCPU_SBI_H__ */

arch/riscv/include/uapi/asm/kvm.h

@@ -15,6 +15,7 @@
 #include <asm/bitsperlong.h>
 #include <asm/ptrace.h>
 
+#define __KVM_HAVE_IRQ_LINE
 #define __KVM_HAVE_READONLY_MEM
 
 #define KVM_COALESCED_MMIO_PAGE_OFFSET 1
@@ -122,6 +123,7 @@ enum KVM_RISCV_ISA_EXT_ID {
 	KVM_RISCV_ISA_EXT_ZBB,
 	KVM_RISCV_ISA_EXT_SSAIA,
 	KVM_RISCV_ISA_EXT_V,
+	KVM_RISCV_ISA_EXT_SVNAPOT,
 	KVM_RISCV_ISA_EXT_MAX,
 };
 
@@ -211,6 +213,77 @@ enum KVM_RISCV_SBI_EXT_ID {
 #define KVM_REG_RISCV_VECTOR_REG(n)	\
 		((n) + sizeof(struct __riscv_v_ext_state) / sizeof(unsigned long))
 
+/* Device Control API: RISC-V AIA */
+#define KVM_DEV_RISCV_APLIC_ALIGN		0x1000
+#define KVM_DEV_RISCV_APLIC_SIZE		0x4000
+#define KVM_DEV_RISCV_APLIC_MAX_HARTS		0x4000
+#define KVM_DEV_RISCV_IMSIC_ALIGN		0x1000
+#define KVM_DEV_RISCV_IMSIC_SIZE		0x1000
+
+#define KVM_DEV_RISCV_AIA_GRP_CONFIG		0
+#define KVM_DEV_RISCV_AIA_CONFIG_MODE		0
+#define KVM_DEV_RISCV_AIA_CONFIG_IDS		1
+#define KVM_DEV_RISCV_AIA_CONFIG_SRCS		2
+#define KVM_DEV_RISCV_AIA_CONFIG_GROUP_BITS	3
+#define KVM_DEV_RISCV_AIA_CONFIG_GROUP_SHIFT	4
+#define KVM_DEV_RISCV_AIA_CONFIG_HART_BITS	5
+#define KVM_DEV_RISCV_AIA_CONFIG_GUEST_BITS	6
+
+/*
+ * Modes of RISC-V AIA device:
+ * 1) EMUL (aka Emulation): Trap-n-emulate IMSIC
+ * 2) HWACCEL (aka HW Acceleration): Virtualize IMSIC using IMSIC guest files
+ * 3) AUTO (aka Automatic): Virtualize IMSIC using IMSIC guest files whenever
+ *    available otherwise fallback to trap-n-emulation
+ */
+#define KVM_DEV_RISCV_AIA_MODE_EMUL		0
+#define KVM_DEV_RISCV_AIA_MODE_HWACCEL		1
+#define KVM_DEV_RISCV_AIA_MODE_AUTO		2
+
+#define KVM_DEV_RISCV_AIA_IDS_MIN		63
+#define KVM_DEV_RISCV_AIA_IDS_MAX		2048
+#define KVM_DEV_RISCV_AIA_SRCS_MAX		1024
+#define KVM_DEV_RISCV_AIA_GROUP_BITS_MAX	8
+#define KVM_DEV_RISCV_AIA_GROUP_SHIFT_MIN	24
+#define KVM_DEV_RISCV_AIA_GROUP_SHIFT_MAX	56
+#define KVM_DEV_RISCV_AIA_HART_BITS_MAX		16
+#define KVM_DEV_RISCV_AIA_GUEST_BITS_MAX	8
+
+#define KVM_DEV_RISCV_AIA_GRP_ADDR		1
+#define KVM_DEV_RISCV_AIA_ADDR_APLIC		0
+#define KVM_DEV_RISCV_AIA_ADDR_IMSIC(__vcpu)	(1 + (__vcpu))
+#define KVM_DEV_RISCV_AIA_ADDR_MAX		\
+		(1 + KVM_DEV_RISCV_APLIC_MAX_HARTS)
+
+#define KVM_DEV_RISCV_AIA_GRP_CTRL		2
+#define KVM_DEV_RISCV_AIA_CTRL_INIT		0
+
+/*
+ * The device attribute type contains the memory mapped offset of the
+ * APLIC register (range 0x0000-0x3FFF) and it must be 4-byte aligned.
+ */
+#define KVM_DEV_RISCV_AIA_GRP_APLIC		3
+
+/*
+ * The lower 12-bits of the device attribute type contains the iselect
+ * value of the IMSIC register (range 0x70-0xFF) whereas the higher order
+ * bits contains the VCPU id.
+ */
+#define KVM_DEV_RISCV_AIA_GRP_IMSIC		4
+#define KVM_DEV_RISCV_AIA_IMSIC_ISEL_BITS	12
+#define KVM_DEV_RISCV_AIA_IMSIC_ISEL_MASK	\
+		((1U << KVM_DEV_RISCV_AIA_IMSIC_ISEL_BITS) - 1)
+#define KVM_DEV_RISCV_AIA_IMSIC_MKATTR(__vcpu, __isel)	\
+		(((__vcpu) << KVM_DEV_RISCV_AIA_IMSIC_ISEL_BITS) | \
+		 ((__isel) & KVM_DEV_RISCV_AIA_IMSIC_ISEL_MASK))
+#define KVM_DEV_RISCV_AIA_IMSIC_GET_ISEL(__attr)	\
+		((__attr) & KVM_DEV_RISCV_AIA_IMSIC_ISEL_MASK)
+#define KVM_DEV_RISCV_AIA_IMSIC_GET_VCPU(__attr)	\
+		((__attr) >> KVM_DEV_RISCV_AIA_IMSIC_ISEL_BITS)
+
+/* One single KVM irqchip, ie. the AIA */
+#define KVM_NR_IRQCHIPS			1
+
 #endif
 
 #endif /* __LINUX_KVM_RISCV_H */

arch/riscv/kvm/Kconfig

@@ -21,6 +21,10 @@ config KVM
 	tristate "Kernel-based Virtual Machine (KVM) support (EXPERIMENTAL)"
 	depends on RISCV_SBI && MMU
 	select HAVE_KVM_EVENTFD
+	select HAVE_KVM_IRQCHIP
+	select HAVE_KVM_IRQFD
+	select HAVE_KVM_IRQ_ROUTING
+	select HAVE_KVM_MSI
 	select HAVE_KVM_VCPU_ASYNC_IOCTL
 	select KVM_GENERIC_DIRTYLOG_READ_PROTECT
 	select KVM_GENERIC_HARDWARE_ENABLING

arch/riscv/kvm/Makefile

@@ -28,3 +28,6 @@ kvm-y += vcpu_sbi_hsm.o
 kvm-y += vcpu_timer.o
 kvm-$(CONFIG_RISCV_PMU_SBI) += vcpu_pmu.o vcpu_sbi_pmu.o
 kvm-y += aia.o
+kvm-y += aia_device.o
+kvm-y += aia_aplic.o
+kvm-y += aia_imsic.o

arch/riscv/kvm/aia.c

@@ -8,11 +8,49 @@
  */
 
 #include <linux/kernel.h>
+#include <linux/bitops.h>
+#include <linux/irq.h>
+#include <linux/irqdomain.h>
 #include <linux/kvm_host.h>
+#include <linux/percpu.h>
+#include <linux/spinlock.h>
 #include <asm/hwcap.h>
+#include <asm/kvm_aia_imsic.h>
 
+struct aia_hgei_control {
+	raw_spinlock_t lock;
+	unsigned long free_bitmap;
+	struct kvm_vcpu *owners[BITS_PER_LONG];
+};
+static DEFINE_PER_CPU(struct aia_hgei_control, aia_hgei);
+static int hgei_parent_irq;
+
+unsigned int kvm_riscv_aia_nr_hgei;
+unsigned int kvm_riscv_aia_max_ids;
 DEFINE_STATIC_KEY_FALSE(kvm_riscv_aia_available);
 
+static int aia_find_hgei(struct kvm_vcpu *owner)
+{
+	int i, hgei;
+	unsigned long flags;
+	struct aia_hgei_control *hgctrl = get_cpu_ptr(&aia_hgei);
+
+	raw_spin_lock_irqsave(&hgctrl->lock, flags);
+
+	hgei = -1;
+	for (i = 1; i <= kvm_riscv_aia_nr_hgei; i++) {
+		if (hgctrl->owners[i] == owner) {
+			hgei = i;
+			break;
+		}
+	}
+
+	raw_spin_unlock_irqrestore(&hgctrl->lock, flags);
+
+	put_cpu_ptr(&aia_hgei);
+	return hgei;
+}
+
 static void aia_set_hvictl(bool ext_irq_pending)
 {
 	unsigned long hvictl;
@@ -56,6 +94,7 @@ void kvm_riscv_vcpu_aia_sync_interrupts(struct kvm_vcpu *vcpu)
 
 bool kvm_riscv_vcpu_aia_has_interrupts(struct kvm_vcpu *vcpu, u64 mask)
 {
+	int hgei;
 	unsigned long seip;
 
 	if (!kvm_riscv_aia_available())
@@ -74,6 +113,10 @@ bool kvm_riscv_vcpu_aia_has_interrupts(struct kvm_vcpu *vcpu, u64 mask)
 	if (!kvm_riscv_aia_initialized(vcpu->kvm) || !seip)
 		return false;
 
+	hgei = aia_find_hgei(vcpu);
+	if (hgei > 0)
+		return !!(csr_read(CSR_HGEIP) & BIT(hgei));
+
 	return false;
 }
 
@@ -323,8 +366,6 @@ static int aia_rmw_iprio(struct kvm_vcpu *vcpu, unsigned int isel,
 	return KVM_INSN_CONTINUE_NEXT_SEPC;
 }
 
-#define IMSIC_FIRST	0x70
-#define IMSIC_LAST	0xff
 int kvm_riscv_vcpu_aia_rmw_ireg(struct kvm_vcpu *vcpu, unsigned int csr_num,
 				unsigned long *val, unsigned long new_val,
 				unsigned long wr_mask)
@@ -348,6 +389,143 @@ int kvm_riscv_vcpu_aia_rmw_ireg(struct kvm_vcpu *vcpu, unsigned int csr_num,
 	return KVM_INSN_EXIT_TO_USER_SPACE;
 }
 
+int kvm_riscv_aia_alloc_hgei(int cpu, struct kvm_vcpu *owner,
+			     void __iomem **hgei_va, phys_addr_t *hgei_pa)
+{
+	int ret = -ENOENT;
+	unsigned long flags;
+	struct aia_hgei_control *hgctrl = per_cpu_ptr(&aia_hgei, cpu);
+
+	if (!kvm_riscv_aia_available() || !hgctrl)
+		return -ENODEV;
+
+	raw_spin_lock_irqsave(&hgctrl->lock, flags);
+
+	if (hgctrl->free_bitmap) {
+		ret = __ffs(hgctrl->free_bitmap);
+		hgctrl->free_bitmap &= ~BIT(ret);
+		hgctrl->owners[ret] = owner;
+	}
+
+	raw_spin_unlock_irqrestore(&hgctrl->lock, flags);
+
+	/* TODO: To be updated later by AIA IMSIC HW guest file support */
+	if (hgei_va)
+		*hgei_va = NULL;
+	if (hgei_pa)
+		*hgei_pa = 0;
+
+	return ret;
+}
+
+void kvm_riscv_aia_free_hgei(int cpu, int hgei)
+{
+	unsigned long flags;
+	struct aia_hgei_control *hgctrl = per_cpu_ptr(&aia_hgei, cpu);
+
+	if (!kvm_riscv_aia_available() || !hgctrl)
+		return;
+
+	raw_spin_lock_irqsave(&hgctrl->lock, flags);
+
+	if (hgei > 0 && hgei <= kvm_riscv_aia_nr_hgei) {
+		if (!(hgctrl->free_bitmap & BIT(hgei))) {
+			hgctrl->free_bitmap |= BIT(hgei);
+			hgctrl->owners[hgei] = NULL;
+		}
+	}
+
+	raw_spin_unlock_irqrestore(&hgctrl->lock, flags);
+}
+
+void kvm_riscv_aia_wakeon_hgei(struct kvm_vcpu *owner, bool enable)
+{
+	int hgei;
+
+	if (!kvm_riscv_aia_available())
+		return;
+
+	hgei = aia_find_hgei(owner);
+	if (hgei > 0) {
+		if (enable)
+			csr_set(CSR_HGEIE, BIT(hgei));
+		else
+			csr_clear(CSR_HGEIE, BIT(hgei));
+	}
+}
+
+static irqreturn_t hgei_interrupt(int irq, void *dev_id)
+{
+	int i;
+	unsigned long hgei_mask, flags;
+	struct aia_hgei_control *hgctrl = get_cpu_ptr(&aia_hgei);
+
+	hgei_mask = csr_read(CSR_HGEIP) & csr_read(CSR_HGEIE);
+	csr_clear(CSR_HGEIE, hgei_mask);
+
+	raw_spin_lock_irqsave(&hgctrl->lock, flags);
+
+	for_each_set_bit(i, &hgei_mask, BITS_PER_LONG) {
+		if (hgctrl->owners[i])
+			kvm_vcpu_kick(hgctrl->owners[i]);
+	}
+
+	raw_spin_unlock_irqrestore(&hgctrl->lock, flags);
+
+	put_cpu_ptr(&aia_hgei);
+	return IRQ_HANDLED;
+}
+
+static int aia_hgei_init(void)
+{
+	int cpu, rc;
+	struct irq_domain *domain;
+	struct aia_hgei_control *hgctrl;
+
+	/* Initialize per-CPU guest external interrupt line management */
+	for_each_possible_cpu(cpu) {
+		hgctrl = per_cpu_ptr(&aia_hgei, cpu);
+		raw_spin_lock_init(&hgctrl->lock);
+		if (kvm_riscv_aia_nr_hgei) {
+			hgctrl->free_bitmap =
+				BIT(kvm_riscv_aia_nr_hgei + 1) - 1;
+			hgctrl->free_bitmap &= ~BIT(0);
+		} else
+			hgctrl->free_bitmap = 0;
+	}
+
+	/* Find INTC irq domain */
+	domain = irq_find_matching_fwnode(riscv_get_intc_hwnode(),
+					  DOMAIN_BUS_ANY);
+	if (!domain) {
+		kvm_err("unable to find INTC domain\n");
+		return -ENOENT;
+	}
+
+	/* Map per-CPU SGEI interrupt from INTC domain */
+	hgei_parent_irq = irq_create_mapping(domain, IRQ_S_GEXT);
+	if (!hgei_parent_irq) {
+		kvm_err("unable to map SGEI IRQ\n");
+		return -ENOMEM;
+	}
+
+	/* Request per-CPU SGEI interrupt */
+	rc = request_percpu_irq(hgei_parent_irq, hgei_interrupt,
+				"riscv-kvm", &aia_hgei);
+	if (rc) {
+		kvm_err("failed to request SGEI IRQ\n");
+		return rc;
+	}
+
+	return 0;
+}
+
+static void aia_hgei_exit(void)
+{
+	/* Free per-CPU SGEI interrupt */
+	free_percpu_irq(hgei_parent_irq, &aia_hgei);
+}
+
 void kvm_riscv_aia_enable(void)
 {
 	if (!kvm_riscv_aia_available())
@@ -362,21 +540,105 @@ void kvm_riscv_aia_enable(void)
 	csr_write(CSR_HVIPRIO1H, 0x0);
 	csr_write(CSR_HVIPRIO2H, 0x0);
 #endif
+
+	/* Enable per-CPU SGEI interrupt */
+	enable_percpu_irq(hgei_parent_irq,
+			  irq_get_trigger_type(hgei_parent_irq));
+	csr_set(CSR_HIE, BIT(IRQ_S_GEXT));
 }
 
 void kvm_riscv_aia_disable(void)
 {
+	int i;
+	unsigned long flags;
+	struct kvm_vcpu *vcpu;
+	struct aia_hgei_control *hgctrl;
+
 	if (!kvm_riscv_aia_available())
 		return;
+	hgctrl = get_cpu_ptr(&aia_hgei);
+
+	/* Disable per-CPU SGEI interrupt */
+	csr_clear(CSR_HIE, BIT(IRQ_S_GEXT));
+	disable_percpu_irq(hgei_parent_irq);
 
 	aia_set_hvictl(false);
+
+	raw_spin_lock_irqsave(&hgctrl->lock, flags);
+
+	for (i = 0; i <= kvm_riscv_aia_nr_hgei; i++) {
+		vcpu = hgctrl->owners[i];
+		if (!vcpu)
+			continue;
+
+		/*
+		 * We release hgctrl->lock before notifying IMSIC
+		 * so that we don't have lock ordering issues.
+		 */
+		raw_spin_unlock_irqrestore(&hgctrl->lock, flags);
+
+		/* Notify IMSIC */
+		kvm_riscv_vcpu_aia_imsic_release(vcpu);
+
+		/*
+		 * Wakeup VCPU if it was blocked so that it can
+		 * run on other HARTs
+		 */
+		if (csr_read(CSR_HGEIE) & BIT(i)) {
+			csr_clear(CSR_HGEIE, BIT(i));
+			kvm_vcpu_kick(vcpu);
+		}
+
+		raw_spin_lock_irqsave(&hgctrl->lock, flags);
+	}
+
+	raw_spin_unlock_irqrestore(&hgctrl->lock, flags);
+
+	put_cpu_ptr(&aia_hgei);
 }
 
 int kvm_riscv_aia_init(void)
 {
+	int rc;
+
 	if (!riscv_isa_extension_available(NULL, SxAIA))
 		return -ENODEV;
 
+	/* Figure-out number of bits in HGEIE */
+	csr_write(CSR_HGEIE, -1UL);
+	kvm_riscv_aia_nr_hgei = fls_long(csr_read(CSR_HGEIE));
+	csr_write(CSR_HGEIE, 0);
+	if (kvm_riscv_aia_nr_hgei)
+		kvm_riscv_aia_nr_hgei--;
+
+	/*
+	 * Number of usable HGEI lines should be minimum of per-HART
+	 * IMSIC guest files and number of bits in HGEIE
+	 *
+	 * TODO: To be updated later by AIA IMSIC HW guest file support
+	 */
+	kvm_riscv_aia_nr_hgei = 0;
+
+	/*
+	 * Find number of guest MSI IDs
+	 *
+	 * TODO: To be updated later by AIA IMSIC HW guest file support
+	 */
+	kvm_riscv_aia_max_ids = IMSIC_MAX_ID;
+
+	/* Initialize guest external interrupt line management */
+	rc = aia_hgei_init();
+	if (rc)
+		return rc;
+
+	/* Register device operations */
+	rc = kvm_register_device_ops(&kvm_riscv_aia_device_ops,
+				     KVM_DEV_TYPE_RISCV_AIA);
+	if (rc) {
+		aia_hgei_exit();
+		return rc;
+	}
+
 	/* Enable KVM AIA support */
 	static_branch_enable(&kvm_riscv_aia_available);
 
