KVM: arm64: Add attribute to control GICD_TYPER2.nASSGIcap

[PATCH v3 0/4] KVM: arm64: Add attribute to control GICD_TYPER2.nASSGIcap

Posted by Raghavendra Rao Ananta 3 months, 4 weeks ago

A shortcoming of the GIC architecture is that there's an absolute limit on
the number of vPEs that can be tracked by the ITS. It is possible that
an operator is running a mix of VMs on a system, only wanting to provide
a specific class of VMs with hardware interrupt injection support.

The series introduces KVM_DEV_ARM_VGIC_FEATURE_nASSGIcap vGIC attribute to allow
the userspace to control GICD_TYPER2.nASSGIcap (GICv4.1) on a per-VM basis.

v1: https://lore.kernel.org/kvmarm/20250514192159.1751538-1-rananta@google.com/

v1 -> v2: https://lore.kernel.org/all/20250531012545.709887-1-oliver.upton@linux.dev/
 - Drop all use of GICv4 in the UAPI and KVM-internal helpers in favor
   of nASSGIcap. This changes things around to model a guest feature,
   not a host feature.

 - Consolidate UAPI into a single attribute and expect userspace to use
   to read the attribute for discovery, much like we do with the ID
   registers

 - Squash documentation together with implementation

 - Clean up maintenance IRQ attribute handling, which I ran into as part
   of reviewing this series

v2 -> v3:
 - Update checks in vgic-v3.c and vgic-v4.c to also include nASSGIcap (via
   vgic_supports_direct_sgis()) that's configured by the userspace. (Oliver)

Oliver Upton (2):
  KVM: arm64: Disambiguate support for vSGIs v. vLPIs
  KVM: arm64: vgic-v3: Consolidate MAINT_IRQ handling

Raghavendra Rao Ananta (2):
  KVM: arm64: Introduce attribute to control GICD_TYPER2.nASSGIcap
  KVM: arm64: selftests: Add test for nASSGIcap attribute

 .../virt/kvm/devices/arm-vgic-v3.rst          | 29 ++++++
 arch/arm64/include/uapi/asm/kvm.h             |  3 +
 arch/arm64/kvm/vgic/vgic-init.c               |  7 +-
 arch/arm64/kvm/vgic/vgic-kvm-device.c         | 88 +++++++++++++------
 arch/arm64/kvm/vgic/vgic-mmio-v3.c            | 24 +++--
 arch/arm64/kvm/vgic/vgic-v3.c                 |  5 +-
 arch/arm64/kvm/vgic/vgic-v4.c                 |  6 +-
 arch/arm64/kvm/vgic/vgic.c                    |  4 +-
 arch/arm64/kvm/vgic/vgic.h                    |  7 ++
 include/kvm/arm_vgic.h                        |  3 +
 tools/testing/selftests/kvm/arm64/vgic_init.c | 41 +++++++++
 11 files changed, 176 insertions(+), 41 deletions(-)


base-commit: 19272b37aa4f83ca52bdf9c16d5d81bdd1354494
--
2.50.0.rc2.692.g299adb8693-goog

