Skip to content

Commit dafa493

Browse files
committed
KVM: arm64: Document vCPU feature selection UAPIs
KVM/arm64 has a couple schemes for handling vCPU feature selection now, which is a lot to put on userspace. Add some documentation about how these interact and provide some recommendations for how to use the writable ID register scheme. Reviewed-by: Marc Zyngier <maz@kernel.org> Link: https://lore.kernel.org/r/20231003230408.3405722-11-oliver.upton@linux.dev Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
1 parent f89fbb3 commit dafa493

File tree

3 files changed

+53
-0
lines changed

3 files changed

+53
-0
lines changed

Documentation/virt/kvm/api.rst

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -3370,6 +3370,8 @@ return indicates the attribute is implemented. It does not necessarily
33703370
indicate that the attribute can be read or written in the device's
33713371
current state. "addr" is ignored.
33723372

3373+
.. _KVM_ARM_VCPU_INIT:
3374+
33733375
4.82 KVM_ARM_VCPU_INIT
33743376
----------------------
33753377

@@ -6070,6 +6072,8 @@ writes to the CNTVCT_EL0 and CNTPCT_EL0 registers using the SET_ONE_REG
60706072
interface. No error will be returned, but the resulting offset will not be
60716073
applied.
60726074

6075+
.. _KVM_ARM_GET_REG_WRITABLE_MASKS:
6076+
60736077
4.139 KVM_ARM_GET_REG_WRITABLE_MASKS
60746078
-------------------------------------------
60756079

Documentation/virt/kvm/arm/index.rst

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -11,3 +11,4 @@ ARM
1111
hypercalls
1212
pvtime
1313
ptp_kvm
14+
vcpu-features
Lines changed: 48 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,48 @@
1+
.. SPDX-License-Identifier: GPL-2.0
2+
3+
===============================
4+
vCPU feature selection on arm64
5+
===============================
6+
7+
KVM/arm64 provides two mechanisms that allow userspace to configure
8+
the CPU features presented to the guest.
9+
10+
KVM_ARM_VCPU_INIT
11+
=================
12+
13+
The ``KVM_ARM_VCPU_INIT`` ioctl accepts a bitmap of feature flags
14+
(``struct kvm_vcpu_init::features``). Features enabled by this interface are
15+
*opt-in* and may change/extend UAPI. See :ref:`KVM_ARM_VCPU_INIT` for complete
16+
documentation of the features controlled by the ioctl.
17+
18+
Otherwise, all CPU features supported by KVM are described by the architected
19+
ID registers.
20+
21+
The ID Registers
22+
================
23+
24+
The Arm architecture specifies a range of *ID Registers* that describe the set
25+
of architectural features supported by the CPU implementation. KVM initializes
26+
the guest's ID registers to the maximum set of CPU features supported by the
27+
system. The ID register values may be VM-scoped in KVM, meaning that the
28+
values could be shared for all vCPUs in a VM.
29+
30+
KVM allows userspace to *opt-out* of certain CPU features described by the ID
31+
registers by writing values to them via the ``KVM_SET_ONE_REG`` ioctl. The ID
32+
registers are mutable until the VM has started, i.e. userspace has called
33+
``KVM_RUN`` on at least one vCPU in the VM. Userspace can discover what fields
34+
are mutable in the ID registers using the ``KVM_ARM_GET_REG_WRITABLE_MASKS``.
35+
See the :ref:`ioctl documentation <KVM_ARM_GET_REG_WRITABLE_MASKS>` for more
36+
details.
37+
38+
Userspace is allowed to *limit* or *mask* CPU features according to the rules
39+
outlined by the architecture in DDI0487J.a D19.1.3 'Principles of the ID
40+
scheme for fields in ID register'. KVM does not allow ID register values that
41+
exceed the capabilities of the system.
42+
43+
.. warning::
44+
It is **strongly recommended** that userspace modify the ID register values
45+
before accessing the rest of the vCPU's CPU register state. KVM may use the
46+
ID register values to control feature emulation. Interleaving ID register
47+
modification with other system register accesses may lead to unpredictable
48+
behavior.

0 commit comments

Comments
 (0)