@@ -385,4 +647,12 @@ int kvm_riscv_aia_init(void)
 
 void kvm_riscv_aia_exit(void)
 {
+	if (!kvm_riscv_aia_available())
+		return;
+
+	/* Unregister device operations */
+	kvm_unregister_device_ops(KVM_DEV_TYPE_RISCV_AIA);
+
+	/* Cleanup the HGEI state */
+	aia_hgei_exit();
 }

arch/riscv/kvm/aia_aplic.c

@@ -0,0 +1,619 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (C) 2022 Ventana Micro Systems Inc.
+ *
+ * Authors:
+ *	Anup Patel <apatel@ventanamicro.com>
+ */
+
+#include <linux/kvm_host.h>
+#include <linux/math.h>
+#include <linux/spinlock.h>
+#include <linux/swab.h>
+#include <kvm/iodev.h>
+#include <asm/kvm_aia_aplic.h>
+
+struct aplic_irq {
+	raw_spinlock_t lock;
+	u32 sourcecfg;
+	u32 state;
+#define APLIC_IRQ_STATE_PENDING		BIT(0)
+#define APLIC_IRQ_STATE_ENABLED		BIT(1)
+#define APLIC_IRQ_STATE_ENPEND		(APLIC_IRQ_STATE_PENDING | \
+					 APLIC_IRQ_STATE_ENABLED)
+#define APLIC_IRQ_STATE_INPUT		BIT(8)
+	u32 target;
+};
+
+struct aplic {
+	struct kvm_io_device iodev;
+
+	u32 domaincfg;
+	u32 genmsi;
+
+	u32 nr_irqs;
+	u32 nr_words;
+	struct aplic_irq *irqs;
+};
+
+static u32 aplic_read_sourcecfg(struct aplic *aplic, u32 irq)
+{
+	u32 ret;
+	unsigned long flags;
+	struct aplic_irq *irqd;
+
+	if (!irq || aplic->nr_irqs <= irq)
+		return 0;
+	irqd = &aplic->irqs[irq];
+
+	raw_spin_lock_irqsave(&irqd->lock, flags);
+	ret = irqd->sourcecfg;
+	raw_spin_unlock_irqrestore(&irqd->lock, flags);
+
+	return ret;
+}
+
+static void aplic_write_sourcecfg(struct aplic *aplic, u32 irq, u32 val)
+{
+	unsigned long flags;
+	struct aplic_irq *irqd;
+
+	if (!irq || aplic->nr_irqs <= irq)
+		return;
+	irqd = &aplic->irqs[irq];
+
+	if (val & APLIC_SOURCECFG_D)
+		val = 0;
+	else
+		val &= APLIC_SOURCECFG_SM_MASK;
+
+	raw_spin_lock_irqsave(&irqd->lock, flags);
+	irqd->sourcecfg = val;
+	raw_spin_unlock_irqrestore(&irqd->lock, flags);
+}
+
+static u32 aplic_read_target(struct aplic *aplic, u32 irq)
+{
+	u32 ret;
+	unsigned long flags;
+	struct aplic_irq *irqd;
+
+	if (!irq || aplic->nr_irqs <= irq)
+		return 0;
+	irqd = &aplic->irqs[irq];
+
+	raw_spin_lock_irqsave(&irqd->lock, flags);
+	ret = irqd->target;
+	raw_spin_unlock_irqrestore(&irqd->lock, flags);
+
+	return ret;
+}
+
+static void aplic_write_target(struct aplic *aplic, u32 irq, u32 val)
+{
+	unsigned long flags;
+	struct aplic_irq *irqd;
+
+	if (!irq || aplic->nr_irqs <= irq)
+		return;
+	irqd = &aplic->irqs[irq];
+
+	val &= APLIC_TARGET_EIID_MASK |
+	       (APLIC_TARGET_HART_IDX_MASK << APLIC_TARGET_HART_IDX_SHIFT) |
+	       (APLIC_TARGET_GUEST_IDX_MASK << APLIC_TARGET_GUEST_IDX_SHIFT);
+
+	raw_spin_lock_irqsave(&irqd->lock, flags);
+	irqd->target = val;
+	raw_spin_unlock_irqrestore(&irqd->lock, flags);
+}
+
+static bool aplic_read_pending(struct aplic *aplic, u32 irq)
+{
+	bool ret;
+	unsigned long flags;
+	struct aplic_irq *irqd;
+
+	if (!irq || aplic->nr_irqs <= irq)
+		return false;
+	irqd = &aplic->irqs[irq];
+
+	raw_spin_lock_irqsave(&irqd->lock, flags);
+	ret = (irqd->state & APLIC_IRQ_STATE_PENDING) ? true : false;
+	raw_spin_unlock_irqrestore(&irqd->lock, flags);
+
+	return ret;
+}
+
+static void aplic_write_pending(struct aplic *aplic, u32 irq, bool pending)
+{
+	unsigned long flags, sm;
+	struct aplic_irq *irqd;
+
+	if (!irq || aplic->nr_irqs <= irq)
+		return;
+	irqd = &aplic->irqs[irq];
+
+	raw_spin_lock_irqsave(&irqd->lock, flags);
+
+	sm = irqd->sourcecfg & APLIC_SOURCECFG_SM_MASK;
+	if (!pending &&
+	    ((sm == APLIC_SOURCECFG_SM_LEVEL_HIGH) ||
+	     (sm == APLIC_SOURCECFG_SM_LEVEL_LOW)))
+		goto skip_write_pending;
+
+	if (pending)
+		irqd->state |= APLIC_IRQ_STATE_PENDING;
+	else
+		irqd->state &= ~APLIC_IRQ_STATE_PENDING;
+
+skip_write_pending:
+	raw_spin_unlock_irqrestore(&irqd->lock, flags);
+}
+
+static bool aplic_read_enabled(struct aplic *aplic, u32 irq)
+{
+	bool ret;
+	unsigned long flags;
+	struct aplic_irq *irqd;
+
+	if (!irq || aplic->nr_irqs <= irq)
+		return false;
+	irqd = &aplic->irqs[irq];
+
+	raw_spin_lock_irqsave(&irqd->lock, flags);
+	ret = (irqd->state & APLIC_IRQ_STATE_ENABLED) ? true : false;
+	raw_spin_unlock_irqrestore(&irqd->lock, flags);
+
+	return ret;
+}
+
+static void aplic_write_enabled(struct aplic *aplic, u32 irq, bool enabled)
+{
+	unsigned long flags;
+	struct aplic_irq *irqd;
+
+	if (!irq || aplic->nr_irqs <= irq)
+		return;
+	irqd = &aplic->irqs[irq];
+
+	raw_spin_lock_irqsave(&irqd->lock, flags);
+	if (enabled)
+		irqd->state |= APLIC_IRQ_STATE_ENABLED;
+	else
+		irqd->state &= ~APLIC_IRQ_STATE_ENABLED;
+	raw_spin_unlock_irqrestore(&irqd->lock, flags);
+}
+
+static bool aplic_read_input(struct aplic *aplic, u32 irq)
+{
+	bool ret;
+	unsigned long flags;
+	struct aplic_irq *irqd;
+
+	if (!irq || aplic->nr_irqs <= irq)
+		return false;
+	irqd = &aplic->irqs[irq];
+
+	raw_spin_lock_irqsave(&irqd->lock, flags);
+	ret = (irqd->state & APLIC_IRQ_STATE_INPUT) ? true : false;
+	raw_spin_unlock_irqrestore(&irqd->lock, flags);
+
+	return ret;
+}
+
+static void aplic_inject_msi(struct kvm *kvm, u32 irq, u32 target)
+{
+	u32 hart_idx, guest_idx, eiid;
+
+	hart_idx = target >> APLIC_TARGET_HART_IDX_SHIFT;
+	hart_idx &= APLIC_TARGET_HART_IDX_MASK;
+	guest_idx = target >> APLIC_TARGET_GUEST_IDX_SHIFT;
+	guest_idx &= APLIC_TARGET_GUEST_IDX_MASK;
+	eiid = target & APLIC_TARGET_EIID_MASK;
+	kvm_riscv_aia_inject_msi_by_id(kvm, hart_idx, guest_idx, eiid);
+}
+
+static void aplic_update_irq_range(struct kvm *kvm, u32 first, u32 last)
+{
+	bool inject;
+	u32 irq, target;
+	unsigned long flags;
+	struct aplic_irq *irqd;
+	struct aplic *aplic = kvm->arch.aia.aplic_state;
+
+	if (!(aplic->domaincfg & APLIC_DOMAINCFG_IE))
+		return;
+
+	for (irq = first; irq <= last; irq++) {
+		if (!irq || aplic->nr_irqs <= irq)
+			continue;
+		irqd = &aplic->irqs[irq];
+
+		raw_spin_lock_irqsave(&irqd->lock, flags);
+
+		inject = false;
+		target = irqd->target;
+		if ((irqd->state & APLIC_IRQ_STATE_ENPEND) ==
+		    APLIC_IRQ_STATE_ENPEND) {
+			irqd->state &= ~APLIC_IRQ_STATE_PENDING;
+			inject = true;
+		}
+
+		raw_spin_unlock_irqrestore(&irqd->lock, flags);
+
+		if (inject)
+			aplic_inject_msi(kvm, irq, target);
+	}
+}
+
+int kvm_riscv_aia_aplic_inject(struct kvm *kvm, u32 source, bool level)
+{
+	u32 target;
+	bool inject = false, ie;
+	unsigned long flags;
+	struct aplic_irq *irqd;
+	struct aplic *aplic = kvm->arch.aia.aplic_state;
+
+	if (!aplic || !source || (aplic->nr_irqs <= source))
+		return -ENODEV;
+	irqd = &aplic->irqs[source];
+	ie = (aplic->domaincfg & APLIC_DOMAINCFG_IE) ? true : false;
+
+	raw_spin_lock_irqsave(&irqd->lock, flags);
+
+	if (irqd->sourcecfg & APLIC_SOURCECFG_D)
+		goto skip_unlock;
+
+	switch (irqd->sourcecfg & APLIC_SOURCECFG_SM_MASK) {
+	case APLIC_SOURCECFG_SM_EDGE_RISE:
+		if (level && !(irqd->state & APLIC_IRQ_STATE_INPUT) &&
+		    !(irqd->state & APLIC_IRQ_STATE_PENDING))
+			irqd->state |= APLIC_IRQ_STATE_PENDING;
+		break;
+	case APLIC_SOURCECFG_SM_EDGE_FALL:
+		if (!level && (irqd->state & APLIC_IRQ_STATE_INPUT) &&
+		    !(irqd->state & APLIC_IRQ_STATE_PENDING))
+			irqd->state |= APLIC_IRQ_STATE_PENDING;
+		break;
+	case APLIC_SOURCECFG_SM_LEVEL_HIGH:
+		if (level && !(irqd->state & APLIC_IRQ_STATE_PENDING))
+			irqd->state |= APLIC_IRQ_STATE_PENDING;
+		break;
+	case APLIC_SOURCECFG_SM_LEVEL_LOW:
+		if (!level && !(irqd->state & APLIC_IRQ_STATE_PENDING))
+			irqd->state |= APLIC_IRQ_STATE_PENDING;
+		break;
+	}
+
+	if (level)
+		irqd->state |= APLIC_IRQ_STATE_INPUT;
+	else
+		irqd->state &= ~APLIC_IRQ_STATE_INPUT;
+
+	target = irqd->target;
+	if (ie && ((irqd->state & APLIC_IRQ_STATE_ENPEND) ==
+		   APLIC_IRQ_STATE_ENPEND)) {
+		irqd->state &= ~APLIC_IRQ_STATE_PENDING;
+		inject = true;
+	}
+
+skip_unlock:
+	raw_spin_unlock_irqrestore(&irqd->lock, flags);
+
+	if (inject)
+		aplic_inject_msi(kvm, source, target);
+
+	return 0;
+}
+
+static u32 aplic_read_input_word(struct aplic *aplic, u32 word)
+{
+	u32 i, ret = 0;
+
+	for (i = 0; i < 32; i++)
+		ret |= aplic_read_input(aplic, word * 32 + i) ? BIT(i) : 0;
+
+	return ret;
+}
+
+static u32 aplic_read_pending_word(struct aplic *aplic, u32 word)
+{
+	u32 i, ret = 0;
+
+	for (i = 0; i < 32; i++)
+		ret |= aplic_read_pending(aplic, word * 32 + i) ? BIT(i) : 0;
+
+	return ret;
+}
+
+static void aplic_write_pending_word(struct aplic *aplic, u32 word,
+				     u32 val, bool pending)
+{
+	u32 i;
+
+	for (i = 0; i < 32; i++) {
+		if (val & BIT(i))
+			aplic_write_pending(aplic, word * 32 + i, pending);
+	}
+}
+
+static u32 aplic_read_enabled_word(struct aplic *aplic, u32 word)
+{
+	u32 i, ret = 0;
+
+	for (i = 0; i < 32; i++)
+		ret |= aplic_read_enabled(aplic, word * 32 + i) ? BIT(i) : 0;
+
+	return ret;
+}
+
+static void aplic_write_enabled_word(struct aplic *aplic, u32 word,
+				     u32 val, bool enabled)
+{
+	u32 i;
+
+	for (i = 0; i < 32; i++) {
+		if (val & BIT(i))
+			aplic_write_enabled(aplic, word * 32 + i, enabled);
+	}
+}
+
+static int aplic_mmio_read_offset(struct kvm *kvm, gpa_t off, u32 *val32)
+{
+	u32 i;
+	struct aplic *aplic = kvm->arch.aia.aplic_state;
+
+	if ((off & 0x3) != 0)
+		return -EOPNOTSUPP;
+
+	if (off == APLIC_DOMAINCFG) {
+		*val32 = APLIC_DOMAINCFG_RDONLY |
+			 aplic->domaincfg | APLIC_DOMAINCFG_DM;
+	} else if ((off >= APLIC_SOURCECFG_BASE) &&
+		 (off < (APLIC_SOURCECFG_BASE + (aplic->nr_irqs - 1) * 4))) {
+		i = ((off - APLIC_SOURCECFG_BASE) >> 2) + 1;
+		*val32 = aplic_read_sourcecfg(aplic, i);
+	} else if ((off >= APLIC_SETIP_BASE) &&
+		   (off < (APLIC_SETIP_BASE + aplic->nr_words * 4))) {
+		i = (off - APLIC_SETIP_BASE) >> 2;
+		*val32 = aplic_read_pending_word(aplic, i);
+	} else if (off == APLIC_SETIPNUM) {
+		*val32 = 0;
+	} else if ((off >= APLIC_CLRIP_BASE) &&
+		   (off < (APLIC_CLRIP_BASE + aplic->nr_words * 4))) {
+		i = (off - APLIC_CLRIP_BASE) >> 2;
+		*val32 = aplic_read_input_word(aplic, i);
+	} else if (off == APLIC_CLRIPNUM) {
+		*val32 = 0;
+	} else if ((off >= APLIC_SETIE_BASE) &&
+		   (off < (APLIC_SETIE_BASE + aplic->nr_words * 4))) {
+		i = (off - APLIC_SETIE_BASE) >> 2;
+		*val32 = aplic_read_enabled_word(aplic, i);
+	} else if (off == APLIC_SETIENUM) {
+		*val32 = 0;
+	} else if ((off >= APLIC_CLRIE_BASE) &&
+		   (off < (APLIC_CLRIE_BASE + aplic->nr_words * 4))) {
+		*val32 = 0;
+	} else if (off == APLIC_CLRIENUM) {
+		*val32 = 0;
+	} else if (off == APLIC_SETIPNUM_LE) {
+		*val32 = 0;
+	} else if (off == APLIC_SETIPNUM_BE) {
+		*val32 = 0;
+	} else if (off == APLIC_GENMSI) {
+		*val32 = aplic->genmsi;
+	} else if ((off >= APLIC_TARGET_BASE) &&
+		   (off < (APLIC_TARGET_BASE + (aplic->nr_irqs - 1) * 4))) {
+		i = ((off - APLIC_TARGET_BASE) >> 2) + 1;
+		*val32 = aplic_read_target(aplic, i);
+	} else
+		return -ENODEV;
+
+	return 0;
+}
+
+static int aplic_mmio_read(struct kvm_vcpu *vcpu, struct kvm_io_device *dev,
+			   gpa_t addr, int len, void *val)
+{
+	if (len != 4)
+		return -EOPNOTSUPP;
+
+	return aplic_mmio_read_offset(vcpu->kvm,
+				      addr - vcpu->kvm->arch.aia.aplic_addr,
+				      val);
+}
+
+static int aplic_mmio_write_offset(struct kvm *kvm, gpa_t off, u32 val32)
+{
+	u32 i;
+	struct aplic *aplic = kvm->arch.aia.aplic_state;
+
+	if ((off & 0x3) != 0)
+		return -EOPNOTSUPP;
+
+	if (off == APLIC_DOMAINCFG) {
+		/* Only IE bit writeable */
+		aplic->domaincfg = val32 & APLIC_DOMAINCFG_IE;
+	} else if ((off >= APLIC_SOURCECFG_BASE) &&
+		 (off < (APLIC_SOURCECFG_BASE + (aplic->nr_irqs - 1) * 4))) {
+		i = ((off - APLIC_SOURCECFG_BASE) >> 2) + 1;
+		aplic_write_sourcecfg(aplic, i, val32);
+	} else if ((off >= APLIC_SETIP_BASE) &&
+		   (off < (APLIC_SETIP_BASE + aplic->nr_words * 4))) {
+		i = (off - APLIC_SETIP_BASE) >> 2;
+		aplic_write_pending_word(aplic, i, val32, true);
+	} else if (off == APLIC_SETIPNUM) {
+		aplic_write_pending(aplic, val32, true);
+	} else if ((off >= APLIC_CLRIP_BASE) &&
+		   (off < (APLIC_CLRIP_BASE + aplic->nr_words * 4))) {
+		i = (off - APLIC_CLRIP_BASE) >> 2;
+		aplic_write_pending_word(aplic, i, val32, false);
+	} else if (off == APLIC_CLRIPNUM) {
+		aplic_write_pending(aplic, val32, false);
+	} else if ((off >= APLIC_SETIE_BASE) &&
+		   (off < (APLIC_SETIE_BASE + aplic->nr_words * 4))) {
+		i = (off - APLIC_SETIE_BASE) >> 2;
+		aplic_write_enabled_word(aplic, i, val32, true);
+	} else if (off == APLIC_SETIENUM) {
+		aplic_write_enabled(aplic, val32, true);
+	} else if ((off >= APLIC_CLRIE_BASE) &&
+		   (off < (APLIC_CLRIE_BASE + aplic->nr_words * 4))) {
+		i = (off - APLIC_CLRIE_BASE) >> 2;
+		aplic_write_enabled_word(aplic, i, val32, false);
+	} else if (off == APLIC_CLRIENUM) {
+		aplic_write_enabled(aplic, val32, false);
+	} else if (off == APLIC_SETIPNUM_LE) {
+		aplic_write_pending(aplic, val32, true);
+	} else if (off == APLIC_SETIPNUM_BE) {
+		aplic_write_pending(aplic, __swab32(val32), true);
+	} else if (off == APLIC_GENMSI) {
+		aplic->genmsi = val32 & ~(APLIC_TARGET_GUEST_IDX_MASK <<
+					  APLIC_TARGET_GUEST_IDX_SHIFT);
+		kvm_riscv_aia_inject_msi_by_id(kvm,
+				val32 >> APLIC_TARGET_HART_IDX_SHIFT, 0,
+				val32 & APLIC_TARGET_EIID_MASK);
+	} else if ((off >= APLIC_TARGET_BASE) &&
+		   (off < (APLIC_TARGET_BASE + (aplic->nr_irqs - 1) * 4))) {
+		i = ((off - APLIC_TARGET_BASE) >> 2) + 1;
+		aplic_write_target(aplic, i, val32);
+	} else
+		return -ENODEV;
+
+	aplic_update_irq_range(kvm, 1, aplic->nr_irqs - 1);
+
+	return 0;
+}
+
+static int aplic_mmio_write(struct kvm_vcpu *vcpu, struct kvm_io_device *dev,
+			    gpa_t addr, int len, const void *val)
+{
+	if (len != 4)
+		return -EOPNOTSUPP;
+
+	return aplic_mmio_write_offset(vcpu->kvm,
+				       addr - vcpu->kvm->arch.aia.aplic_addr,
+				       *((const u32 *)val));
+}
+
+static struct kvm_io_device_ops aplic_iodoev_ops = {
+	.read = aplic_mmio_read,
+	.write = aplic_mmio_write,
+};
+
+int kvm_riscv_aia_aplic_set_attr(struct kvm *kvm, unsigned long type, u32 v)
+{
+	int rc;
+
+	if (!kvm->arch.aia.aplic_state)
+		return -ENODEV;
+
+	rc = aplic_mmio_write_offset(kvm, type, v);
+	if (rc)
+		return rc;
+
+	return 0;
+}
+
+int kvm_riscv_aia_aplic_get_attr(struct kvm *kvm, unsigned long type, u32 *v)
+{
+	int rc;
+
+	if (!kvm->arch.aia.aplic_state)
+		return -ENODEV;
+
+	rc = aplic_mmio_read_offset(kvm, type, v);
+	if (rc)
+		return rc;
+
+	return 0;
+}
+
+int kvm_riscv_aia_aplic_has_attr(struct kvm *kvm, unsigned long type)
+{
+	int rc;
+	u32 val;
+
+	if (!kvm->arch.aia.aplic_state)
+		return -ENODEV;
+
+	rc = aplic_mmio_read_offset(kvm, type, &val);
+	if (rc)
+		return rc;
+
+	return 0;
+}
+
+int kvm_riscv_aia_aplic_init(struct kvm *kvm)
+{
+	int i, ret = 0;
+	struct aplic *aplic;
+
+	/* Do nothing if we have zero sources */
+	if (!kvm->arch.aia.nr_sources)
+		return 0;
+
+	/* Allocate APLIC global state */
+	aplic = kzalloc(sizeof(*aplic), GFP_KERNEL);
+	if (!aplic)
+		return -ENOMEM;
+	kvm->arch.aia.aplic_state = aplic;
+
+	/* Setup APLIC IRQs */
+	aplic->nr_irqs = kvm->arch.aia.nr_sources + 1;
+	aplic->nr_words = DIV_ROUND_UP(aplic->nr_irqs, 32);
+	aplic->irqs = kcalloc(aplic->nr_irqs,
+			      sizeof(*aplic->irqs), GFP_KERNEL);
+	if (!aplic->irqs) {
+		ret = -ENOMEM;
+		goto fail_free_aplic;
+	}
+	for (i = 0; i < aplic->nr_irqs; i++)
+		raw_spin_lock_init(&aplic->irqs[i].lock);
+
+	/* Setup IO device */
+	kvm_iodevice_init(&aplic->iodev, &aplic_iodoev_ops);
+	mutex_lock(&kvm->slots_lock);
+	ret = kvm_io_bus_register_dev(kvm, KVM_MMIO_BUS,
+				      kvm->arch.aia.aplic_addr,
+				      KVM_DEV_RISCV_APLIC_SIZE,
+				      &aplic->iodev);
+	mutex_unlock(&kvm->slots_lock);
+	if (ret)
+		goto fail_free_aplic_irqs;
+
+	/* Setup default IRQ routing */
+	ret = kvm_riscv_setup_default_irq_routing(kvm, aplic->nr_irqs);
+	if (ret)
+		goto fail_unreg_iodev;
+
+	return 0;
+
+fail_unreg_iodev:
+	mutex_lock(&kvm->slots_lock);
+	kvm_io_bus_unregister_dev(kvm, KVM_MMIO_BUS, &aplic->iodev);
+	mutex_unlock(&kvm->slots_lock);
+fail_free_aplic_irqs:
+	kfree(aplic->irqs);
+fail_free_aplic:
+	kvm->arch.aia.aplic_state = NULL;
+	kfree(aplic);
+	return ret;
+}
+
+void kvm_riscv_aia_aplic_cleanup(struct kvm *kvm)
+{
+	struct aplic *aplic = kvm->arch.aia.aplic_state;
+
+	if (!aplic)
+		return;
+
+	mutex_lock(&kvm->slots_lock);
+	kvm_io_bus_unregister_dev(kvm, KVM_MMIO_BUS, &aplic->iodev);
+	mutex_unlock(&kvm->slots_lock);
+
+	kfree(aplic->irqs);
+
+	kvm->arch.aia.aplic_state = NULL;
+	kfree(aplic);
+}

