Re: [PATCH 27/30] KVM: arm64: Add some initial documentation for pKVM

[Vincent Donnefort <vdonnefort@google.com>]

On Mon, Jan 05, 2026 at 03:49:35PM +0000, Will Deacon wrote:
> Add some initial documentation for pKVM to help people understand what
> is supported, the limitations of protected VMs when compared to
> non-protected VMs and also what is left to do.
> 
> Signed-off-by: Will Deacon <will@kernel.org>
> ---
>  .../admin-guide/kernel-parameters.txt         |   4 +-
>  Documentation/virt/kvm/arm/index.rst          |   1 +
>  Documentation/virt/kvm/arm/pkvm.rst           | 101 ++++++++++++++++++
>  3 files changed, 104 insertions(+), 2 deletions(-)
>  create mode 100644 Documentation/virt/kvm/arm/pkvm.rst
> 
> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
> index a8d0afde7f85..9939dc5654d2 100644
> --- a/Documentation/admin-guide/kernel-parameters.txt
> +++ b/Documentation/admin-guide/kernel-parameters.txt
> @@ -3141,8 +3141,8 @@ Kernel parameters
>  			for the host. To force nVHE on VHE hardware, add
>  			"arm64_sw.hvhe=0 id_aa64mmfr1.vh=0" to the
>  			command-line.
> -			"nested" is experimental and should be used with
> -			extreme caution.
> +			"nested" and "protected" are experimental and should be
> +			used with extreme caution.
>  
>  	kvm-arm.vgic_v3_group0_trap=
>  			[KVM,ARM,EARLY] Trap guest accesses to GICv3 group-0
> diff --git a/Documentation/virt/kvm/arm/index.rst b/Documentation/virt/kvm/arm/index.rst
> index ec09881de4cf..0856b4942e05 100644
> --- a/Documentation/virt/kvm/arm/index.rst
> +++ b/Documentation/virt/kvm/arm/index.rst
> @@ -10,6 +10,7 @@ ARM
>     fw-pseudo-registers
>     hyp-abi
>     hypercalls
> +   pkvm
>     pvtime
>     ptp_kvm
>     vcpu-features
> diff --git a/Documentation/virt/kvm/arm/pkvm.rst b/Documentation/virt/kvm/arm/pkvm.rst
> new file mode 100644
> index 000000000000..1882bde8cc0b
> --- /dev/null
> +++ b/Documentation/virt/kvm/arm/pkvm.rst
> @@ -0,0 +1,101 @@
> +.. SPDX-License-Identifier: GPL-2.0
> +
> +====================
> +Protected KVM (pKVM)
> +====================
> +
> +**NOTE**: pKVM is currently an experimental, development feature and
> +subject to breaking changes as new isolation features are implemented.
> +Please reach out to the developers at kvmarm@lists.linux.dev if you have
> +any questions.
> +
> +Overview
> +========
> +
> +Booting a host kernel with '``kvm-arm.mode=protected``' enables
> +"Protected KVM" (pKVM). During boot, pKVM installs a stage-2 identity
> +map page-table for the host and uses it to isolate the hypervisor
> +running at EL2 from the rest of the host running at EL1/0.
> +
> +If ``CONFIG_PROTECTED_VM_UAPI=y``, pKVM permits creation of protected
> +virtual machines (pVMs) by passing the ``KVM_VM_TYPE_ARM_PROTECTED``
> +machine type identifier to the ``KVM_CREATE_VM`` ioctl(). The hypervisor
> +isolates pVMs from the host by unmapping pages from the stage-2 identity
> +map as they are accessed by a pVM. Hypercalls are provided for a pVM to
> +share specific regions of its IPA space back with the host, allowing
> +for communication with the VMM. See hypercalls.rst for more details.
> +
> +Isolation mechanisms
> +====================
> +
> +pKVM relies on a number of mechanisms to isolate PVMs from the host:
> +
> +CPU memory isolation
> +--------------------
> +
> +Status: Isolation of anonymous memory and metadata pages.
> +
> +Metadata pages (e.g. page-table pages and '``struct kvm_vcpu``' pages)
> +are donated from the host to the hypervisor during pVM creation and
> +are consequently unmapped from the stage-2 identity map until the pVM is
> +destroyed.
> +
> +Similarly to regular KVM, pages are lazily mapped into the guest in
> +response to stage-2 page faults handled by the host. However, when
> +running a pVM, these pages are first pinned and then unmapped from the
> +stage-2 identity map as part of the donation procedure. This gives rise
> +to some user-visible differences when compared to non-protected VMs,
> +largely due to the lack of MMU notifiers:
> +
> +* Memslots cannot be moved or deleted once the pVM has started running.
> +* Read-only memslots and dirty logging are not supported.
> +* With the exception of swap, file-backed pages cannot be mapped into a
> +  pVM.
> +* Donated pages are accounted against ``RLIMIT_MLOCK`` and so the VMM
> +  must have a sufficient resource limit or be granted ``CAP_IPC_LOCK``.

Perhaps worth to add that there's no runtime reclaim either so the accounting
will only grow until the VM is destroyed?

> +* Changes to the VMM address space (e.g. a ``MAP_FIXED`` mmap() over a
> +  mapping associated with a memslot) are not reflected in the guest and
> +  may lead to loss of coherency.
> +* Accessing pVM memory that has not been shared back will result in the
> +  delivery of a SIGSEGV.
> +* If a system call accesses pVM memory that has not been shared back
> +  then it will either return ``-EFAULT`` or forcefully reclaim the
> +  memory pages. Reclaimed memory is zeroed by the hypervisor and a
> +  subsequent attempt to access it in the pVM will return ``-EFAULT``
> +  from the ``VCPU_RUN`` ioctl().
> +
> +CPU state isolation
> +-------------------
> +
> +Status: **Unimplemented.**
> +
> +DMA isolation using an IOMMU
> +----------------------------
> +
> +Status: **Unimplemented.**
> +
> +Proxying of Trustzone services
> +------------------------------
> +
> +Status: FF-A and PSCI calls from the host are proxied by the pKVM
> +hypervisor.
> +
> +The FF-A proxy ensures that the host cannot share pVM or hypervisor
> +memory with Trustzone as part of a "confused deputy" attack.
> +
> +The PSCI proxy ensures that CPUs always have the stage-2 identity map
> +installed when they are executing in the host.
> +
> +Protected VM firmware (pvmfw)
> +-----------------------------
> +
> +Status: **Unimplemented.**
> +
> +Resources
> +=========
> +
> +Quentin Perret's KVM Forum 2022 talk entitled "Protected KVM on arm64: A
> +technical deep dive" remains a good resource for learning more about
> +pKVM, despite some of the details having changed in the meantime:
> +
> +https://www.youtube.com/watch?v=9npebeVFbFw
> -- 
> 2.52.0.351.gbe84eed79e-goog
>