Re: [PATCH v3 0/4] KVM: arm64: Add attribute to control GICD_TYPER2.nASSGIcap

Posted by Oliver Upton 3 months, 4 weeks ago

On Fri, Jun 13, 2025 at 03:52:34PM +0000, Raghavendra Rao Ananta wrote:
> A shortcoming of the GIC architecture is that there's an absolute limit on
> the number of vPEs that can be tracked by the ITS. It is possible that
> an operator is running a mix of VMs on a system, only wanting to provide
> a specific class of VMs with hardware interrupt injection support.
> 
> The series introduces KVM_DEV_ARM_VGIC_FEATURE_nASSGIcap vGIC attribute to allow
> the userspace to control GICD_TYPER2.nASSGIcap (GICv4.1) on a per-VM basis.
> 
> v1: https://lore.kernel.org/kvmarm/20250514192159.1751538-1-rananta@google.com/
> 
> v1 -> v2: https://lore.kernel.org/all/20250531012545.709887-1-oliver.upton@linux.dev/
>  - Drop all use of GICv4 in the UAPI and KVM-internal helpers in favor
>    of nASSGIcap. This changes things around to model a guest feature,
>    not a host feature.
> 
>  - Consolidate UAPI into a single attribute and expect userspace to use
>    to read the attribute for discovery, much like we do with the ID
>    registers
> 
>  - Squash documentation together with implementation
> 
>  - Clean up maintenance IRQ attribute handling, which I ran into as part
>    of reviewing this series
> 
> v2 -> v3:
>  - Update checks in vgic-v3.c and vgic-v4.c to also include nASSGIcap (via
>    vgic_supports_direct_sgis()) that's configured by the userspace. (Oliver)
> 
> Oliver Upton (2):
>   KVM: arm64: Disambiguate support for vSGIs v. vLPIs
>   KVM: arm64: vgic-v3: Consolidate MAINT_IRQ handling

Make sure you run checkpatch next time before sending out, it should've
warned you about sending patches w/o including your SOB.

Thanks,
Oliver

Re: [PATCH v3 0/4] KVM: arm64: Add attribute to control GICD_TYPER2.nASSGIcap

Posted by Raghavendra Rao Ananta 3 months, 4 weeks ago

On Fri, Jun 13, 2025 at 1:53 PM Oliver Upton <oliver.upton@linux.dev> wrote:
>
> On Fri, Jun 13, 2025 at 03:52:34PM +0000, Raghavendra Rao Ananta wrote:
> > A shortcoming of the GIC architecture is that there's an absolute limit on
> > the number of vPEs that can be tracked by the ITS. It is possible that
> > an operator is running a mix of VMs on a system, only wanting to provide
> > a specific class of VMs with hardware interrupt injection support.
> >
> > The series introduces KVM_DEV_ARM_VGIC_FEATURE_nASSGIcap vGIC attribute to allow
> > the userspace to control GICD_TYPER2.nASSGIcap (GICv4.1) on a per-VM basis.
> >
> > v1: https://lore.kernel.org/kvmarm/20250514192159.1751538-1-rananta@google.com/
> >
> > v1 -> v2: https://lore.kernel.org/all/20250531012545.709887-1-oliver.upton@linux.dev/
> >  - Drop all use of GICv4 in the UAPI and KVM-internal helpers in favor
> >    of nASSGIcap. This changes things around to model a guest feature,
> >    not a host feature.
> >
> >  - Consolidate UAPI into a single attribute and expect userspace to use
> >    to read the attribute for discovery, much like we do with the ID
> >    registers
> >
> >  - Squash documentation together with implementation
> >
> >  - Clean up maintenance IRQ attribute handling, which I ran into as part
> >    of reviewing this series
> >
> > v2 -> v3:
> >  - Update checks in vgic-v3.c and vgic-v4.c to also include nASSGIcap (via
> >    vgic_supports_direct_sgis()) that's configured by the userspace. (Oliver)
> >
> > Oliver Upton (2):
> >   KVM: arm64: Disambiguate support for vSGIs v. vLPIs
> >   KVM: arm64: vgic-v3: Consolidate MAINT_IRQ handling
>
> Make sure you run checkpatch next time before sending out, it should've
> warned you about sending patches w/o including your SOB.
>
Hmm, I do run checkpatch before sending, but I don't see any warning as such.

Example:
$ ./scripts/checkpatch.pl
v3-0001-KVM-arm64-Disambiguate-support-for-vSGIs-v.-vLPIs.patch
total: 0 errors, 0 warnings, 107 lines checked

v3-0001-KVM-arm64-Disambiguate-support-for-vSGIs-v.-vLPIs.patch has no
obvious style problems and is ready for submission.

I do see an option to tell the script to ignore the check:
--no-signoff, so I'm guessing it should check by default? Or is there
any other option?

Thank you.
Raghavendra