arch/riscv/kvm/aia_device.c

@@ -0,0 +1,673 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2021 Western Digital Corporation or its affiliates.
+ * Copyright (C) 2022 Ventana Micro Systems Inc.
+ *
+ * Authors:
+ *	Anup Patel <apatel@ventanamicro.com>
+ */
+
+#include <linux/bits.h>
+#include <linux/kvm_host.h>
+#include <linux/uaccess.h>
+#include <asm/kvm_aia_imsic.h>
+
+static void unlock_vcpus(struct kvm *kvm, int vcpu_lock_idx)
+{
+	struct kvm_vcpu *tmp_vcpu;
+
+	for (; vcpu_lock_idx >= 0; vcpu_lock_idx--) {
+		tmp_vcpu = kvm_get_vcpu(kvm, vcpu_lock_idx);
+		mutex_unlock(&tmp_vcpu->mutex);
+	}
+}
+
+static void unlock_all_vcpus(struct kvm *kvm)
+{
+	unlock_vcpus(kvm, atomic_read(&kvm->online_vcpus) - 1);
+}
+
+static bool lock_all_vcpus(struct kvm *kvm)
+{
+	struct kvm_vcpu *tmp_vcpu;
+	unsigned long c;
+
+	kvm_for_each_vcpu(c, tmp_vcpu, kvm) {
+		if (!mutex_trylock(&tmp_vcpu->mutex)) {
+			unlock_vcpus(kvm, c - 1);
+			return false;
+		}
+	}
+
+	return true;
+}
+
+static int aia_create(struct kvm_device *dev, u32 type)
+{
+	int ret;
+	unsigned long i;
+	struct kvm *kvm = dev->kvm;
+	struct kvm_vcpu *vcpu;
+
+	if (irqchip_in_kernel(kvm))
+		return -EEXIST;
+
+	ret = -EBUSY;
+	if (!lock_all_vcpus(kvm))
+		return ret;
+
+	kvm_for_each_vcpu(i, vcpu, kvm) {
+		if (vcpu->arch.ran_atleast_once)
+			goto out_unlock;
+	}
+	ret = 0;
+
+	kvm->arch.aia.in_kernel = true;
+
+out_unlock:
+	unlock_all_vcpus(kvm);
+	return ret;
+}
+
+static void aia_destroy(struct kvm_device *dev)
+{
+	kfree(dev);
+}
+
+static int aia_config(struct kvm *kvm, unsigned long type,
+		      u32 *nr, bool write)
+{
+	struct kvm_aia *aia = &kvm->arch.aia;
+
+	/* Writes can only be done before irqchip is initialized */
+	if (write && kvm_riscv_aia_initialized(kvm))
+		return -EBUSY;
+
+	switch (type) {
+	case KVM_DEV_RISCV_AIA_CONFIG_MODE:
+		if (write) {
+			switch (*nr) {
+			case KVM_DEV_RISCV_AIA_MODE_EMUL:
+				break;
+			case KVM_DEV_RISCV_AIA_MODE_HWACCEL:
+			case KVM_DEV_RISCV_AIA_MODE_AUTO:
+				/*
+				 * HW Acceleration and Auto modes only
+				 * supported on host with non-zero guest
+				 * external interrupts (i.e. non-zero
+				 * VS-level IMSIC pages).
+				 */
+				if (!kvm_riscv_aia_nr_hgei)
+					return -EINVAL;
+				break;
+			default:
+				return -EINVAL;
+			}
+			aia->mode = *nr;
+		} else
+			*nr = aia->mode;
+		break;
+	case KVM_DEV_RISCV_AIA_CONFIG_IDS:
+		if (write) {
+			if ((*nr < KVM_DEV_RISCV_AIA_IDS_MIN) ||
+			    (*nr >= KVM_DEV_RISCV_AIA_IDS_MAX) ||
+			    ((*nr & KVM_DEV_RISCV_AIA_IDS_MIN) !=
+			     KVM_DEV_RISCV_AIA_IDS_MIN) ||
+			    (kvm_riscv_aia_max_ids <= *nr))
+				return -EINVAL;
+			aia->nr_ids = *nr;
+		} else
+			*nr = aia->nr_ids;
+		break;
+	case KVM_DEV_RISCV_AIA_CONFIG_SRCS:
+		if (write) {
+			if ((*nr >= KVM_DEV_RISCV_AIA_SRCS_MAX) ||
+			    (*nr >= kvm_riscv_aia_max_ids))
+				return -EINVAL;
+			aia->nr_sources = *nr;
+		} else
+			*nr = aia->nr_sources;
+		break;
+	case KVM_DEV_RISCV_AIA_CONFIG_GROUP_BITS:
+		if (write) {
+			if (*nr >= KVM_DEV_RISCV_AIA_GROUP_BITS_MAX)
+				return -EINVAL;
+			aia->nr_group_bits = *nr;
+		} else
+			*nr = aia->nr_group_bits;
+		break;
+	case KVM_DEV_RISCV_AIA_CONFIG_GROUP_SHIFT:
+		if (write) {
+			if ((*nr < KVM_DEV_RISCV_AIA_GROUP_SHIFT_MIN) ||
+			    (*nr >= KVM_DEV_RISCV_AIA_GROUP_SHIFT_MAX))
+				return -EINVAL;
+			aia->nr_group_shift = *nr;
+		} else
+			*nr = aia->nr_group_shift;
+		break;
+	case KVM_DEV_RISCV_AIA_CONFIG_HART_BITS:
+		if (write) {
+			if (*nr >= KVM_DEV_RISCV_AIA_HART_BITS_MAX)
+				return -EINVAL;
+			aia->nr_hart_bits = *nr;
+		} else
+			*nr = aia->nr_hart_bits;
+		break;
+	case KVM_DEV_RISCV_AIA_CONFIG_GUEST_BITS:
+		if (write) {
+			if (*nr >= KVM_DEV_RISCV_AIA_GUEST_BITS_MAX)
+				return -EINVAL;
+			aia->nr_guest_bits = *nr;
+		} else
+			*nr = aia->nr_guest_bits;
+		break;
+	default:
+		return -ENXIO;
+	}
+
+	return 0;
+}
+
+static int aia_aplic_addr(struct kvm *kvm, u64 *addr, bool write)
+{
+	struct kvm_aia *aia = &kvm->arch.aia;
+
+	if (write) {
+		/* Writes can only be done before irqchip is initialized */
+		if (kvm_riscv_aia_initialized(kvm))
+			return -EBUSY;
+
+		if (*addr & (KVM_DEV_RISCV_APLIC_ALIGN - 1))
+			return -EINVAL;
+
+		aia->aplic_addr = *addr;
+	} else
+		*addr = aia->aplic_addr;
+
+	return 0;
+}
+
+static int aia_imsic_addr(struct kvm *kvm, u64 *addr,
+			  unsigned long vcpu_idx, bool write)
+{
+	struct kvm_vcpu *vcpu;
+	struct kvm_vcpu_aia *vcpu_aia;
+
+	vcpu = kvm_get_vcpu(kvm, vcpu_idx);
+	if (!vcpu)
+		return -EINVAL;
+	vcpu_aia = &vcpu->arch.aia_context;
+
+	if (write) {
+		/* Writes can only be done before irqchip is initialized */
+		if (kvm_riscv_aia_initialized(kvm))
+			return -EBUSY;
+
+		if (*addr & (KVM_DEV_RISCV_IMSIC_ALIGN - 1))
+			return -EINVAL;
+	}
+
+	mutex_lock(&vcpu->mutex);
+	if (write)
+		vcpu_aia->imsic_addr = *addr;
+	else
+		*addr = vcpu_aia->imsic_addr;
+	mutex_unlock(&vcpu->mutex);
+
+	return 0;
+}
+
+static gpa_t aia_imsic_ppn(struct kvm_aia *aia, gpa_t addr)
+{
+	u32 h, l;
+	gpa_t mask = 0;
+
+	h = aia->nr_hart_bits + aia->nr_guest_bits +
+	    IMSIC_MMIO_PAGE_SHIFT - 1;
+	mask = GENMASK_ULL(h, 0);
+
+	if (aia->nr_group_bits) {
+		h = aia->nr_group_bits + aia->nr_group_shift - 1;
+		l = aia->nr_group_shift;
+		mask |= GENMASK_ULL(h, l);
+	}
+
+	return (addr & ~mask) >> IMSIC_MMIO_PAGE_SHIFT;
+}
+
+static u32 aia_imsic_hart_index(struct kvm_aia *aia, gpa_t addr)
+{
+	u32 hart, group = 0;
+
+	hart = (addr >> (aia->nr_guest_bits + IMSIC_MMIO_PAGE_SHIFT)) &
+		GENMASK_ULL(aia->nr_hart_bits - 1, 0);
+	if (aia->nr_group_bits)
+		group = (addr >> aia->nr_group_shift) &
+			GENMASK_ULL(aia->nr_group_bits - 1, 0);
+
+	return (group << aia->nr_hart_bits) | hart;
+}
+
+static int aia_init(struct kvm *kvm)
+{
+	int ret, i;
+	unsigned long idx;
+	struct kvm_vcpu *vcpu;
+	struct kvm_vcpu_aia *vaia;
+	struct kvm_aia *aia = &kvm->arch.aia;
+	gpa_t base_ppn = KVM_RISCV_AIA_UNDEF_ADDR;
+
+	/* Irqchip can be initialized only once */
+	if (kvm_riscv_aia_initialized(kvm))
+		return -EBUSY;
+
+	/* We might be in the middle of creating a VCPU? */
+	if (kvm->created_vcpus != atomic_read(&kvm->online_vcpus))
+		return -EBUSY;
+
+	/* Number of sources should be less than or equals number of IDs */
+	if (aia->nr_ids < aia->nr_sources)
+		return -EINVAL;
+
+	/* APLIC base is required for non-zero number of sources */
+	if (aia->nr_sources && aia->aplic_addr == KVM_RISCV_AIA_UNDEF_ADDR)
+		return -EINVAL;
+
+	/* Initialize APLIC */
+	ret = kvm_riscv_aia_aplic_init(kvm);
+	if (ret)
+		return ret;
+
+	/* Iterate over each VCPU */
+	kvm_for_each_vcpu(idx, vcpu, kvm) {
+		vaia = &vcpu->arch.aia_context;
+
+		/* IMSIC base is required */
+		if (vaia->imsic_addr == KVM_RISCV_AIA_UNDEF_ADDR) {
+			ret = -EINVAL;
+			goto fail_cleanup_imsics;
+		}
+
+		/* All IMSICs should have matching base PPN */
+		if (base_ppn == KVM_RISCV_AIA_UNDEF_ADDR)
+			base_ppn = aia_imsic_ppn(aia, vaia->imsic_addr);
+		if (base_ppn != aia_imsic_ppn(aia, vaia->imsic_addr)) {
+			ret = -EINVAL;
+			goto fail_cleanup_imsics;
+		}
+
+		/* Update HART index of the IMSIC based on IMSIC base */
+		vaia->hart_index = aia_imsic_hart_index(aia,
+							vaia->imsic_addr);
+
+		/* Initialize IMSIC for this VCPU */
+		ret = kvm_riscv_vcpu_aia_imsic_init(vcpu);
+		if (ret)
+			goto fail_cleanup_imsics;
+	}
+
+	/* Set the initialized flag */
+	kvm->arch.aia.initialized = true;
+
+	return 0;
+
+fail_cleanup_imsics:
+	for (i = idx - 1; i >= 0; i--) {
+		vcpu = kvm_get_vcpu(kvm, i);
+		if (!vcpu)
+			continue;
+		kvm_riscv_vcpu_aia_imsic_cleanup(vcpu);
+	}
+	kvm_riscv_aia_aplic_cleanup(kvm);
+	return ret;
+}
+
+static int aia_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
+{
+	u32 nr;
+	u64 addr;
+	int nr_vcpus, r = -ENXIO;
+	unsigned long v, type = (unsigned long)attr->attr;
+	void __user *uaddr = (void __user *)(long)attr->addr;
+
+	switch (attr->group) {
+	case KVM_DEV_RISCV_AIA_GRP_CONFIG:
+		if (copy_from_user(&nr, uaddr, sizeof(nr)))
+			return -EFAULT;
+
+		mutex_lock(&dev->kvm->lock);
+		r = aia_config(dev->kvm, type, &nr, true);
+		mutex_unlock(&dev->kvm->lock);
+
+		break;
+
+	case KVM_DEV_RISCV_AIA_GRP_ADDR:
+		if (copy_from_user(&addr, uaddr, sizeof(addr)))
+			return -EFAULT;
+
+		nr_vcpus = atomic_read(&dev->kvm->online_vcpus);
+		mutex_lock(&dev->kvm->lock);
+		if (type == KVM_DEV_RISCV_AIA_ADDR_APLIC)
+			r = aia_aplic_addr(dev->kvm, &addr, true);
+		else if (type < KVM_DEV_RISCV_AIA_ADDR_IMSIC(nr_vcpus))
+			r = aia_imsic_addr(dev->kvm, &addr,
+			    type - KVM_DEV_RISCV_AIA_ADDR_IMSIC(0), true);
+		mutex_unlock(&dev->kvm->lock);
+
+		break;
+
+	case KVM_DEV_RISCV_AIA_GRP_CTRL:
+		switch (type) {
+		case KVM_DEV_RISCV_AIA_CTRL_INIT:
+			mutex_lock(&dev->kvm->lock);
+			r = aia_init(dev->kvm);
+			mutex_unlock(&dev->kvm->lock);
+			break;
+		}
+
+		break;
+	case KVM_DEV_RISCV_AIA_GRP_APLIC:
+		if (copy_from_user(&nr, uaddr, sizeof(nr)))
+			return -EFAULT;
+
+		mutex_lock(&dev->kvm->lock);
+		r = kvm_riscv_aia_aplic_set_attr(dev->kvm, type, nr);
+		mutex_unlock(&dev->kvm->lock);
+
+		break;
+	case KVM_DEV_RISCV_AIA_GRP_IMSIC:
+		if (copy_from_user(&v, uaddr, sizeof(v)))
+			return -EFAULT;
+
+		mutex_lock(&dev->kvm->lock);
+		r = kvm_riscv_aia_imsic_rw_attr(dev->kvm, type, true, &v);
+		mutex_unlock(&dev->kvm->lock);
+
+		break;
+	}
+
+	return r;
+}
+
+static int aia_get_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
+{
+	u32 nr;
+	u64 addr;
+	int nr_vcpus, r = -ENXIO;
+	void __user *uaddr = (void __user *)(long)attr->addr;
+	unsigned long v, type = (unsigned long)attr->attr;
+
+	switch (attr->group) {
+	case KVM_DEV_RISCV_AIA_GRP_CONFIG:
+		if (copy_from_user(&nr, uaddr, sizeof(nr)))
+			return -EFAULT;
+
+		mutex_lock(&dev->kvm->lock);
+		r = aia_config(dev->kvm, type, &nr, false);
+		mutex_unlock(&dev->kvm->lock);
+		if (r)
+			return r;
+
+		if (copy_to_user(uaddr, &nr, sizeof(nr)))
+			return -EFAULT;
+
+		break;
+	case KVM_DEV_RISCV_AIA_GRP_ADDR:
+		if (copy_from_user(&addr, uaddr, sizeof(addr)))
+			return -EFAULT;
+
+		nr_vcpus = atomic_read(&dev->kvm->online_vcpus);
+		mutex_lock(&dev->kvm->lock);
+		if (type == KVM_DEV_RISCV_AIA_ADDR_APLIC)
+			r = aia_aplic_addr(dev->kvm, &addr, false);
+		else if (type < KVM_DEV_RISCV_AIA_ADDR_IMSIC(nr_vcpus))
+			r = aia_imsic_addr(dev->kvm, &addr,
+			    type - KVM_DEV_RISCV_AIA_ADDR_IMSIC(0), false);
+		mutex_unlock(&dev->kvm->lock);
+		if (r)
+			return r;
+
+		if (copy_to_user(uaddr, &addr, sizeof(addr)))
+			return -EFAULT;
+
+		break;
+	case KVM_DEV_RISCV_AIA_GRP_APLIC:
+		if (copy_from_user(&nr, uaddr, sizeof(nr)))
+			return -EFAULT;
+
+		mutex_lock(&dev->kvm->lock);
+		r = kvm_riscv_aia_aplic_get_attr(dev->kvm, type, &nr);
+		mutex_unlock(&dev->kvm->lock);
+		if (r)
+			return r;
+
+		if (copy_to_user(uaddr, &nr, sizeof(nr)))
+			return -EFAULT;
+
+		break;
+	case KVM_DEV_RISCV_AIA_GRP_IMSIC:
+		if (copy_from_user(&v, uaddr, sizeof(v)))
+			return -EFAULT;
+
+		mutex_lock(&dev->kvm->lock);
+		r = kvm_riscv_aia_imsic_rw_attr(dev->kvm, type, false, &v);
+		mutex_unlock(&dev->kvm->lock);
+		if (r)
+			return r;
+
+		if (copy_to_user(uaddr, &v, sizeof(v)))
+			return -EFAULT;
+
+		break;
+	}
+
+	return r;
+}
+
+static int aia_has_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
+{
+	int nr_vcpus;
+
+	switch (attr->group) {
+	case KVM_DEV_RISCV_AIA_GRP_CONFIG:
+		switch (attr->attr) {
+		case KVM_DEV_RISCV_AIA_CONFIG_MODE:
+		case KVM_DEV_RISCV_AIA_CONFIG_IDS:
+		case KVM_DEV_RISCV_AIA_CONFIG_SRCS:
+		case KVM_DEV_RISCV_AIA_CONFIG_GROUP_BITS:
+		case KVM_DEV_RISCV_AIA_CONFIG_GROUP_SHIFT:
+		case KVM_DEV_RISCV_AIA_CONFIG_HART_BITS:
+		case KVM_DEV_RISCV_AIA_CONFIG_GUEST_BITS:
+			return 0;
+		}
+		break;
+	case KVM_DEV_RISCV_AIA_GRP_ADDR:
+		nr_vcpus = atomic_read(&dev->kvm->online_vcpus);
+		if (attr->attr == KVM_DEV_RISCV_AIA_ADDR_APLIC)
+			return 0;
+		else if (attr->attr < KVM_DEV_RISCV_AIA_ADDR_IMSIC(nr_vcpus))
+			return 0;
+		break;
+	case KVM_DEV_RISCV_AIA_GRP_CTRL:
+		switch (attr->attr) {
+		case KVM_DEV_RISCV_AIA_CTRL_INIT:
+			return 0;
+		}
+		break;
+	case KVM_DEV_RISCV_AIA_GRP_APLIC:
+		return kvm_riscv_aia_aplic_has_attr(dev->kvm, attr->attr);
+	case KVM_DEV_RISCV_AIA_GRP_IMSIC:
+		return kvm_riscv_aia_imsic_has_attr(dev->kvm, attr->attr);
+	}
+
+	return -ENXIO;
+}
+
+struct kvm_device_ops kvm_riscv_aia_device_ops = {
+	.name = "kvm-riscv-aia",
+	.create = aia_create,
+	.destroy = aia_destroy,
+	.set_attr = aia_set_attr,
+	.get_attr = aia_get_attr,
+	.has_attr = aia_has_attr,
+};
+
+int kvm_riscv_vcpu_aia_update(struct kvm_vcpu *vcpu)
+{
+	/* Proceed only if AIA was initialized successfully */
+	if (!kvm_riscv_aia_initialized(vcpu->kvm))
+		return 1;
+
+	/* Update the IMSIC HW state before entering guest mode */
+	return kvm_riscv_vcpu_aia_imsic_update(vcpu);
+}
+
+void kvm_riscv_vcpu_aia_reset(struct kvm_vcpu *vcpu)
+{
+	struct kvm_vcpu_aia_csr *csr = &vcpu->arch.aia_context.guest_csr;
+	struct kvm_vcpu_aia_csr *reset_csr =
+				&vcpu->arch.aia_context.guest_reset_csr;
+
+	if (!kvm_riscv_aia_available())
+		return;
+	memcpy(csr, reset_csr, sizeof(*csr));
+
+	/* Proceed only if AIA was initialized successfully */
+	if (!kvm_riscv_aia_initialized(vcpu->kvm))
+		return;
+
+	/* Reset the IMSIC context */
+	kvm_riscv_vcpu_aia_imsic_reset(vcpu);
+}
+
+int kvm_riscv_vcpu_aia_init(struct kvm_vcpu *vcpu)
+{
+	struct kvm_vcpu_aia *vaia = &vcpu->arch.aia_context;
+
+	if (!kvm_riscv_aia_available())
+		return 0;
+
+	/*
+	 * We don't do any memory allocations over here because these
+	 * will be done after AIA device is initialized by the user-space.
+	 *
+	 * Refer, aia_init() implementation for more details.
+	 */
+
+	/* Initialize default values in AIA vcpu context */
+	vaia->imsic_addr = KVM_RISCV_AIA_UNDEF_ADDR;
+	vaia->hart_index = vcpu->vcpu_idx;
+
+	return 0;
+}
+
+void kvm_riscv_vcpu_aia_deinit(struct kvm_vcpu *vcpu)
+{
+	/* Proceed only if AIA was initialized successfully */
+	if (!kvm_riscv_aia_initialized(vcpu->kvm))
+		return;
+
+	/* Cleanup IMSIC context */
+	kvm_riscv_vcpu_aia_imsic_cleanup(vcpu);
+}
+
+int kvm_riscv_aia_inject_msi_by_id(struct kvm *kvm, u32 hart_index,
+				   u32 guest_index, u32 iid)
+{
+	unsigned long idx;
+	struct kvm_vcpu *vcpu;
+
+	/* Proceed only if AIA was initialized successfully */
+	if (!kvm_riscv_aia_initialized(kvm))
+		return -EBUSY;
+
+	/* Inject MSI to matching VCPU */
+	kvm_for_each_vcpu(idx, vcpu, kvm) {
+		if (vcpu->arch.aia_context.hart_index == hart_index)
+			return kvm_riscv_vcpu_aia_imsic_inject(vcpu,
+							       guest_index,
+							       0, iid);
+	}
+
+	return 0;
+}
+
+int kvm_riscv_aia_inject_msi(struct kvm *kvm, struct kvm_msi *msi)
+{
+	gpa_t tppn, ippn;
+	unsigned long idx;
+	struct kvm_vcpu *vcpu;
+	u32 g, toff, iid = msi->data;
+	struct kvm_aia *aia = &kvm->arch.aia;
+	gpa_t target = (((gpa_t)msi->address_hi) << 32) | msi->address_lo;
+
+	/* Proceed only if AIA was initialized successfully */
+	if (!kvm_riscv_aia_initialized(kvm))
+		return -EBUSY;
+
+	/* Convert target address to target PPN */
+	tppn = target >> IMSIC_MMIO_PAGE_SHIFT;
+
+	/* Extract and clear Guest ID from target PPN */
+	g = tppn & (BIT(aia->nr_guest_bits) - 1);
+	tppn &= ~((gpa_t)(BIT(aia->nr_guest_bits) - 1));
+
+	/* Inject MSI to matching VCPU */
+	kvm_for_each_vcpu(idx, vcpu, kvm) {
+		ippn = vcpu->arch.aia_context.imsic_addr >>
+					IMSIC_MMIO_PAGE_SHIFT;
+		if (ippn == tppn) {
+			toff = target & (IMSIC_MMIO_PAGE_SZ - 1);
+			return kvm_riscv_vcpu_aia_imsic_inject(vcpu, g,
+							       toff, iid);
+		}
+	}
+
+	return 0;
+}
+
+int kvm_riscv_aia_inject_irq(struct kvm *kvm, unsigned int irq, bool level)
+{
+	/* Proceed only if AIA was initialized successfully */
+	if (!kvm_riscv_aia_initialized(kvm))
+		return -EBUSY;
+
+	/* Inject interrupt level change in APLIC */
+	return kvm_riscv_aia_aplic_inject(kvm, irq, level);
+}
+
+void kvm_riscv_aia_init_vm(struct kvm *kvm)
+{
+	struct kvm_aia *aia = &kvm->arch.aia;
+
+	if (!kvm_riscv_aia_available())
+		return;
+
+	/*
+	 * We don't do any memory allocations over here because these
+	 * will be done after AIA device is initialized by the user-space.
+	 *
+	 * Refer, aia_init() implementation for more details.
+	 */
+
+	/* Initialize default values in AIA global context */
+	aia->mode = (kvm_riscv_aia_nr_hgei) ?
+		KVM_DEV_RISCV_AIA_MODE_AUTO : KVM_DEV_RISCV_AIA_MODE_EMUL;
+	aia->nr_ids = kvm_riscv_aia_max_ids - 1;
+	aia->nr_sources = 0;
+	aia->nr_group_bits = 0;
+	aia->nr_group_shift = KVM_DEV_RISCV_AIA_GROUP_SHIFT_MIN;
+	aia->nr_hart_bits = 0;
+	aia->nr_guest_bits = 0;
+	aia->aplic_addr = KVM_RISCV_AIA_UNDEF_ADDR;
+}
+
+void kvm_riscv_aia_destroy_vm(struct kvm *kvm)
+{
+	/* Proceed only if AIA was initialized successfully */
+	if (!kvm_riscv_aia_initialized(kvm))
+		return;
+
+	/* Cleanup APLIC context */
+	kvm_riscv_aia_aplic_cleanup(kvm);
+}

arch/riscv/kvm/main.c

@@ -116,7 +116,8 @@ static int __init riscv_kvm_init(void)
 	kvm_info("VMID %ld bits available\n", kvm_riscv_gstage_vmid_bits());
 
 	if (kvm_riscv_aia_available())
-		kvm_info("AIA available\n");
+		kvm_info("AIA available with %d guest external interrupts\n",
+			 kvm_riscv_aia_nr_hgei);
 
 	rc = kvm_init(sizeof(struct kvm_vcpu), 0, THIS_MODULE);
 	if (rc) {

arch/riscv/kvm/tlb.c

@@ -296,7 +296,7 @@ static void make_xfence_request(struct kvm *kvm,
 	unsigned int actual_req = req;
 	DECLARE_BITMAP(vcpu_mask, KVM_MAX_VCPUS);
 
-	bitmap_clear(vcpu_mask, 0, KVM_MAX_VCPUS);
+	bitmap_zero(vcpu_mask, KVM_MAX_VCPUS);
 	kvm_for_each_vcpu(i, vcpu, kvm) {
 		if (hbase != -1UL) {
 			if (vcpu->vcpu_id < hbase)

arch/riscv/kvm/vcpu.c

@@ -64,6 +64,7 @@ static const unsigned long kvm_isa_ext_arr[] = {
 	KVM_ISA_EXT_ARR(SSAIA),
 	KVM_ISA_EXT_ARR(SSTC),
 	KVM_ISA_EXT_ARR(SVINVAL),
+	KVM_ISA_EXT_ARR(SVNAPOT),
 	KVM_ISA_EXT_ARR(SVPBMT),
 	KVM_ISA_EXT_ARR(ZBB),
 	KVM_ISA_EXT_ARR(ZIHINTPAUSE),
@@ -107,6 +108,7 @@ static bool kvm_riscv_vcpu_isa_disable_allowed(unsigned long ext)
 	case KVM_RISCV_ISA_EXT_SSAIA:
 	case KVM_RISCV_ISA_EXT_SSTC:
 	case KVM_RISCV_ISA_EXT_SVINVAL:
+	case KVM_RISCV_ISA_EXT_SVNAPOT:
 	case KVM_RISCV_ISA_EXT_ZIHINTPAUSE:
 	case KVM_RISCV_ISA_EXT_ZBB:
 		return false;
@@ -263,10 +265,12 @@ int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
 
 void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
 {
+	kvm_riscv_aia_wakeon_hgei(vcpu, true);
 }
 
 void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
 {
+	kvm_riscv_aia_wakeon_hgei(vcpu, false);
 }
 
 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu)

arch/riscv/kvm/vcpu_exit.c

@@ -183,6 +183,8 @@ int kvm_riscv_vcpu_exit(struct kvm_vcpu *vcpu, struct kvm_run *run,
 	run->exit_reason = KVM_EXIT_UNKNOWN;
 	switch (trap->scause) {
 	case EXC_INST_ILLEGAL:
+	case EXC_LOAD_MISALIGNED:
+	case EXC_STORE_MISALIGNED:
 		if (vcpu->arch.guest_context.hstatus & HSTATUS_SPV) {
 			kvm_riscv_vcpu_trap_redirect(vcpu, trap);
 			ret = 1;

arch/riscv/kvm/vcpu_sbi.c

@@ -20,9 +20,7 @@ static const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_v01 = {
 };
 #endif
 
-#ifdef CONFIG_RISCV_PMU_SBI
+#ifndef CONFIG_RISCV_PMU_SBI
-extern const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_pmu;
-#else
 static const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_pmu = {
 	.extid_start = -1UL,
 	.extid_end = -1UL,
@@ -31,49 +29,49 @@ static const struct kvm_vcpu_sbi_extension vcpu_sbi_ext_pmu = {
 #endif
 
 struct kvm_riscv_sbi_extension_entry {
-	enum KVM_RISCV_SBI_EXT_ID dis_idx;
+	enum KVM_RISCV_SBI_EXT_ID ext_idx;
 	const struct kvm_vcpu_sbi_extension *ext_ptr;
 };
 
 static const struct kvm_riscv_sbi_extension_entry sbi_ext[] = {
 	{
-		.dis_idx = KVM_RISCV_SBI_EXT_V01,
+		.ext_idx = KVM_RISCV_SBI_EXT_V01,
 		.ext_ptr = &vcpu_sbi_ext_v01,
 	},
 	{
-		.dis_idx = KVM_RISCV_SBI_EXT_MAX, /* Can't be disabled */
+		.ext_idx = KVM_RISCV_SBI_EXT_MAX, /* Can't be disabled */
 		.ext_ptr = &vcpu_sbi_ext_base,
 	},
 	{
-		.dis_idx = KVM_RISCV_SBI_EXT_TIME,
+		.ext_idx = KVM_RISCV_SBI_EXT_TIME,
 		.ext_ptr = &vcpu_sbi_ext_time,
 	},
 	{
-		.dis_idx = KVM_RISCV_SBI_EXT_IPI,
+		.ext_idx = KVM_RISCV_SBI_EXT_IPI,
 		.ext_ptr = &vcpu_sbi_ext_ipi,
 	},
 	{
-		.dis_idx = KVM_RISCV_SBI_EXT_RFENCE,
+		.ext_idx = KVM_RISCV_SBI_EXT_RFENCE,
 		.ext_ptr = &vcpu_sbi_ext_rfence,
 	},
 	{
-		.dis_idx = KVM_RISCV_SBI_EXT_SRST,
+		.ext_idx = KVM_RISCV_SBI_EXT_SRST,
 		.ext_ptr = &vcpu_sbi_ext_srst,
 	},
 	{
-		.dis_idx = KVM_RISCV_SBI_EXT_HSM,
+		.ext_idx = KVM_RISCV_SBI_EXT_HSM,
 		.ext_ptr = &vcpu_sbi_ext_hsm,
 	},
 	{
-		.dis_idx = KVM_RISCV_SBI_EXT_PMU,
+		.ext_idx = KVM_RISCV_SBI_EXT_PMU,
 		.ext_ptr = &vcpu_sbi_ext_pmu,
 	},
 	{
-		.dis_idx = KVM_RISCV_SBI_EXT_EXPERIMENTAL,
+		.ext_idx = KVM_RISCV_SBI_EXT_EXPERIMENTAL,
 		.ext_ptr = &vcpu_sbi_ext_experimental,
 	},
 	{
-		.dis_idx = KVM_RISCV_SBI_EXT_VENDOR,
+		.ext_idx = KVM_RISCV_SBI_EXT_VENDOR,
 		.ext_ptr = &vcpu_sbi_ext_vendor,
 	},
 };
@@ -147,7 +145,7 @@ static int riscv_vcpu_set_sbi_ext_single(struct kvm_vcpu *vcpu,
 		return -EINVAL;
 
 	for (i = 0; i < ARRAY_SIZE(sbi_ext); i++) {
-		if (sbi_ext[i].dis_idx == reg_num) {
+		if (sbi_ext[i].ext_idx == reg_num) {
 			sext = &sbi_ext[i];
 			break;
 		}
@@ -155,7 +153,15 @@ static int riscv_vcpu_set_sbi_ext_single(struct kvm_vcpu *vcpu,
 	if (!sext)
 		return -ENOENT;
 
-	scontext->extension_disabled[sext->dis_idx] = !reg_val;
+	/*
+	 * We can't set the extension status to available here, since it may
+	 * have a probe() function which needs to confirm availability first,
+	 * but it may be too early to call that here. We can set the status to
+	 * unavailable, though.
+	 */
+	if (!reg_val)
+		scontext->ext_status[sext->ext_idx] =
+			KVM_RISCV_SBI_EXT_UNAVAILABLE;
 
 	return 0;
 }
@@ -172,7 +178,7 @@ static int riscv_vcpu_get_sbi_ext_single(struct kvm_vcpu *vcpu,
 		return -EINVAL;
 
 	for (i = 0; i < ARRAY_SIZE(sbi_ext); i++) {
-		if (sbi_ext[i].dis_idx == reg_num) {
+		if (sbi_ext[i].ext_idx == reg_num) {
 			sext = &sbi_ext[i];
 			break;
 		}
@@ -180,7 +186,15 @@ static int riscv_vcpu_get_sbi_ext_single(struct kvm_vcpu *vcpu,
 	if (!sext)
 		return -ENOENT;
 
-	*reg_val = !scontext->extension_disabled[sext->dis_idx];
+	/*
+	 * If the extension status is still uninitialized, then we should probe
+	 * to determine if it's available, but it may be too early to do that
+	 * here. The best we can do is report that the extension has not been
+	 * disabled, i.e. we return 1 when the extension is available and also
+	 * when it only may be available.
+	 */
+	*reg_val = scontext->ext_status[sext->ext_idx] !=
+				KVM_RISCV_SBI_EXT_UNAVAILABLE;
 
 	return 0;
 }
@@ -307,18 +321,32 @@ int kvm_riscv_vcpu_get_reg_sbi_ext(struct kvm_vcpu *vcpu,
 const struct kvm_vcpu_sbi_extension *kvm_vcpu_sbi_find_ext(
 				struct kvm_vcpu *vcpu, unsigned long extid)
 {
-	int i;
-	const struct kvm_riscv_sbi_extension_entry *sext;
 	struct kvm_vcpu_sbi_context *scontext = &vcpu->arch.sbi_context;
+	const struct kvm_riscv_sbi_extension_entry *entry;
+	const struct kvm_vcpu_sbi_extension *ext;
+	int i;
 
 	for (i = 0; i < ARRAY_SIZE(sbi_ext); i++) {
-		sext = &sbi_ext[i];
+		entry = &sbi_ext[i];
-		if (sext->ext_ptr->extid_start <= extid &&
+		ext = entry->ext_ptr;
-		    sext->ext_ptr->extid_end >= extid) {
+
-			if (sext->dis_idx < KVM_RISCV_SBI_EXT_MAX &&
+		if (ext->extid_start <= extid && ext->extid_end >= extid) {
-			    scontext->extension_disabled[sext->dis_idx])
+			if (entry->ext_idx >= KVM_RISCV_SBI_EXT_MAX ||
+			    scontext->ext_status[entry->ext_idx] ==
+						KVM_RISCV_SBI_EXT_AVAILABLE)
+				return ext;
+			if (scontext->ext_status[entry->ext_idx] ==
+						KVM_RISCV_SBI_EXT_UNAVAILABLE)
 				return NULL;
-			return sbi_ext[i].ext_ptr;
+			if (ext->probe && !ext->probe(vcpu)) {
+				scontext->ext_status[entry->ext_idx] =
+					KVM_RISCV_SBI_EXT_UNAVAILABLE;
+				return NULL;
+			}
+
+			scontext->ext_status[entry->ext_idx] =
+				KVM_RISCV_SBI_EXT_AVAILABLE;
+			return ext;
 		}
 	}
 

arch/riscv/kvm/vm.c

@@ -55,11 +55,129 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
 	kvm_riscv_aia_destroy_vm(kvm);
 }
 
+int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irql,
+			  bool line_status)
+{
+	if (!irqchip_in_kernel(kvm))
+		return -ENXIO;
+
+	return kvm_riscv_aia_inject_irq(kvm, irql->irq, irql->level);
+}
+
+int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e,
+		struct kvm *kvm, int irq_source_id,
+		int level, bool line_status)
+{
+	struct kvm_msi msi;
+
+	if (!level)
+		return -1;
+
+	msi.address_lo = e->msi.address_lo;
+	msi.address_hi = e->msi.address_hi;
+	msi.data = e->msi.data;
+	msi.flags = e->msi.flags;
+	msi.devid = e->msi.devid;
+
+	return kvm_riscv_aia_inject_msi(kvm, &msi);
+}
+
+static int kvm_riscv_set_irq(struct kvm_kernel_irq_routing_entry *e,
+			     struct kvm *kvm, int irq_source_id,
+			     int level, bool line_status)
+{
+	return kvm_riscv_aia_inject_irq(kvm, e->irqchip.pin, level);
+}
+
+int kvm_riscv_setup_default_irq_routing(struct kvm *kvm, u32 lines)
+{
+	struct kvm_irq_routing_entry *ents;
+	int i, rc;
+
+	ents = kcalloc(lines, sizeof(*ents), GFP_KERNEL);
+	if (!ents)
+		return -ENOMEM;
+
+	for (i = 0; i < lines; i++) {
+		ents[i].gsi = i;
+		ents[i].type = KVM_IRQ_ROUTING_IRQCHIP;
+		ents[i].u.irqchip.irqchip = 0;
+		ents[i].u.irqchip.pin = i;
+	}
+	rc = kvm_set_irq_routing(kvm, ents, lines, 0);
+	kfree(ents);
+
+	return rc;
+}
+
+bool kvm_arch_can_set_irq_routing(struct kvm *kvm)
+{
+	return irqchip_in_kernel(kvm);
+}
+
+int kvm_set_routing_entry(struct kvm *kvm,
+			  struct kvm_kernel_irq_routing_entry *e,
+			  const struct kvm_irq_routing_entry *ue)
+{
+	int r = -EINVAL;
+
+	switch (ue->type) {
+	case KVM_IRQ_ROUTING_IRQCHIP:
+		e->set = kvm_riscv_set_irq;
+		e->irqchip.irqchip = ue->u.irqchip.irqchip;
+		e->irqchip.pin = ue->u.irqchip.pin;
+		if ((e->irqchip.pin >= KVM_IRQCHIP_NUM_PINS) ||
+		    (e->irqchip.irqchip >= KVM_NR_IRQCHIPS))
+			goto out;
+		break;
+	case KVM_IRQ_ROUTING_MSI:
+		e->set = kvm_set_msi;
+		e->msi.address_lo = ue->u.msi.address_lo;
+		e->msi.address_hi = ue->u.msi.address_hi;
+		e->msi.data = ue->u.msi.data;
+		e->msi.flags = ue->flags;
+		e->msi.devid = ue->u.msi.devid;
+		break;
+	default:
+		goto out;
+	}
+	r = 0;
+out:
+	return r;
+}
+
+int kvm_arch_set_irq_inatomic(struct kvm_kernel_irq_routing_entry *e,
+			      struct kvm *kvm, int irq_source_id, int level,
+			      bool line_status)
+{
+	if (!level)
+		return -EWOULDBLOCK;
+
+	switch (e->type) {
+	case KVM_IRQ_ROUTING_MSI:
+		return kvm_set_msi(e, kvm, irq_source_id, level, line_status);
+
+	case KVM_IRQ_ROUTING_IRQCHIP:
+		return kvm_riscv_set_irq(e, kvm, irq_source_id,
+					 level, line_status);
+	}
+
+	return -EWOULDBLOCK;
+}
+
+bool kvm_arch_irqchip_in_kernel(struct kvm *kvm)
+{
+	return irqchip_in_kernel(kvm);
+}
+
 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 {
 	int r;
 
 	switch (ext) {
+	case KVM_CAP_IRQCHIP:
+		r = kvm_riscv_aia_available();
+		break;
 	case KVM_CAP_IOEVENTFD:
 	case KVM_CAP_DEVICE_CTRL:
 	case KVM_CAP_USER_MEMORY:

arch/s390/boot/uv.c

@@ -47,6 +47,10 @@ void uv_query_info(void)
 		uv_info.conf_dump_finalize_len = uvcb.conf_dump_finalize_len;
 		uv_info.supp_att_req_hdr_ver = uvcb.supp_att_req_hdr_ver;
 		uv_info.supp_att_pflags = uvcb.supp_att_pflags;
+		uv_info.supp_add_secret_req_ver = uvcb.supp_add_secret_req_ver;
+		uv_info.supp_add_secret_pcf = uvcb.supp_add_secret_pcf;
+		uv_info.supp_secret_types = uvcb.supp_secret_types;
+		uv_info.max_secrets = uvcb.max_secrets;
 	}
 
 #ifdef CONFIG_PROTECTED_VIRTUALIZATION_GUEST

arch/s390/include/asm/uv.h

@@ -58,6 +58,9 @@
 #define UVC_CMD_SET_SHARED_ACCESS	0x1000
 #define UVC_CMD_REMOVE_SHARED_ACCESS	0x1001
 #define UVC_CMD_RETR_ATTEST		0x1020
+#define UVC_CMD_ADD_SECRET		0x1031
+#define UVC_CMD_LIST_SECRETS		0x1033
+#define UVC_CMD_LOCK_SECRETS		0x1034
 
 /* Bits in installed uv calls */
 enum uv_cmds_inst {
@@ -88,6 +91,9 @@ enum uv_cmds_inst {
 	BIT_UVC_CMD_DUMP_CPU = 26,
 	BIT_UVC_CMD_DUMP_COMPLETE = 27,
 	BIT_UVC_CMD_RETR_ATTEST = 28,
+	BIT_UVC_CMD_ADD_SECRET = 29,
+	BIT_UVC_CMD_LIST_SECRETS = 30,
+	BIT_UVC_CMD_LOCK_SECRETS = 31,
 };
 
 enum uv_feat_ind {
@@ -117,7 +123,7 @@ struct uv_cb_qui {
 	u32 reserved70[3];			/* 0x0070 */
 	u32 max_num_sec_conf;			/* 0x007c */
 	u64 max_guest_stor_addr;		/* 0x0080 */
-	u8  reserved88[158 - 136];		/* 0x0088 */
+	u8  reserved88[0x9e - 0x88];		/* 0x0088 */
 	u16 max_guest_cpu_id;			/* 0x009e */
 	u64 uv_feature_indications;		/* 0x00a0 */
 	u64 reserveda8;				/* 0x00a8 */
@@ -129,7 +135,12 @@ struct uv_cb_qui {
 	u64 reservedd8;				/* 0x00d8 */
 	u64 supp_att_req_hdr_ver;		/* 0x00e0 */
 	u64 supp_att_pflags;			/* 0x00e8 */
-	u8 reservedf0[256 - 240];		/* 0x00f0 */
+	u64 reservedf0;				/* 0x00f0 */
+	u64 supp_add_secret_req_ver;		/* 0x00f8 */
+	u64 supp_add_secret_pcf;		/* 0x0100 */
+	u64 supp_secret_types;			/* 0x0180 */
+	u16 max_secrets;			/* 0x0110 */
+	u8 reserved112[0x120 - 0x112];		/* 0x0112 */
 } __packed __aligned(8);
 
 /* Initialize Ultravisor */
@@ -292,6 +303,19 @@ struct uv_cb_dump_complete {
 	u64 reserved30[5];
 } __packed __aligned(8);
 
+/*
+ * A common UV call struct for pv guests that contains a single address
+ * Examples:
+ * Add Secret
+ * List Secrets
+ */
+struct uv_cb_guest_addr {
+	struct uv_cb_header header;
+	u64 reserved08[3];
+	u64 addr;
+	u64 reserved28[4];
+} __packed __aligned(8);
+
 static inline int __uv_call(unsigned long r1, unsigned long r2)
 {
 	int cc;
@@ -365,6 +389,10 @@ struct uv_info {
 	unsigned long conf_dump_finalize_len;
 	unsigned long supp_att_req_hdr_ver;
 	unsigned long supp_att_pflags;
+	unsigned long supp_add_secret_req_ver;
+	unsigned long supp_add_secret_pcf;
+	unsigned long supp_secret_types;
+	unsigned short max_secrets;
 };
 
 extern struct uv_info uv_info;

arch/s390/include/uapi/asm/uvdevice.h

@@ -32,6 +32,33 @@ struct uvio_attest {
 	__u16 reserved136;				/* 0x0136 */
 };
 
+/**
+ * uvio_uvdev_info - Information of supported functions
+ * @supp_uvio_cmds - supported IOCTLs by this device
+ * @supp_uv_cmds - supported UVCs corresponding to the IOCTL
+ *
+ * UVIO request to get information about supported request types by this
+ * uvdevice and the Ultravisor.  Everything is output. Bits are in LSB0
+ * ordering.  If the bit is set in both, @supp_uvio_cmds and @supp_uv_cmds, the
+ * uvdevice and the Ultravisor support that call.
+ *
+ * Note that bit 0 (UVIO_IOCTL_UVDEV_INFO_NR) is always zero for `supp_uv_cmds`
+ * as there is no corresponding UV-call.
+ */
+struct uvio_uvdev_info {
+	/*
+	 * If bit `n` is set, this device supports the IOCTL with nr `n`.
+	 */
+	__u64 supp_uvio_cmds;
+	/*
+	 * If bit `n` is set, the Ultravisor(UV) supports the UV-call
+	 * corresponding to the IOCTL with nr `n` in the calling contextx (host
+	 * or guest).  The value is only valid if the corresponding bit in
+	 * @supp_uvio_cmds is set as well.
+	 */
+	__u64 supp_uv_cmds;
+};
+
 /*
  * The following max values define an upper length for the IOCTL in/out buffers.
  * However, they do not represent the maximum the Ultravisor allows which is
@@ -42,10 +69,34 @@ struct uvio_attest {
 #define UVIO_ATT_ARCB_MAX_LEN		0x100000
 #define UVIO_ATT_MEASUREMENT_MAX_LEN	0x8000
 #define UVIO_ATT_ADDITIONAL_MAX_LEN	0x8000
+#define UVIO_ADD_SECRET_MAX_LEN		0x100000
+#define UVIO_LIST_SECRETS_LEN		0x1000
 
 #define UVIO_DEVICE_NAME "uv"
 #define UVIO_TYPE_UVC 'u'
 
-#define UVIO_IOCTL_ATT _IOWR(UVIO_TYPE_UVC, 0x01, struct uvio_ioctl_cb)
+enum UVIO_IOCTL_NR {
+	UVIO_IOCTL_UVDEV_INFO_NR = 0x00,
+	UVIO_IOCTL_ATT_NR,
+	UVIO_IOCTL_ADD_SECRET_NR,
+	UVIO_IOCTL_LIST_SECRETS_NR,
+	UVIO_IOCTL_LOCK_SECRETS_NR,
+	/* must be the last entry */
+	UVIO_IOCTL_NUM_IOCTLS
+};
+
+#define UVIO_IOCTL(nr)		_IOWR(UVIO_TYPE_UVC, nr, struct uvio_ioctl_cb)
+#define UVIO_IOCTL_UVDEV_INFO	UVIO_IOCTL(UVIO_IOCTL_UVDEV_INFO_NR)
+#define UVIO_IOCTL_ATT		UVIO_IOCTL(UVIO_IOCTL_ATT_NR)
+#define UVIO_IOCTL_ADD_SECRET	UVIO_IOCTL(UVIO_IOCTL_ADD_SECRET_NR)
+#define UVIO_IOCTL_LIST_SECRETS	UVIO_IOCTL(UVIO_IOCTL_LIST_SECRETS_NR)
+#define UVIO_IOCTL_LOCK_SECRETS	UVIO_IOCTL(UVIO_IOCTL_LOCK_SECRETS_NR)
+
+#define UVIO_SUPP_CALL(nr)	(1ULL << (nr))
+#define UVIO_SUPP_UDEV_INFO	UVIO_SUPP_CALL(UVIO_IOCTL_UDEV_INFO_NR)
+#define UVIO_SUPP_ATT		UVIO_SUPP_CALL(UVIO_IOCTL_ATT_NR)
+#define UVIO_SUPP_ADD_SECRET	UVIO_SUPP_CALL(UVIO_IOCTL_ADD_SECRET_NR)
+#define UVIO_SUPP_LIST_SECRETS	UVIO_SUPP_CALL(UVIO_IOCTL_LIST_SECRETS_NR)
+#define UVIO_SUPP_LOCK_SECRETS	UVIO_SUPP_CALL(UVIO_IOCTL_LOCK_SECRETS_NR)
 
 #endif /* __S390_ASM_UVDEVICE_H */

arch/s390/kernel/uv.c

@@ -23,12 +23,20 @@
 int __bootdata_preserved(prot_virt_guest);
 #endif
 
+/*
+ * uv_info contains both host and guest information but it's currently only
+ * expected to be used within modules if it's the KVM module or for
+ * any PV guest module.
+ *
+ * The kernel itself will write these values once in uv_query_info()
+ * and then make some of them readable via a sysfs interface.
+ */
 struct uv_info __bootdata_preserved(uv_info);
+EXPORT_SYMBOL(uv_info);
 
 #if IS_ENABLED(CONFIG_KVM)
 int __bootdata_preserved(prot_virt_host);
 EXPORT_SYMBOL(prot_virt_host);
-EXPORT_SYMBOL(uv_info);
 
 static int __init uv_init(phys_addr_t stor_base, unsigned long stor_len)
 {
@@ -462,13 +470,13 @@ EXPORT_SYMBOL_GPL(arch_make_page_accessible);
 
 #if defined(CONFIG_PROTECTED_VIRTUALIZATION_GUEST) || IS_ENABLED(CONFIG_KVM)
 static ssize_t uv_query_facilities(struct kobject *kobj,
-				   struct kobj_attribute *attr, char *page)
+				   struct kobj_attribute *attr, char *buf)
 {
-	return scnprintf(page, PAGE_SIZE, "%lx\n%lx\n%lx\n%lx\n",
+	return sysfs_emit(buf, "%lx\n%lx\n%lx\n%lx\n",
-			uv_info.inst_calls_list[0],
+			  uv_info.inst_calls_list[0],
-			uv_info.inst_calls_list[1],
+			  uv_info.inst_calls_list[1],
-			uv_info.inst_calls_list[2],
+			  uv_info.inst_calls_list[2],
-			uv_info.inst_calls_list[3]);
+			  uv_info.inst_calls_list[3]);
 }
 
 static struct kobj_attribute uv_query_facilities_attr =
@@ -493,30 +501,27 @@ static struct kobj_attribute uv_query_supp_se_hdr_pcf_attr =
 	__ATTR(supp_se_hdr_pcf, 0444, uv_query_supp_se_hdr_pcf, NULL);
 
 static ssize_t uv_query_dump_cpu_len(struct kobject *kobj,
-				     struct kobj_attribute *attr, char *page)
+				     struct kobj_attribute *attr, char *buf)
 {
-	return scnprintf(page, PAGE_SIZE, "%lx\n",
+	return sysfs_emit(buf, "%lx\n", uv_info.guest_cpu_stor_len);
-			uv_info.guest_cpu_stor_len);
 }
 
 static struct kobj_attribute uv_query_dump_cpu_len_attr =
 	__ATTR(uv_query_dump_cpu_len, 0444, uv_query_dump_cpu_len, NULL);
 
 static ssize_t uv_query_dump_storage_state_len(struct kobject *kobj,
-					       struct kobj_attribute *attr, char *page)
+					       struct kobj_attribute *attr, char *buf)
 {
-	return scnprintf(page, PAGE_SIZE, "%lx\n",
+	return sysfs_emit(buf, "%lx\n", uv_info.conf_dump_storage_state_len);
-			uv_info.conf_dump_storage_state_len);
 }
 
 static struct kobj_attribute uv_query_dump_storage_state_len_attr =
 	__ATTR(dump_storage_state_len, 0444, uv_query_dump_storage_state_len, NULL);
 
 static ssize_t uv_query_dump_finalize_len(struct kobject *kobj,
-					  struct kobj_attribute *attr, char *page)
+					  struct kobj_attribute *attr, char *buf)
 {
-	return scnprintf(page, PAGE_SIZE, "%lx\n",
+	return sysfs_emit(buf, "%lx\n", uv_info.conf_dump_finalize_len);
-			uv_info.conf_dump_finalize_len);
 }
 
 static struct kobj_attribute uv_query_dump_finalize_len_attr =
@@ -532,53 +537,86 @@ static struct kobj_attribute uv_query_feature_indications_attr =
 	__ATTR(feature_indications, 0444, uv_query_feature_indications, NULL);
 
 static ssize_t uv_query_max_guest_cpus(struct kobject *kobj,
-				       struct kobj_attribute *attr, char *page)
+				       struct kobj_attribute *attr, char *buf)
 {
-	return scnprintf(page, PAGE_SIZE, "%d\n",
+	return sysfs_emit(buf, "%d\n", uv_info.max_guest_cpu_id + 1);
-			uv_info.max_guest_cpu_id + 1);
 }
 
 static struct kobj_attribute uv_query_max_guest_cpus_attr =
 	__ATTR(max_cpus, 0444, uv_query_max_guest_cpus, NULL);
 
 static ssize_t uv_query_max_guest_vms(struct kobject *kobj,
-				      struct kobj_attribute *attr, char *page)
+				      struct kobj_attribute *attr, char *buf)
 {
-	return scnprintf(page, PAGE_SIZE, "%d\n",
+	return sysfs_emit(buf, "%d\n", uv_info.max_num_sec_conf);
-			uv_info.max_num_sec_conf);
 }
 
 static struct kobj_attribute uv_query_max_guest_vms_attr =
 	__ATTR(max_guests, 0444, uv_query_max_guest_vms, NULL);
 
 static ssize_t uv_query_max_guest_addr(struct kobject *kobj,
-				       struct kobj_attribute *attr, char *page)
+				       struct kobj_attribute *attr, char *buf)
 {
-	return scnprintf(page, PAGE_SIZE, "%lx\n",
+	return sysfs_emit(buf, "%lx\n", uv_info.max_sec_stor_addr);
-			uv_info.max_sec_stor_addr);
 }
 
 static struct kobj_attribute uv_query_max_guest_addr_attr =
 	__ATTR(max_address, 0444, uv_query_max_guest_addr, NULL);
 
 static ssize_t uv_query_supp_att_req_hdr_ver(struct kobject *kobj,
-					     struct kobj_attribute *attr, char *page)
+					     struct kobj_attribute *attr, char *buf)
 {
-	return scnprintf(page, PAGE_SIZE, "%lx\n", uv_info.supp_att_req_hdr_ver);
+	return sysfs_emit(buf, "%lx\n", uv_info.supp_att_req_hdr_ver);
 }
 
 static struct kobj_attribute uv_query_supp_att_req_hdr_ver_attr =
 	__ATTR(supp_att_req_hdr_ver, 0444, uv_query_supp_att_req_hdr_ver, NULL);
 
 static ssize_t uv_query_supp_att_pflags(struct kobject *kobj,
-					struct kobj_attribute *attr, char *page)
+					struct kobj_attribute *attr, char *buf)
 {
-	return scnprintf(page, PAGE_SIZE, "%lx\n", uv_info.supp_att_pflags);
+	return sysfs_emit(buf, "%lx\n", uv_info.supp_att_pflags);
 }
 
 static struct kobj_attribute uv_query_supp_att_pflags_attr =
 	__ATTR(supp_att_pflags, 0444, uv_query_supp_att_pflags, NULL);
 
+static ssize_t uv_query_supp_add_secret_req_ver(struct kobject *kobj,
+						struct kobj_attribute *attr, char *buf)
+{
+	return sysfs_emit(buf, "%lx\n", uv_info.supp_add_secret_req_ver);
+}
+
+static struct kobj_attribute uv_query_supp_add_secret_req_ver_attr =
+	__ATTR(supp_add_secret_req_ver, 0444, uv_query_supp_add_secret_req_ver, NULL);
+
+static ssize_t uv_query_supp_add_secret_pcf(struct kobject *kobj,
+					    struct kobj_attribute *attr, char *buf)
+{
+	return sysfs_emit(buf, "%lx\n", uv_info.supp_add_secret_pcf);
+}
+
+static struct kobj_attribute uv_query_supp_add_secret_pcf_attr =
+	__ATTR(supp_add_secret_pcf, 0444, uv_query_supp_add_secret_pcf, NULL);
+
+static ssize_t uv_query_supp_secret_types(struct kobject *kobj,
+					  struct kobj_attribute *attr, char *buf)
+{
+	return sysfs_emit(buf, "%lx\n", uv_info.supp_secret_types);
+}
+
+static struct kobj_attribute uv_query_supp_secret_types_attr =
+	__ATTR(supp_secret_types, 0444, uv_query_supp_secret_types, NULL);
+
+static ssize_t uv_query_max_secrets(struct kobject *kobj,
+				    struct kobj_attribute *attr, char *buf)
+{
+	return sysfs_emit(buf, "%d\n", uv_info.max_secrets);
+}
+
+static struct kobj_attribute uv_query_max_secrets_attr =
+	__ATTR(max_secrets, 0444, uv_query_max_secrets, NULL);
+
 static struct attribute *uv_query_attrs[] = {
 	&uv_query_facilities_attr.attr,
 	&uv_query_feature_indications_attr.attr,
@@ -592,6 +630,10 @@ static struct attribute *uv_query_attrs[] = {
 	&uv_query_dump_cpu_len_attr.attr,
 	&uv_query_supp_att_req_hdr_ver_attr.attr,
 	&uv_query_supp_att_pflags_attr.attr,
+	&uv_query_supp_add_secret_req_ver_attr.attr,
+	&uv_query_supp_add_secret_pcf_attr.attr,
+	&uv_query_supp_secret_types_attr.attr,
+	&uv_query_max_secrets_attr.attr,
 	NULL,
 };
 
@@ -600,18 +642,18 @@ static struct attribute_group uv_query_attr_group = {
 };
 
 static ssize_t uv_is_prot_virt_guest(struct kobject *kobj,
-				     struct kobj_attribute *attr, char *page)
+				     struct kobj_attribute *attr, char *buf)
 {
 	int val = 0;
 
 #ifdef CONFIG_PROTECTED_VIRTUALIZATION_GUEST
 	val = prot_virt_guest;
 #endif
-	return scnprintf(page, PAGE_SIZE, "%d\n", val);
+	return sysfs_emit(buf, "%d\n", val);
 }
 
 static ssize_t uv_is_prot_virt_host(struct kobject *kobj,
-				    struct kobj_attribute *attr, char *page)
+				    struct kobj_attribute *attr, char *buf)
 {
 	int val = 0;
 
@@ -619,7 +661,7 @@ static ssize_t uv_is_prot_virt_host(struct kobject *kobj,
 	val = prot_virt_host;
 #endif
 
-	return scnprintf(page, PAGE_SIZE, "%d\n", val);
+	return sysfs_emit(buf, "%d\n", val);
 }
 
 static struct kobj_attribute uv_prot_virt_guest =

arch/s390/kvm/diag.c

@@ -166,6 +166,7 @@ static int diag9c_forwarding_overrun(void)
 static int __diag_time_slice_end_directed(struct kvm_vcpu *vcpu)
 {
 	struct kvm_vcpu *tcpu;
+	int tcpu_cpu;
 	int tid;
 
 	tid = vcpu->run->s.regs.gprs[(vcpu->arch.sie_block->ipa & 0xf0) >> 4];
@@ -181,14 +182,15 @@ static int __diag_time_slice_end_directed(struct kvm_vcpu *vcpu)
 		goto no_yield;
 
 	/* target guest VCPU already running */
-	if (READ_ONCE(tcpu->cpu) >= 0) {
+	tcpu_cpu = READ_ONCE(tcpu->cpu);
+	if (tcpu_cpu >= 0) {
 		if (!diag9c_forwarding_hz || diag9c_forwarding_overrun())
 			goto no_yield;
 
 		/* target host CPU already running */
-		if (!vcpu_is_preempted(tcpu->cpu))
+		if (!vcpu_is_preempted(tcpu_cpu))
 			goto no_yield;
-		smp_yield_cpu(tcpu->cpu);
+		smp_yield_cpu(tcpu_cpu);
 		VCPU_EVENT(vcpu, 5,
 			   "diag time slice end directed to %d: yield forwarded",
 			   tid);

arch/s390/kvm/kvm-s390.c

@@ -2156,6 +2156,10 @@ static unsigned long kvm_s390_next_dirty_cmma(struct kvm_memslots *slots,
 		ms = container_of(mnode, struct kvm_memory_slot, gfn_node[slots->node_idx]);
 		ofs = 0;
 	}
+
+	if (cur_gfn < ms->base_gfn)
+		ofs = 0;
+
 	ofs = find_next_bit(kvm_second_dirty_bitmap(ms), ms->npages, ofs);
 	while (ofs >= ms->npages && (mnode = rb_next(mnode))) {
 		ms = container_of(mnode, struct kvm_memory_slot, gfn_node[slots->node_idx]);

arch/s390/kvm/vsie.c

@@ -177,7 +177,8 @@ static int setup_apcb00(struct kvm_vcpu *vcpu, unsigned long *apcb_s,
 			    sizeof(struct kvm_s390_apcb0)))
 		return -EFAULT;
 
-	bitmap_and(apcb_s, apcb_s, apcb_h, sizeof(struct kvm_s390_apcb0));
+	bitmap_and(apcb_s, apcb_s, apcb_h,
+		   BITS_PER_BYTE * sizeof(struct kvm_s390_apcb0));
 
 	return 0;
 }
@@ -203,7 +204,8 @@ static int setup_apcb11(struct kvm_vcpu *vcpu, unsigned long *apcb_s,
 			    sizeof(struct kvm_s390_apcb1)))
 		return -EFAULT;
 
-	bitmap_and(apcb_s, apcb_s, apcb_h, sizeof(struct kvm_s390_apcb1));
+	bitmap_and(apcb_s, apcb_s, apcb_h,
+		   BITS_PER_BYTE * sizeof(struct kvm_s390_apcb1));
 
 	return 0;
 }

arch/x86/include/asm/kvm-x86-pmu-ops.h

@@ -13,7 +13,6 @@ BUILD_BUG_ON(1)
  * at the call sites.
  */
 KVM_X86_PMU_OP(hw_event_available)
-KVM_X86_PMU_OP(pmc_is_enabled)
 KVM_X86_PMU_OP(pmc_idx_to_pmc)
 KVM_X86_PMU_OP(rdpmc_ecx_to_pmc)
 KVM_X86_PMU_OP(msr_idx_to_pmc)

arch/x86/include/asm/kvm_host.h

@@ -523,7 +523,7 @@ struct kvm_pmu {
 	u64 global_status;
 	u64 counter_bitmask[2];
 	u64 global_ctrl_mask;
-	u64 global_ovf_ctrl_mask;
+	u64 global_status_mask;
 	u64 reserved_bits;
 	u64 raw_event_mask;
 	struct kvm_pmc gp_counters[KVM_INTEL_PMC_MAX_GENERIC];

arch/x86/kvm/cpuid.c

@@ -501,20 +501,15 @@ int kvm_vcpu_ioctl_get_cpuid2(struct kvm_vcpu *vcpu,
 			      struct kvm_cpuid2 *cpuid,
 			      struct kvm_cpuid_entry2 __user *entries)
 {
-	int r;
-
-	r = -E2BIG;
 	if (cpuid->nent < vcpu->arch.cpuid_nent)
-		goto out;
+		return -E2BIG;
-	r = -EFAULT;
+
 	if (copy_to_user(entries, vcpu->arch.cpuid_entries,
 			 vcpu->arch.cpuid_nent * sizeof(struct kvm_cpuid_entry2)))
-		goto out;
+		return -EFAULT;
-	return 0;
 
-out:
 	cpuid->nent = vcpu->arch.cpuid_nent;
-	return r;
+	return 0;
 }
 
 /* Mask kvm_cpu_caps for @leaf with the raw CPUID capabilities of this CPU. */
@@ -734,6 +729,10 @@ void kvm_set_cpu_caps(void)
 		F(NULL_SEL_CLR_BASE) | F(AUTOIBRS) | 0 /* PrefetchCtlMsr */
 	);
 
+	kvm_cpu_cap_init_kvm_defined(CPUID_8000_0022_EAX,
+		F(PERFMON_V2)
+	);
+
 	/*
 	 * Synthesize "LFENCE is serializing" into the AMD-defined entry in
 	 * KVM's supported CPUID if the feature is reported as supported by the
@@ -948,7 +947,7 @@ static inline int __do_cpuid_func(struct kvm_cpuid_array *array, u32 function)
 		union cpuid10_eax eax;
 		union cpuid10_edx edx;
 
-		if (!static_cpu_has(X86_FEATURE_ARCH_PERFMON)) {
+		if (!enable_pmu || !static_cpu_has(X86_FEATURE_ARCH_PERFMON)) {
 			entry->eax = entry->ebx = entry->ecx = entry->edx = 0;
 			break;
 		}
@@ -1128,7 +1127,7 @@ static inline int __do_cpuid_func(struct kvm_cpuid_array *array, u32 function)
 		entry->edx = 0;
 		break;
 	case 0x80000000:
-		entry->eax = min(entry->eax, 0x80000021);
+		entry->eax = min(entry->eax, 0x80000022);
 		/*
 		 * Serializing LFENCE is reported in a multitude of ways, and
 		 * NullSegClearsBase is not reported in CPUID on Zen2; help
@@ -1233,6 +1232,28 @@ static inline int __do_cpuid_func(struct kvm_cpuid_array *array, u32 function)
 		entry->ebx = entry->ecx = entry->edx = 0;
 		cpuid_entry_override(entry, CPUID_8000_0021_EAX);
 		break;
+	/* AMD Extended Performance Monitoring and Debug */
+	case 0x80000022: {
+		union cpuid_0x80000022_ebx ebx;
+
+		entry->ecx = entry->edx = 0;
+		if (!enable_pmu || !kvm_cpu_cap_has(X86_FEATURE_PERFMON_V2)) {
+			entry->eax = entry->ebx;
+			break;
+		}
+
+		cpuid_entry_override(entry, CPUID_8000_0022_EAX);
+
+		if (kvm_cpu_cap_has(X86_FEATURE_PERFMON_V2))
+			ebx.split.num_core_pmc = kvm_pmu_cap.num_counters_gp;
+		else if (kvm_cpu_cap_has(X86_FEATURE_PERFCTR_CORE))
+			ebx.split.num_core_pmc = AMD64_NUM_COUNTERS_CORE;
+		else
+			ebx.split.num_core_pmc = AMD64_NUM_COUNTERS;
+
+		entry->ebx = ebx.full;
+		break;
+	}
 	/*Add support for Centaur's CPUID instruction*/
 	case 0xC0000000:
 		/*Just support up to 0xC0000004 now*/

arch/x86/kvm/i8259.c

@@ -411,7 +411,10 @@ static u32 pic_poll_read(struct kvm_kpic_state *s, u32 addr1)
 		pic_clear_isr(s, ret);
 		if (addr1 >> 7 || ret != 2)
 			pic_update_irq(s->pics_state);
+		/* Bit 7 is 1, means there's an interrupt */
+		ret |= 0x80;
 	} else {
+		/* Bit 7 is 0, means there's no interrupt */
 		ret = 0x07;
 		pic_update_irq(s->pics_state);
 	}

arch/x86/kvm/lapic.c

@@ -51,11 +51,6 @@
 #define mod_64(x, y) ((x) % (y))
 #endif
 
-#define PRId64 "d"
-#define PRIx64 "llx"
-#define PRIu64 "u"
-#define PRIo64 "o"
-
 /* 14 is the version for Xeon and Pentium 8.4.8*/
 #define APIC_VERSION			0x14UL
 #define LAPIC_MMIO_LENGTH		(1 << 12)

arch/x86/kvm/mmu/mmu.c

@@ -58,6 +58,8 @@
 
 extern bool itlb_multihit_kvm_mitigation;
 
+static bool nx_hugepage_mitigation_hard_disabled;
+
 int __read_mostly nx_huge_pages = -1;
 static uint __read_mostly nx_huge_pages_recovery_period_ms;
 #ifdef CONFIG_PREEMPT_RT
@@ -67,12 +69,13 @@ static uint __read_mostly nx_huge_pages_recovery_ratio = 0;
 static uint __read_mostly nx_huge_pages_recovery_ratio = 60;
 #endif
 
+static int get_nx_huge_pages(char *buffer, const struct kernel_param *kp);
 static int set_nx_huge_pages(const char *val, const struct kernel_param *kp);
 static int set_nx_huge_pages_recovery_param(const char *val, const struct kernel_param *kp);
 
 static const struct kernel_param_ops nx_huge_pages_ops = {
 	.set = set_nx_huge_pages,
-	.get = param_get_bool,
+	.get = get_nx_huge_pages,
 };
 
 static const struct kernel_param_ops nx_huge_pages_recovery_param_ops = {
@@ -1600,6 +1603,10 @@ bool kvm_unmap_gfn_range(struct kvm *kvm, struct kvm_gfn_range *range)
 	if (tdp_mmu_enabled)
 		flush = kvm_tdp_mmu_unmap_gfn_range(kvm, range, flush);
 
+	if (kvm_x86_ops.set_apic_access_page_addr &&
+	    range->slot->id == APIC_ACCESS_PAGE_PRIVATE_MEMSLOT)
+		kvm_make_all_cpus_request(kvm, KVM_REQ_APIC_PAGE_RELOAD);
+
 	return flush;
 }
 
@@ -5797,6 +5804,14 @@ static void __kvm_mmu_invalidate_addr(struct kvm_vcpu *vcpu, struct kvm_mmu *mmu
 
 	vcpu_clear_mmio_info(vcpu, addr);
 
+	/*
+	 * Walking and synchronizing SPTEs both assume they are operating in
+	 * the context of the current MMU, and would need to be reworked if
+	 * this is ever used to sync the guest_mmu, e.g. to emulate INVEPT.
+	 */
+	if (WARN_ON_ONCE(mmu != vcpu->arch.mmu))
+		return;
+
 	if (!VALID_PAGE(root_hpa))
 		return;
 
@@ -6844,6 +6859,14 @@ static void mmu_destroy_caches(void)
 	kmem_cache_destroy(mmu_page_header_cache);
 }
 
+static int get_nx_huge_pages(char *buffer, const struct kernel_param *kp)
+{
+	if (nx_hugepage_mitigation_hard_disabled)
+		return sprintf(buffer, "never\n");
+
+	return param_get_bool(buffer, kp);
+}
+
 static bool get_nx_auto_mode(void)
 {
 	/* Return true when CPU has the bug, and mitigations are ON */
@@ -6860,15 +6883,29 @@ static int set_nx_huge_pages(const char *val, const struct kernel_param *kp)
 	bool old_val = nx_huge_pages;
 	bool new_val;
 
+	if (nx_hugepage_mitigation_hard_disabled)
+		return -EPERM;
+
 	/* In "auto" mode deploy workaround only if CPU has the bug. */
-	if (sysfs_streq(val, "off"))
+	if (sysfs_streq(val, "off")) {
 		new_val = 0;
-	else if (sysfs_streq(val, "force"))
+	} else if (sysfs_streq(val, "force")) {
 		new_val = 1;
-	else if (sysfs_streq(val, "auto"))
+	} else if (sysfs_streq(val, "auto")) {
 		new_val = get_nx_auto_mode();
-	else if (kstrtobool(val, &new_val) < 0)
+	} else if (sysfs_streq(val, "never")) {
+		new_val = 0;
+
+		mutex_lock(&kvm_lock);
+		if (!list_empty(&vm_list)) {
+			mutex_unlock(&kvm_lock);
+			return -EBUSY;
+		}
+		nx_hugepage_mitigation_hard_disabled = true;
+		mutex_unlock(&kvm_lock);
+	} else if (kstrtobool(val, &new_val) < 0) {
 		return -EINVAL;
+	}
 
 	__set_nx_huge_pages(new_val);
 
@@ -7006,6 +7043,9 @@ static int set_nx_huge_pages_recovery_param(const char *val, const struct kernel
 	uint old_period, new_period;
 	int err;
 
+	if (nx_hugepage_mitigation_hard_disabled)
+		return -EPERM;
+
 	was_recovery_enabled = calc_nx_huge_pages_recovery_period(&old_period);
 
 	err = param_set_uint(val, kp);
@@ -7164,6 +7204,9 @@ int kvm_mmu_post_init_vm(struct kvm *kvm)
 {
 	int err;
 
+	if (nx_hugepage_mitigation_hard_disabled)
+		return 0;
+
 	err = kvm_vm_create_worker_thread(kvm, kvm_nx_huge_page_recovery_worker, 0,
 					  "kvm-nx-lpage-recovery",
 					  &kvm->arch.nx_huge_page_recovery_thread);

arch/x86/kvm/mmu/tdp_mmu.c

@@ -592,7 +592,10 @@ static inline int tdp_mmu_set_spte_atomic(struct kvm *kvm,
 
 	/*
 	 * Note, fast_pf_fix_direct_spte() can also modify TDP MMU SPTEs and
-	 * does not hold the mmu_lock.
+	 * does not hold the mmu_lock.  On failure, i.e. if a different logical
+	 * CPU modified the SPTE, try_cmpxchg64() updates iter->old_spte with
+	 * the current value, so the caller operates on fresh data, e.g. if it
+	 * retries tdp_mmu_set_spte_atomic()
 	 */
 	if (!try_cmpxchg64(sptep, &iter->old_spte, new_spte))
 		return -EBUSY;

arch/x86/kvm/mtrr.c

@@ -25,10 +25,24 @@
 #define IA32_MTRR_DEF_TYPE_FE		(1ULL << 10)
 #define IA32_MTRR_DEF_TYPE_TYPE_MASK	(0xff)
 
+static bool is_mtrr_base_msr(unsigned int msr)
+{
+	/* MTRR base MSRs use even numbers, masks use odd numbers. */
+	return !(msr & 0x1);
+}
+
+static struct kvm_mtrr_range *var_mtrr_msr_to_range(struct kvm_vcpu *vcpu,
+						    unsigned int msr)
+{
+	int index = (msr - MTRRphysBase_MSR(0)) / 2;
+
+	return &vcpu->arch.mtrr_state.var_ranges[index];
+}
+
 static bool msr_mtrr_valid(unsigned msr)
 {
 	switch (msr) {
-	case 0x200 ... 0x200 + 2 * KVM_NR_VAR_MTRR - 1:
+	case MTRRphysBase_MSR(0) ... MTRRphysMask_MSR(KVM_NR_VAR_MTRR - 1):
 	case MSR_MTRRfix64K_00000:
 	case MSR_MTRRfix16K_80000:
 	case MSR_MTRRfix16K_A0000:
@@ -41,7 +55,6 @@ static bool msr_mtrr_valid(unsigned msr)
 	case MSR_MTRRfix4K_F0000:
 	case MSR_MTRRfix4K_F8000:
 	case MSR_MTRRdefType:
-	case MSR_IA32_CR_PAT:
 		return true;
 	}
 	return false;
@@ -52,7 +65,7 @@ static bool valid_mtrr_type(unsigned t)
 	return t < 8 && (1 << t) & 0x73; /* 0, 1, 4, 5, 6 */
 }
 
-bool kvm_mtrr_valid(struct kvm_vcpu *vcpu, u32 msr, u64 data)
+static bool kvm_mtrr_valid(struct kvm_vcpu *vcpu, u32 msr, u64 data)
 {
 	int i;
 	u64 mask;
@@ -60,9 +73,7 @@ bool kvm_mtrr_valid(struct kvm_vcpu *vcpu, u32 msr, u64 data)
 	if (!msr_mtrr_valid(msr))
 		return false;
 
-	if (msr == MSR_IA32_CR_PAT) {
+	if (msr == MSR_MTRRdefType) {
-		return kvm_pat_valid(data);
-	} else if (msr == MSR_MTRRdefType) {
 		if (data & ~0xcff)
 			return false;
 		return valid_mtrr_type(data & 0xff);
@@ -74,7 +85,8 @@ bool kvm_mtrr_valid(struct kvm_vcpu *vcpu, u32 msr, u64 data)
 	}
 
 	/* variable MTRRs */
-	WARN_ON(!(msr >= 0x200 && msr < 0x200 + 2 * KVM_NR_VAR_MTRR));
+	WARN_ON(!(msr >= MTRRphysBase_MSR(0) &&
+		  msr <= MTRRphysMask_MSR(KVM_NR_VAR_MTRR - 1)));
 
 	mask = kvm_vcpu_reserved_gpa_bits_raw(vcpu);
 	if ((msr & 1) == 0) {
@@ -88,7 +100,6 @@ bool kvm_mtrr_valid(struct kvm_vcpu *vcpu, u32 msr, u64 data)
 
 	return (data & mask) == 0;
 }
-EXPORT_SYMBOL_GPL(kvm_mtrr_valid);
 
 static bool mtrr_is_enabled(struct kvm_mtrr *mtrr_state)
 {
@@ -308,10 +319,8 @@ static void update_mtrr(struct kvm_vcpu *vcpu, u32 msr)
 {
 	struct kvm_mtrr *mtrr_state = &vcpu->arch.mtrr_state;
 	gfn_t start, end;
-	int index;
 
-	if (msr == MSR_IA32_CR_PAT || !tdp_enabled ||
+	if (!tdp_enabled || !kvm_arch_has_noncoherent_dma(vcpu->kvm))
-	      !kvm_arch_has_noncoherent_dma(vcpu->kvm))
 		return;
 
 	if (!mtrr_is_enabled(mtrr_state) && msr != MSR_MTRRdefType)
@@ -326,8 +335,7 @@ static void update_mtrr(struct kvm_vcpu *vcpu, u32 msr)
 		end = ~0ULL;
 	} else {
 		/* variable range MTRRs. */
-		index = (msr - 0x200) / 2;
+		var_mtrr_range(var_mtrr_msr_to_range(vcpu, msr), &start, &end);
-		var_mtrr_range(&mtrr_state->var_ranges[index], &start, &end);
 	}
 
 	kvm_zap_gfn_range(vcpu->kvm, gpa_to_gfn(start), gpa_to_gfn(end));
@@ -342,21 +350,18 @@ static void set_var_mtrr_msr(struct kvm_vcpu *vcpu, u32 msr, u64 data)
 {
 	struct kvm_mtrr *mtrr_state = &vcpu->arch.mtrr_state;
 	struct kvm_mtrr_range *tmp, *cur;
-	int index, is_mtrr_mask;
 
-	index = (msr - 0x200) / 2;
+	cur = var_mtrr_msr_to_range(vcpu, msr);
-	is_mtrr_mask = msr - 0x200 - 2 * index;
-	cur = &mtrr_state->var_ranges[index];
 
 	/* remove the entry if it's in the list. */
 	if (var_mtrr_range_is_valid(cur))
-		list_del(&mtrr_state->var_ranges[index].node);
+		list_del(&cur->node);
 
 	/*
 	 * Set all illegal GPA bits in the mask, since those bits must
 	 * implicitly be 0.  The bits are then cleared when reading them.
 	 */
-	if (!is_mtrr_mask)
+	if (is_mtrr_base_msr(msr))
 		cur->base = data;
 	else
 		cur->mask = data | kvm_vcpu_reserved_gpa_bits_raw(vcpu);
@@ -382,8 +387,6 @@ int kvm_mtrr_set_msr(struct kvm_vcpu *vcpu, u32 msr, u64 data)
 		*(u64 *)&vcpu->arch.mtrr_state.fixed_ranges[index] = data;
 	else if (msr == MSR_MTRRdefType)
 		vcpu->arch.mtrr_state.deftype = data;
-	else if (msr == MSR_IA32_CR_PAT)
-		vcpu->arch.pat = data;
 	else
 		set_var_mtrr_msr(vcpu, msr, data);
 
@@ -411,21 +414,16 @@ int kvm_mtrr_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata)
 		return 1;
 
 	index = fixed_msr_to_range_index(msr);
-	if (index >= 0)
+	if (index >= 0) {
 		*pdata = *(u64 *)&vcpu->arch.mtrr_state.fixed_ranges[index];
-	else if (msr == MSR_MTRRdefType)
+	} else if (msr == MSR_MTRRdefType) {
 		*pdata = vcpu->arch.mtrr_state.deftype;
-	else if (msr == MSR_IA32_CR_PAT)
+	} else {
-		*pdata = vcpu->arch.pat;
+		/* Variable MTRRs */
-	else {	/* Variable MTRRs */
+		if (is_mtrr_base_msr(msr))
-		int is_mtrr_mask;
+			*pdata = var_mtrr_msr_to_range(vcpu, msr)->base;
-
-		index = (msr - 0x200) / 2;
-		is_mtrr_mask = msr - 0x200 - 2 * index;
-		if (!is_mtrr_mask)
-			*pdata = vcpu->arch.mtrr_state.var_ranges[index].base;
 		else
-			*pdata = vcpu->arch.mtrr_state.var_ranges[index].mask;
+			*pdata = var_mtrr_msr_to_range(vcpu, msr)->mask;
 
 		*pdata &= ~kvm_vcpu_reserved_gpa_bits_raw(vcpu);
 	}

arch/x86/kvm/pmu.c

@@ -93,11 +93,6 @@ void kvm_pmu_ops_update(const struct kvm_pmu_ops *pmu_ops)
 #undef __KVM_X86_PMU_OP
 }
 
-static inline bool pmc_is_globally_enabled(struct kvm_pmc *pmc)
-{
-	return static_call(kvm_x86_pmu_pmc_is_enabled)(pmc);
-}
-
 static void kvm_pmi_trigger_fn(struct irq_work *irq_work)
 {
 	struct kvm_pmu *pmu = container_of(irq_work, struct kvm_pmu, irq_work);
@@ -562,6 +557,14 @@ void kvm_pmu_deliver_pmi(struct kvm_vcpu *vcpu)
 
 bool kvm_pmu_is_valid_msr(struct kvm_vcpu *vcpu, u32 msr)
 {
+	switch (msr) {
+	case MSR_CORE_PERF_GLOBAL_STATUS:
+	case MSR_CORE_PERF_GLOBAL_CTRL:
+	case MSR_CORE_PERF_GLOBAL_OVF_CTRL:
+		return kvm_pmu_has_perf_global_ctrl(vcpu_to_pmu(vcpu));
+	default:
+		break;
+	}
 	return static_call(kvm_x86_pmu_msr_idx_to_pmc)(vcpu, msr) ||
 		static_call(kvm_x86_pmu_is_valid_msr)(vcpu, msr);
 }
@@ -577,13 +580,86 @@ static void kvm_pmu_mark_pmc_in_use(struct kvm_vcpu *vcpu, u32 msr)
 
 int kvm_pmu_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 {
-	return static_call(kvm_x86_pmu_get_msr)(vcpu, msr_info);
+	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
+	u32 msr = msr_info->index;
+
+	switch (msr) {
+	case MSR_CORE_PERF_GLOBAL_STATUS:
+	case MSR_AMD64_PERF_CNTR_GLOBAL_STATUS:
+		msr_info->data = pmu->global_status;
+		break;
+	case MSR_AMD64_PERF_CNTR_GLOBAL_CTL:
+	case MSR_CORE_PERF_GLOBAL_CTRL:
+		msr_info->data = pmu->global_ctrl;
+		break;
+	case MSR_AMD64_PERF_CNTR_GLOBAL_STATUS_CLR:
+	case MSR_CORE_PERF_GLOBAL_OVF_CTRL:
+		msr_info->data = 0;
+		break;
+	default:
+		return static_call(kvm_x86_pmu_get_msr)(vcpu, msr_info);
+	}
+
+	return 0;
 }
 
 int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 {
-	kvm_pmu_mark_pmc_in_use(vcpu, msr_info->index);
+	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
-	return static_call(kvm_x86_pmu_set_msr)(vcpu, msr_info);
+	u32 msr = msr_info->index;
+	u64 data = msr_info->data;
+	u64 diff;
+
+	/*
+	 * Note, AMD ignores writes to reserved bits and read-only PMU MSRs,
+	 * whereas Intel generates #GP on attempts to write reserved/RO MSRs.
+	 */
+	switch (msr) {
+	case MSR_CORE_PERF_GLOBAL_STATUS:
+		if (!msr_info->host_initiated)
+			return 1; /* RO MSR */
+		fallthrough;
+	case MSR_AMD64_PERF_CNTR_GLOBAL_STATUS:
+		/* Per PPR, Read-only MSR. Writes are ignored. */
+		if (!msr_info->host_initiated)
+			break;
+
+		if (data & pmu->global_status_mask)
+			return 1;
+
+		pmu->global_status = data;
+		break;
+	case MSR_AMD64_PERF_CNTR_GLOBAL_CTL:
+		data &= ~pmu->global_ctrl_mask;
+		fallthrough;
+	case MSR_CORE_PERF_GLOBAL_CTRL:
+		if (!kvm_valid_perf_global_ctrl(pmu, data))
+			return 1;
+
+		if (pmu->global_ctrl != data) {
+			diff = pmu->global_ctrl ^ data;
+			pmu->global_ctrl = data;
+			reprogram_counters(pmu, diff);
+		}
+		break;
+	case MSR_CORE_PERF_GLOBAL_OVF_CTRL:
+		/*
+		 * GLOBAL_OVF_CTRL, a.k.a. GLOBAL STATUS_RESET, clears bits in
+		 * GLOBAL_STATUS, and so the set of reserved bits is the same.
+		 */
+		if (data & pmu->global_status_mask)
+			return 1;
+		fallthrough;
+	case MSR_AMD64_PERF_CNTR_GLOBAL_STATUS_CLR:
+		if (!msr_info->host_initiated)
+			pmu->global_status &= ~data;
+		break;
+	default:
+		kvm_pmu_mark_pmc_in_use(vcpu, msr_info->index);
+		return static_call(kvm_x86_pmu_set_msr)(vcpu, msr_info);
+	}
+
+	return 0;
 }
 
 /* refresh PMU settings. This function generally is called when underlying

arch/x86/kvm/pmu.h

@@ -20,7 +20,6 @@
 
 struct kvm_pmu_ops {
 	bool (*hw_event_available)(struct kvm_pmc *pmc);
-	bool (*pmc_is_enabled)(struct kvm_pmc *pmc);
 	struct kvm_pmc *(*pmc_idx_to_pmc)(struct kvm_pmu *pmu, int pmc_idx);
 	struct kvm_pmc *(*rdpmc_ecx_to_pmc)(struct kvm_vcpu *vcpu,
 		unsigned int idx, u64 *mask);
@@ -37,10 +36,25 @@ struct kvm_pmu_ops {
 
 	const u64 EVENTSEL_EVENT;
 	const int MAX_NR_GP_COUNTERS;
+	const int MIN_NR_GP_COUNTERS;
 };
 
 void kvm_pmu_ops_update(const struct kvm_pmu_ops *pmu_ops);
 
+static inline bool kvm_pmu_has_perf_global_ctrl(struct kvm_pmu *pmu)
+{
+	/*
+	 * Architecturally, Intel's SDM states that IA32_PERF_GLOBAL_CTRL is
+	 * supported if "CPUID.0AH: EAX[7:0] > 0", i.e. if the PMU version is
+	 * greater than zero.  However, KVM only exposes and emulates the MSR
+	 * to/for the guest if the guest PMU supports at least "Architectural
+	 * Performance Monitoring Version 2".
+	 *
+	 * AMD's version of PERF_GLOBAL_CTRL conveniently shows up with v2.
+	 */
+	return pmu->version > 1;
+}
+
 static inline u64 pmc_bitmask(struct kvm_pmc *pmc)
 {
 	struct kvm_pmu *pmu = pmc_to_pmu(pmc);
@@ -161,6 +175,7 @@ extern struct x86_pmu_capability kvm_pmu_cap;
 static inline void kvm_init_pmu_capability(const struct kvm_pmu_ops *pmu_ops)
 {
 	bool is_intel = boot_cpu_data.x86_vendor == X86_VENDOR_INTEL;
+	int min_nr_gp_ctrs = pmu_ops->MIN_NR_GP_COUNTERS;
 
 	/*
 	 * Hybrid PMUs don't play nice with virtualization without careful
@@ -175,11 +190,15 @@ static inline void kvm_init_pmu_capability(const struct kvm_pmu_ops *pmu_ops)
 		perf_get_x86_pmu_capability(&kvm_pmu_cap);
 
 		/*
-		 * For Intel, only support guest architectural pmu
+		 * WARN if perf did NOT disable hardware PMU if the number of
-		 * on a host with architectural pmu.
+		 * architecturally required GP counters aren't present, i.e. if
+		 * there are a non-zero number of counters, but fewer than what
+		 * is architecturally required.
 		 */
-		if ((is_intel && !kvm_pmu_cap.version) ||
+		if (!kvm_pmu_cap.num_counters_gp ||
-		    !kvm_pmu_cap.num_counters_gp)
+		    WARN_ON_ONCE(kvm_pmu_cap.num_counters_gp < min_nr_gp_ctrs))
+			enable_pmu = false;
+		else if (is_intel && !kvm_pmu_cap.version)
 			enable_pmu = false;
 	}
 
@@ -201,6 +220,33 @@ static inline void kvm_pmu_request_counter_reprogram(struct kvm_pmc *pmc)
 	kvm_make_request(KVM_REQ_PMU, pmc->vcpu);
 }
 
+static inline void reprogram_counters(struct kvm_pmu *pmu, u64 diff)
+{
+	int bit;
+
+	if (!diff)
+		return;
+
+	for_each_set_bit(bit, (unsigned long *)&diff, X86_PMC_IDX_MAX)
+		set_bit(bit, pmu->reprogram_pmi);
+	kvm_make_request(KVM_REQ_PMU, pmu_to_vcpu(pmu));
+}
+
+/*
+ * Check if a PMC is enabled by comparing it against global_ctrl bits.
+ *
+ * If the vPMU doesn't have global_ctrl MSR, all vPMCs are enabled.
+ */
+static inline bool pmc_is_globally_enabled(struct kvm_pmc *pmc)
+{
+	struct kvm_pmu *pmu = pmc_to_pmu(pmc);
+
+	if (!kvm_pmu_has_perf_global_ctrl(pmu))
+		return true;
+
+	return test_bit(pmc->idx, (unsigned long *)&pmu->global_ctrl);
+}
+
 void kvm_pmu_deliver_pmi(struct kvm_vcpu *vcpu);
 void kvm_pmu_handle_event(struct kvm_vcpu *vcpu);
 int kvm_pmu_rdpmc(struct kvm_vcpu *vcpu, unsigned pmc, u64 *data);

arch/x86/kvm/reverse_cpuid.h

@@ -15,6 +15,7 @@ enum kvm_only_cpuid_leafs {
 	CPUID_12_EAX	 = NCAPINTS,
 	CPUID_7_1_EDX,
 	CPUID_8000_0007_EDX,
+	CPUID_8000_0022_EAX,
 	NR_KVM_CPU_CAPS,
 
 	NKVMCAPINTS = NR_KVM_CPU_CAPS - NCAPINTS,
@@ -47,6 +48,9 @@ enum kvm_only_cpuid_leafs {
 /* CPUID level 0x80000007 (EDX). */
 #define KVM_X86_FEATURE_CONSTANT_TSC	KVM_X86_FEATURE(CPUID_8000_0007_EDX, 8)
 
+/* CPUID level 0x80000022 (EAX) */
+#define KVM_X86_FEATURE_PERFMON_V2	KVM_X86_FEATURE(CPUID_8000_0022_EAX, 0)
+
 struct cpuid_reg {
 	u32 function;
 	u32 index;
@@ -74,6 +78,7 @@ static const struct cpuid_reg reverse_cpuid[] = {
 	[CPUID_7_1_EDX]       = {         7, 1, CPUID_EDX},
 	[CPUID_8000_0007_EDX] = {0x80000007, 0, CPUID_EDX},
 	[CPUID_8000_0021_EAX] = {0x80000021, 0, CPUID_EAX},
+	[CPUID_8000_0022_EAX] = {0x80000022, 0, CPUID_EAX},
 };
 
 /*
@@ -108,6 +113,8 @@ static __always_inline u32 __feature_translate(int x86_feature)
 		return KVM_X86_FEATURE_SGX_EDECCSSA;
 	else if (x86_feature == X86_FEATURE_CONSTANT_TSC)
 		return KVM_X86_FEATURE_CONSTANT_TSC;
+	else if (x86_feature == X86_FEATURE_PERFMON_V2)
+		return KVM_X86_FEATURE_PERFMON_V2;
 
 	return x86_feature;
 }

arch/x86/kvm/svm/pmu.c

@@ -78,14 +78,6 @@ static bool amd_hw_event_available(struct kvm_pmc *pmc)
 	return true;
 }
 
-/* check if a PMC is enabled by comparing it against global_ctrl bits. Because
- * AMD CPU doesn't have global_ctrl MSR, all PMCs are enabled (return TRUE).
- */
-static bool amd_pmc_is_enabled(struct kvm_pmc *pmc)
-{
-	return true;
-}
-
 static bool amd_is_valid_rdpmc_ecx(struct kvm_vcpu *vcpu, unsigned int idx)
 {
 	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
@@ -102,12 +94,6 @@ static struct kvm_pmc *amd_rdpmc_ecx_to_pmc(struct kvm_vcpu *vcpu,
 	return amd_pmc_idx_to_pmc(vcpu_to_pmu(vcpu), idx & ~(3u << 30));
 }
 
-static bool amd_is_valid_msr(struct kvm_vcpu *vcpu, u32 msr)
-{
-	/* All MSRs refer to exactly one PMC, so msr_idx_to_pmc is enough.  */
-	return false;
-}
-
 static struct kvm_pmc *amd_msr_idx_to_pmc(struct kvm_vcpu *vcpu, u32 msr)
 {
 	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
@@ -119,6 +105,29 @@ static struct kvm_pmc *amd_msr_idx_to_pmc(struct kvm_vcpu *vcpu, u32 msr)
 	return pmc;
 }
 
+static bool amd_is_valid_msr(struct kvm_vcpu *vcpu, u32 msr)
+{
+	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
+
+	switch (msr) {
+	case MSR_K7_EVNTSEL0 ... MSR_K7_PERFCTR3:
+		return pmu->version > 0;
+	case MSR_F15H_PERF_CTL0 ... MSR_F15H_PERF_CTR5:
+		return guest_cpuid_has(vcpu, X86_FEATURE_PERFCTR_CORE);
+	case MSR_AMD64_PERF_CNTR_GLOBAL_STATUS:
+	case MSR_AMD64_PERF_CNTR_GLOBAL_CTL:
+	case MSR_AMD64_PERF_CNTR_GLOBAL_STATUS_CLR:
+		return pmu->version > 1;
+	default:
+		if (msr > MSR_F15H_PERF_CTR5 &&
+		    msr < MSR_F15H_PERF_CTL0 + 2 * pmu->nr_arch_gp_counters)
+			return pmu->version > 1;
+		break;
+	}
+
+	return amd_msr_idx_to_pmc(vcpu, msr);
+}
+
 static int amd_pmu_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 {
 	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
@@ -172,20 +181,39 @@ static int amd_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 static void amd_pmu_refresh(struct kvm_vcpu *vcpu)
 {
 	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
+	union cpuid_0x80000022_ebx ebx;
 
-	if (guest_cpuid_has(vcpu, X86_FEATURE_PERFCTR_CORE))
+	pmu->version = 1;
+	if (guest_cpuid_has(vcpu, X86_FEATURE_PERFMON_V2)) {
+		pmu->version = 2;
+		/*
+		 * Note, PERFMON_V2 is also in 0x80000022.0x0, i.e. the guest
+		 * CPUID entry is guaranteed to be non-NULL.
+		 */
+		BUILD_BUG_ON(x86_feature_cpuid(X86_FEATURE_PERFMON_V2).function != 0x80000022 ||
+			     x86_feature_cpuid(X86_FEATURE_PERFMON_V2).index);
+		ebx.full = kvm_find_cpuid_entry_index(vcpu, 0x80000022, 0)->ebx;
+		pmu->nr_arch_gp_counters = ebx.split.num_core_pmc;
+	} else if (guest_cpuid_has(vcpu, X86_FEATURE_PERFCTR_CORE)) {
 		pmu->nr_arch_gp_counters = AMD64_NUM_COUNTERS_CORE;
-	else
+	} else {
 		pmu->nr_arch_gp_counters = AMD64_NUM_COUNTERS;
+	}
+
+	pmu->nr_arch_gp_counters = min_t(unsigned int, pmu->nr_arch_gp_counters,
+					 kvm_pmu_cap.num_counters_gp);
+
+	if (pmu->version > 1) {
+		pmu->global_ctrl_mask = ~((1ull << pmu->nr_arch_gp_counters) - 1);
+		pmu->global_status_mask = pmu->global_ctrl_mask;
+	}
 
 	pmu->counter_bitmask[KVM_PMC_GP] = ((u64)1 << 48) - 1;
 	pmu->reserved_bits = 0xfffffff000280000ull;
 	pmu->raw_event_mask = AMD64_RAW_EVENT_MASK;
-	pmu->version = 1;
 	/* not applicable to AMD; but clean them to prevent any fall out */
 	pmu->counter_bitmask[KVM_PMC_FIXED] = 0;
 	pmu->nr_arch_fixed_counters = 0;
-	pmu->global_status = 0;
 	bitmap_set(pmu->all_valid_pmc_idx, 0, pmu->nr_arch_gp_counters);
 }
 
@@ -216,11 +244,12 @@ static void amd_pmu_reset(struct kvm_vcpu *vcpu)
 		pmc_stop_counter(pmc);
 		pmc->counter = pmc->prev_counter = pmc->eventsel = 0;
 	}
+
+	pmu->global_ctrl = pmu->global_status = 0;
 }
 
 struct kvm_pmu_ops amd_pmu_ops __initdata = {
 	.hw_event_available = amd_hw_event_available,
-	.pmc_is_enabled = amd_pmc_is_enabled,
 	.pmc_idx_to_pmc = amd_pmc_idx_to_pmc,
 	.rdpmc_ecx_to_pmc = amd_rdpmc_ecx_to_pmc,
 	.msr_idx_to_pmc = amd_msr_idx_to_pmc,
@@ -233,4 +262,5 @@ struct kvm_pmu_ops amd_pmu_ops __initdata = {
 	.reset = amd_pmu_reset,
 	.EVENTSEL_EVENT = AMD64_EVENTSEL_EVENT,
 	.MAX_NR_GP_COUNTERS = KVM_AMD_PMC_MAX_GENERIC,
+	.MIN_NR_GP_COUNTERS = AMD64_NUM_COUNTERS,
 };

arch/x86/kvm/svm/sev.c

@@ -2216,10 +2216,7 @@ void __init sev_hardware_setup(void)
 	}
 
 	sev_asid_count = max_sev_asid - min_sev_asid + 1;
-	if (misc_cg_set_capacity(MISC_CG_RES_SEV, sev_asid_count))
+	WARN_ON_ONCE(misc_cg_set_capacity(MISC_CG_RES_SEV, sev_asid_count));
-		goto out;
-
-	pr_info("SEV supported: %u ASIDs\n", sev_asid_count);
 	sev_supported = true;
 
 	/* SEV-ES support requested? */
@@ -2244,13 +2241,19 @@ void __init sev_hardware_setup(void)
 		goto out;
 
 	sev_es_asid_count = min_sev_asid - 1;
-	if (misc_cg_set_capacity(MISC_CG_RES_SEV_ES, sev_es_asid_count))
+	WARN_ON_ONCE(misc_cg_set_capacity(MISC_CG_RES_SEV_ES, sev_es_asid_count));
-		goto out;
-
-	pr_info("SEV-ES supported: %u ASIDs\n", sev_es_asid_count);
 	sev_es_supported = true;
 
 out:
+	if (boot_cpu_has(X86_FEATURE_SEV))
+		pr_info("SEV %s (ASIDs %u - %u)\n",
+			sev_supported ? "enabled" : "disabled",
+			min_sev_asid, max_sev_asid);
+	if (boot_cpu_has(X86_FEATURE_SEV_ES))
+		pr_info("SEV-ES %s (ASIDs %u - %u)\n",
+			sev_es_supported ? "enabled" : "disabled",
+			min_sev_asid > 1 ? 1 : 0, min_sev_asid - 1);
+
 	sev_enabled = sev_supported;
 	sev_es_enabled = sev_es_supported;
 #endif

arch/x86/kvm/svm/svm.c

@@ -244,15 +244,6 @@ static u8 rsm_ins_bytes[] = "\x0f\xaa";
 
 static unsigned long iopm_base;
 
-struct kvm_ldttss_desc {
-	u16 limit0;
-	u16 base0;
-	unsigned base1:8, type:5, dpl:2, p:1;
-	unsigned limit1:4, zero0:3, g:1, base2:8;
-	u32 base3;
-	u32 zero1;
-} __attribute__((packed));
-
 DEFINE_PER_CPU(struct svm_cpu_data, svm_data);
 
 /*
@@ -588,7 +579,6 @@ static int svm_hardware_enable(void)
 
 	struct svm_cpu_data *sd;
 	uint64_t efer;
-	struct desc_struct *gdt;
 	int me = raw_smp_processor_id();
 
 	rdmsrl(MSR_EFER, efer);
@@ -601,9 +591,6 @@ static int svm_hardware_enable(void)
 	sd->next_asid = sd->max_asid + 1;
 	sd->min_asid = max_sev_asid + 1;
 
-	gdt = get_current_gdt_rw();
-	sd->tss_desc = (struct kvm_ldttss_desc *)(gdt + GDT_ENTRY_TSS);
-
 	wrmsrl(MSR_EFER, efer | EFER_SVME);
 
 	wrmsrl(MSR_VM_HSAVE_PA, sd->save_area_pa);
@@ -752,7 +739,7 @@ static bool msr_write_intercepted(struct kvm_vcpu *vcpu, u32 msr)
 
 	BUG_ON(offset == MSR_INVALID);
 
-	return !!test_bit(bit_write,  &tmp);
+	return test_bit(bit_write, &tmp);
 }
 
 static void set_msr_interception_bitmap(struct kvm_vcpu *vcpu, u32 *msrpm,
@@ -2939,9 +2926,10 @@ static int svm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr)
 
 		break;
 	case MSR_IA32_CR_PAT:
-		if (!kvm_mtrr_valid(vcpu, MSR_IA32_CR_PAT, data))
+		ret = kvm_set_msr_common(vcpu, msr);
-			return 1;
+		if (ret)
-		vcpu->arch.pat = data;
+			break;
+
 		svm->vmcb01.ptr->save.g_pat = data;
 		if (is_guest_mode(vcpu))
 			nested_vmcb02_compute_g_pat(svm);
@@ -3418,8 +3406,6 @@ static int svm_handle_exit(struct kvm_vcpu *vcpu, fastpath_t exit_fastpath)
 	struct kvm_run *kvm_run = vcpu->run;
 	u32 exit_code = svm->vmcb->control.exit_code;
 
-	trace_kvm_exit(vcpu, KVM_ISA_SVM);
-
 	/* SEV-ES guests must use the CR write traps to track CR registers. */
 	if (!sev_es_guest(vcpu->kvm)) {
 		if (!svm_is_intercept(svm, INTERCEPT_CR0_WRITE))
@@ -3457,14 +3443,6 @@ static int svm_handle_exit(struct kvm_vcpu *vcpu, fastpath_t exit_fastpath)
 	return svm_invoke_exit_handler(vcpu, exit_code);
 }
 
-static void reload_tss(struct kvm_vcpu *vcpu)
-{
-	struct svm_cpu_data *sd = per_cpu_ptr(&svm_data, vcpu->cpu);
-
-	sd->tss_desc->type = 9; /* available 32/64-bit TSS */
-	load_TR_desc();
-}
-
 static void pre_svm_run(struct kvm_vcpu *vcpu)
 {
 	struct svm_cpu_data *sd = per_cpu_ptr(&svm_data, vcpu->cpu);
@@ -4099,9 +4077,6 @@ static __no_kcsan fastpath_t svm_vcpu_run(struct kvm_vcpu *vcpu)
 
 	svm_vcpu_enter_exit(vcpu, spec_ctrl_intercepted);
 
-	if (!sev_es_guest(vcpu->kvm))
-		reload_tss(vcpu);
-
 	if (!static_cpu_has(X86_FEATURE_V_SPEC_CTRL))
 		x86_spec_ctrl_restore_host(svm->virt_spec_ctrl);
 
@@ -4156,6 +4131,8 @@ static __no_kcsan fastpath_t svm_vcpu_run(struct kvm_vcpu *vcpu)
 		     SVM_EXIT_EXCP_BASE + MC_VECTOR))
 		svm_handle_mce(vcpu);
 
+	trace_kvm_exit(vcpu, KVM_ISA_SVM);
+
 	svm_complete_interrupts(vcpu);
 
 	if (is_guest_mode(vcpu))
@@ -5025,9 +5002,22 @@ static __init void svm_set_cpu_caps(void)
 	    boot_cpu_has(X86_FEATURE_AMD_SSBD))
 		kvm_cpu_cap_set(X86_FEATURE_VIRT_SSBD);
 
-	/* AMD PMU PERFCTR_CORE CPUID */
+	if (enable_pmu) {
-	if (enable_pmu && boot_cpu_has(X86_FEATURE_PERFCTR_CORE))
+		/*
-		kvm_cpu_cap_set(X86_FEATURE_PERFCTR_CORE);
+		 * Enumerate support for PERFCTR_CORE if and only if KVM has
+		 * access to enough counters to virtualize "core" support,
+		 * otherwise limit vPMU support to the legacy number of counters.
+		 */
+		if (kvm_pmu_cap.num_counters_gp < AMD64_NUM_COUNTERS_CORE)
+			kvm_pmu_cap.num_counters_gp = min(AMD64_NUM_COUNTERS,
+							  kvm_pmu_cap.num_counters_gp);
+		else
+			kvm_cpu_cap_check_and_set(X86_FEATURE_PERFCTR_CORE);
+
+		if (kvm_pmu_cap.version != 2 ||
+		    !kvm_cpu_cap_has(X86_FEATURE_PERFCTR_CORE))
+			kvm_cpu_cap_clear(X86_FEATURE_PERFMON_V2);
+	}
 
 	/* CPUID 0x8000001F (SME/SEV features) */
 	sev_set_cpu_caps();

arch/x86/kvm/svm/svm.h

@@ -303,7 +303,6 @@ struct svm_cpu_data {
 	u32 max_asid;
 	u32 next_asid;
 	u32 min_asid;
-	struct kvm_ldttss_desc *tss_desc;
 
 	struct page *save_area;
 	unsigned long save_area_pa;

arch/x86/kvm/vmx/capabilities.h

@@ -152,8 +152,8 @@ static inline bool cpu_has_vmx_ept(void)
 
 static inline bool vmx_umip_emulated(void)
 {
-	return vmcs_config.cpu_based_2nd_exec_ctrl &
+	return !boot_cpu_has(X86_FEATURE_UMIP) &&
-		SECONDARY_EXEC_DESC;
+	       (vmcs_config.cpu_based_2nd_exec_ctrl & SECONDARY_EXEC_DESC);
 }
 
 static inline bool cpu_has_vmx_rdtscp(void)

arch/x86/kvm/vmx/nested.c

@@ -2328,8 +2328,7 @@ static void prepare_vmcs02_early(struct vcpu_vmx *vmx, struct loaded_vmcs *vmcs0
 		 * Preset *DT exiting when emulating UMIP, so that vmx_set_cr4()
 		 * will not have to rewrite the controls just for this bit.
 		 */
-		if (!boot_cpu_has(X86_FEATURE_UMIP) && vmx_umip_emulated() &&
+		if (vmx_umip_emulated() && (vmcs12->guest_cr4 & X86_CR4_UMIP))
-		    (vmcs12->guest_cr4 & X86_CR4_UMIP))
 			exec_control |= SECONDARY_EXEC_DESC;
 
 		if (exec_control & SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY)
@@ -2649,7 +2648,7 @@ static int prepare_vmcs02(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
 	}
 
 	if ((vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL) &&
-	    intel_pmu_has_perf_global_ctrl(vcpu_to_pmu(vcpu)) &&
+	    kvm_pmu_has_perf_global_ctrl(vcpu_to_pmu(vcpu)) &&
 	    WARN_ON_ONCE(kvm_set_msr(vcpu, MSR_CORE_PERF_GLOBAL_CTRL,
 				     vmcs12->guest_ia32_perf_global_ctrl))) {
 		*entry_failure_code = ENTRY_FAIL_DEFAULT;
@@ -4524,7 +4523,7 @@ static void load_vmcs12_host_state(struct kvm_vcpu *vcpu,
 		vcpu->arch.pat = vmcs12->host_ia32_pat;
 	}
 	if ((vmcs12->vm_exit_controls & VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL) &&
-	    intel_pmu_has_perf_global_ctrl(vcpu_to_pmu(vcpu)))
+	    kvm_pmu_has_perf_global_ctrl(vcpu_to_pmu(vcpu)))
 		WARN_ON_ONCE(kvm_set_msr(vcpu, MSR_CORE_PERF_GLOBAL_CTRL,
 					 vmcs12->host_ia32_perf_global_ctrl));
 

arch/x86/kvm/vmx/pmu_intel.c

@@ -73,18 +73,6 @@ static struct kvm_pmc *intel_pmc_idx_to_pmc(struct kvm_pmu *pmu, int pmc_idx)
 	}
 }
 
-static void reprogram_counters(struct kvm_pmu *pmu, u64 diff)
-{
-	int bit;
-
-	if (!diff)
-		return;
-
-	for_each_set_bit(bit, (unsigned long *)&diff, X86_PMC_IDX_MAX)
-		set_bit(bit, pmu->reprogram_pmi);
-	kvm_make_request(KVM_REQ_PMU, pmu_to_vcpu(pmu));
-}
-
 static bool intel_hw_event_available(struct kvm_pmc *pmc)
 {
 	struct kvm_pmu *pmu = pmc_to_pmu(pmc);
@@ -107,17 +95,6 @@ static bool intel_hw_event_available(struct kvm_pmc *pmc)
 	return true;
 }
 
-/* check if a PMC is enabled by comparing it with globl_ctrl bits. */
-static bool intel_pmc_is_enabled(struct kvm_pmc *pmc)
-{
-	struct kvm_pmu *pmu = pmc_to_pmu(pmc);
-
-	if (!intel_pmu_has_perf_global_ctrl(pmu))
-		return true;
-
-	return test_bit(pmc->idx, (unsigned long *)&pmu->global_ctrl);
-}
-
 static bool intel_is_valid_rdpmc_ecx(struct kvm_vcpu *vcpu, unsigned int idx)
 {
 	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
@@ -198,11 +175,7 @@ static bool intel_is_valid_msr(struct kvm_vcpu *vcpu, u32 msr)
 
 	switch (msr) {
 	case MSR_CORE_PERF_FIXED_CTR_CTRL:
-	case MSR_CORE_PERF_GLOBAL_STATUS:
+		return kvm_pmu_has_perf_global_ctrl(pmu);
-	case MSR_CORE_PERF_GLOBAL_CTRL:
-	case MSR_CORE_PERF_GLOBAL_OVF_CTRL:
-		return intel_pmu_has_perf_global_ctrl(pmu);
-		break;
 	case MSR_IA32_PEBS_ENABLE:
 		ret = vcpu_get_perf_capabilities(vcpu) & PERF_CAP_PEBS_FORMAT;
 		break;
@@ -352,15 +325,6 @@ static int intel_pmu_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 	case MSR_CORE_PERF_FIXED_CTR_CTRL:
 		msr_info->data = pmu->fixed_ctr_ctrl;
 		break;
-	case MSR_CORE_PERF_GLOBAL_STATUS:
-		msr_info->data = pmu->global_status;
-		break;
-	case MSR_CORE_PERF_GLOBAL_CTRL:
-		msr_info->data = pmu->global_ctrl;
-		break;
-	case MSR_CORE_PERF_GLOBAL_OVF_CTRL:
-		msr_info->data = 0;
-		break;
 	case MSR_IA32_PEBS_ENABLE:
 		msr_info->data = pmu->pebs_enable;
 		break;
@@ -410,29 +374,6 @@ static int intel_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 		if (pmu->fixed_ctr_ctrl != data)
 			reprogram_fixed_counters(pmu, data);
 		break;
-	case MSR_CORE_PERF_GLOBAL_STATUS:
-		if (!msr_info->host_initiated)
-			return 1; /* RO MSR */
-
-		pmu->global_status = data;
-		break;
-	case MSR_CORE_PERF_GLOBAL_CTRL:
-		if (!kvm_valid_perf_global_ctrl(pmu, data))
-			return 1;
-
-		if (pmu->global_ctrl != data) {
-			diff = pmu->global_ctrl ^ data;
-			pmu->global_ctrl = data;
-			reprogram_counters(pmu, diff);
-		}
-		break;
-	case MSR_CORE_PERF_GLOBAL_OVF_CTRL:
-		if (data & pmu->global_ovf_ctrl_mask)
-			return 1;
-
-		if (!msr_info->host_initiated)
-			pmu->global_status &= ~data;
-		break;
 	case MSR_IA32_PEBS_ENABLE:
 		if (data & pmu->pebs_enable_mask)
 			return 1;
@@ -444,8 +385,6 @@ static int intel_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 		}
 		break;
 	case MSR_IA32_DS_AREA:
-		if (msr_info->host_initiated && data && !guest_cpuid_has(vcpu, X86_FEATURE_DS))
-			return 1;
 		if (is_noncanonical_address(data, vcpu))
 			return 1;
 
@@ -531,7 +470,7 @@ static void intel_pmu_refresh(struct kvm_vcpu *vcpu)
 	pmu->reserved_bits = 0xffffffff00200000ull;
 	pmu->raw_event_mask = X86_RAW_EVENT_MASK;
 	pmu->global_ctrl_mask = ~0ull;
-	pmu->global_ovf_ctrl_mask = ~0ull;
+	pmu->global_status_mask = ~0ull;
 	pmu->fixed_ctr_ctrl_mask = ~0ull;
 	pmu->pebs_enable_mask = ~0ull;
 	pmu->pebs_data_cfg_mask = ~0ull;
@@ -585,11 +524,17 @@ static void intel_pmu_refresh(struct kvm_vcpu *vcpu)
 	counter_mask = ~(((1ull << pmu->nr_arch_gp_counters) - 1) |
 		(((1ull << pmu->nr_arch_fixed_counters) - 1) << INTEL_PMC_IDX_FIXED));
 	pmu->global_ctrl_mask = counter_mask;
-	pmu->global_ovf_ctrl_mask = pmu->global_ctrl_mask
+
+	/*
+	 * GLOBAL_STATUS and GLOBAL_OVF_CONTROL (a.k.a. GLOBAL_STATUS_RESET)
+	 * share reserved bit definitions.  The kernel just happens to use
+	 * OVF_CTRL for the names.
+	 */
+	pmu->global_status_mask = pmu->global_ctrl_mask
 			& ~(MSR_CORE_PERF_GLOBAL_OVF_CTRL_OVF_BUF |
 			    MSR_CORE_PERF_GLOBAL_OVF_CTRL_COND_CHGD);
 	if (vmx_pt_mode_is_host_guest())
-		pmu->global_ovf_ctrl_mask &=
+		pmu->global_status_mask &=
 				~MSR_CORE_PERF_GLOBAL_OVF_CTRL_TRACE_TOPA_PMI;
 
 	entry = kvm_find_cpuid_entry_index(vcpu, 7, 0);
@@ -801,7 +746,7 @@ void intel_pmu_cross_mapped_check(struct kvm_pmu *pmu)
 		pmc = intel_pmc_idx_to_pmc(pmu, bit);
 
 		if (!pmc || !pmc_speculative_in_use(pmc) ||
-		    !intel_pmc_is_enabled(pmc) || !pmc->perf_event)
+		    !pmc_is_globally_enabled(pmc) || !pmc->perf_event)
 			continue;
 
 		/*
@@ -816,7 +761,6 @@ void intel_pmu_cross_mapped_check(struct kvm_pmu *pmu)
 
 struct kvm_pmu_ops intel_pmu_ops __initdata = {
 	.hw_event_available = intel_hw_event_available,
-	.pmc_is_enabled = intel_pmc_is_enabled,
 	.pmc_idx_to_pmc = intel_pmc_idx_to_pmc,
 	.rdpmc_ecx_to_pmc = intel_rdpmc_ecx_to_pmc,
 	.msr_idx_to_pmc = intel_msr_idx_to_pmc,
@@ -831,4 +775,5 @@ struct kvm_pmu_ops intel_pmu_ops __initdata = {
 	.cleanup = intel_pmu_cleanup,
 	.EVENTSEL_EVENT = ARCH_PERFMON_EVENTSEL_EVENT,
 	.MAX_NR_GP_COUNTERS = KVM_INTEL_PMC_MAX_GENERIC,
+	.MIN_NR_GP_COUNTERS = 1,
 };

arch/x86/kvm/vmx/sgx.c

@@ -357,11 +357,12 @@ static int handle_encls_einit(struct kvm_vcpu *vcpu)
 
 static inline bool encls_leaf_enabled_in_guest(struct kvm_vcpu *vcpu, u32 leaf)
 {
-	if (!enable_sgx || !guest_cpuid_has(vcpu, X86_FEATURE_SGX))
+	/*
-		return false;
+	 * ENCLS generates a #UD if SGX1 isn't supported, i.e. this point will
-
+	 * be reached if and only if the SGX1 leafs are enabled.
+	 */
 	if (leaf >= ECREATE && leaf <= ETRACK)
-		return guest_cpuid_has(vcpu, X86_FEATURE_SGX1);
+		return true;
 
 	if (leaf >= EAUG && leaf <= EMODT)
 		return guest_cpuid_has(vcpu, X86_FEATURE_SGX2);
@@ -380,9 +381,11 @@ int handle_encls(struct kvm_vcpu *vcpu)
 {
 	u32 leaf = (u32)kvm_rax_read(vcpu);
 
-	if (!encls_leaf_enabled_in_guest(vcpu, leaf)) {
+	if (!enable_sgx || !guest_cpuid_has(vcpu, X86_FEATURE_SGX) ||
+	    !guest_cpuid_has(vcpu, X86_FEATURE_SGX1)) {
 		kvm_queue_exception(vcpu, UD_VECTOR);
-	} else if (!sgx_enabled_in_guest_bios(vcpu)) {
+	} else if (!encls_leaf_enabled_in_guest(vcpu, leaf) ||
+		   !sgx_enabled_in_guest_bios(vcpu) || !is_paging(vcpu)) {
 		kvm_inject_gp(vcpu, 0);
 	} else {
 		if (leaf == ECREATE)

arch/x86/kvm/vmx/vmenter.S

@@ -187,7 +187,7 @@ SYM_FUNC_START(__vmx_vcpu_run)
 	_ASM_EXTABLE(.Lvmresume, .Lfixup)
 	_ASM_EXTABLE(.Lvmlaunch, .Lfixup)
 
-SYM_INNER_LABEL(vmx_vmexit, SYM_L_GLOBAL)
+SYM_INNER_LABEL_ALIGN(vmx_vmexit, SYM_L_GLOBAL)
 
 	/* Restore unwind state from before the VMRESUME/VMLAUNCH. */
 	UNWIND_HINT_RESTORE

arch/x86/kvm/vmx/vmx.c

@@ -2287,19 +2287,16 @@ static int vmx_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 			return 1;
 		goto find_uret_msr;
 	case MSR_IA32_CR_PAT:
-		if (!kvm_pat_valid(data))
+		ret = kvm_set_msr_common(vcpu, msr_info);
-			return 1;
+		if (ret)
+			break;
 
 		if (is_guest_mode(vcpu) &&
 		    get_vmcs12(vcpu)->vm_exit_controls & VM_EXIT_SAVE_IA32_PAT)
 			get_vmcs12(vcpu)->guest_ia32_pat = data;
 
-		if (vmcs_config.vmentry_ctrl & VM_ENTRY_LOAD_IA32_PAT) {
+		if (vmcs_config.vmentry_ctrl & VM_ENTRY_LOAD_IA32_PAT)
 			vmcs_write64(GUEST_IA32_PAT, data);
-			vcpu->arch.pat = data;
-			break;
-		}
-		ret = kvm_set_msr_common(vcpu, msr_info);
 		break;
 	case MSR_IA32_MCG_EXT_CTL:
 		if ((!msr_info->host_initiated &&
@@ -3387,15 +3384,15 @@ static bool vmx_is_valid_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 
 void vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 {
-	unsigned long old_cr4 = vcpu->arch.cr4;
+	unsigned long old_cr4 = kvm_read_cr4(vcpu);
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
+	unsigned long hw_cr4;
+
 	/*
 	 * Pass through host's Machine Check Enable value to hw_cr4, which
 	 * is in force while we are in guest mode.  Do not let guests control
 	 * this bit, even if host CR4.MCE == 0.
 	 */
-	unsigned long hw_cr4;
-
 	hw_cr4 = (cr4_read_shadow() & X86_CR4_MCE) | (cr4 & ~X86_CR4_MCE);
 	if (is_unrestricted_guest(vcpu))
 		hw_cr4 |= KVM_VM_CR4_ALWAYS_ON_UNRESTRICTED_GUEST;
@@ -3404,7 +3401,7 @@ void vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 	else
 		hw_cr4 |= KVM_PMODE_VM_CR4_ALWAYS_ON;
 
-	if (!boot_cpu_has(X86_FEATURE_UMIP) && vmx_umip_emulated()) {
+	if (vmx_umip_emulated()) {
 		if (cr4 & X86_CR4_UMIP) {
 			secondary_exec_controls_setbit(vmx, SECONDARY_EXEC_DESC);
 			hw_cr4 &= ~X86_CR4_UMIP;
@@ -5402,7 +5399,13 @@ static int handle_set_cr4(struct kvm_vcpu *vcpu, unsigned long val)
 
 static int handle_desc(struct kvm_vcpu *vcpu)
 {
-	WARN_ON(!(vcpu->arch.cr4 & X86_CR4_UMIP));
+	/*
+	 * UMIP emulation relies on intercepting writes to CR4.UMIP, i.e. this
+	 * and other code needs to be updated if UMIP can be guest owned.
+	 */
+	BUILD_BUG_ON(KVM_POSSIBLE_CR4_GUEST_BITS & X86_CR4_UMIP);
+
+	WARN_ON_ONCE(!kvm_is_cr4_bit_set(vcpu, X86_CR4_UMIP));
 	return kvm_emulate_instruction(vcpu, 0);
 }
 
@@ -6708,7 +6711,12 @@ void vmx_set_virtual_apic_mode(struct kvm_vcpu *vcpu)
 
 static void vmx_set_apic_access_page_addr(struct kvm_vcpu *vcpu)
 {
-	struct page *page;
+	const gfn_t gfn = APIC_DEFAULT_PHYS_BASE >> PAGE_SHIFT;
+	struct kvm *kvm = vcpu->kvm;
+	struct kvm_memslots *slots = kvm_memslots(kvm);
+	struct kvm_memory_slot *slot;
+	unsigned long mmu_seq;
+	kvm_pfn_t pfn;
 
 	/* Defer reload until vmcs01 is the current VMCS. */
 	if (is_guest_mode(vcpu)) {
@@ -6720,18 +6728,53 @@ static void vmx_set_apic_access_page_addr(struct kvm_vcpu *vcpu)
 	    SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES))
 		return;
 
-	page = gfn_to_page(vcpu->kvm, APIC_DEFAULT_PHYS_BASE >> PAGE_SHIFT);
+	/*
-	if (is_error_page(page))
+	 * Grab the memslot so that the hva lookup for the mmu_notifier retry
+	 * is guaranteed to use the same memslot as the pfn lookup, i.e. rely
+	 * on the pfn lookup's validation of the memslot to ensure a valid hva
+	 * is used for the retry check.
+	 */
+	slot = id_to_memslot(slots, APIC_ACCESS_PAGE_PRIVATE_MEMSLOT);
+	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
 		return;
 
-	vmcs_write64(APIC_ACCESS_ADDR, page_to_phys(page));
+	/*
+	 * Ensure that the mmu_notifier sequence count is read before KVM
+	 * retrieves the pfn from the primary MMU.  Note, the memslot is
+	 * protected by SRCU, not the mmu_notifier.  Pairs with the smp_wmb()
+	 * in kvm_mmu_invalidate_end().
+	 */
+	mmu_seq = kvm->mmu_invalidate_seq;
+	smp_rmb();
+
+	/*
+	 * No need to retry if the memslot does not exist or is invalid.  KVM
+	 * controls the APIC-access page memslot, and only deletes the memslot
+	 * if APICv is permanently inhibited, i.e. the memslot won't reappear.
+	 */
+	pfn = gfn_to_pfn_memslot(slot, gfn);
+	if (is_error_noslot_pfn(pfn))
+		return;
+
+	read_lock(&vcpu->kvm->mmu_lock);
+	if (mmu_invalidate_retry_hva(kvm, mmu_seq,
+				     gfn_to_hva_memslot(slot, gfn))) {
+		kvm_make_request(KVM_REQ_APIC_PAGE_RELOAD, vcpu);
+		read_unlock(&vcpu->kvm->mmu_lock);
+		goto out;
+	}
+
+	vmcs_write64(APIC_ACCESS_ADDR, pfn_to_hpa(pfn));
+	read_unlock(&vcpu->kvm->mmu_lock);
+
 	vmx_flush_tlb_current(vcpu);
 
+out:
 	/*
 	 * Do not pin apic access page in memory, the MMU notifier
 	 * will call us again if it is migrated or swapped out.
 	 */
-	put_page(page);
+	kvm_release_pfn_clean(pfn);
 }
 
 static void vmx_hwapic_isr_update(int max_isr)

arch/x86/kvm/vmx/vmx.h

@@ -93,18 +93,6 @@ union vmx_exit_reason {
 	u32 full;
 };
 
-static inline bool intel_pmu_has_perf_global_ctrl(struct kvm_pmu *pmu)
-{
-	/*
-	 * Architecturally, Intel's SDM states that IA32_PERF_GLOBAL_CTRL is
-	 * supported if "CPUID.0AH: EAX[7:0] > 0", i.e. if the PMU version is
-	 * greater than zero.  However, KVM only exposes and emulates the MSR
-	 * to/for the guest if the guest PMU supports at least "Architectural
-	 * Performance Monitoring Version 2".
-	 */
-	return pmu->version > 1;
-}
-
 struct lbr_desc {
 	/* Basic info about guest LBR records. */
 	struct x86_pmu_lbr records;