Re: [PATCH 2/3] KVM: arm/arm64: Add ARM arch timer interrupts ABI

[Christoffer Dall <christoffer.dall@linaro.org>]

On Tue, Nov 01, 2016 at 11:26:54AM +0000, Peter Maydell wrote:
> On 27 September 2016 at 20:08, Christoffer Dall
> <christoffer.dall@linaro.org> wrote:
> > From: Alexander Graf <agraf@suse.de>
> >
> > We have 2 modes for dealing with interrupts in the ARM world. We can
> > either handle them all using hardware acceleration through the vgic or
> > we can emulate a gic in user space and only drive CPU IRQ pins from
> > there.
> >
> > Unfortunately, when driving IRQs from user space, we never tell user
> > space about timer events that may result in interrupt line state
> > changes, so we lose out on timer events if we run with user space gic
> > emulation.
> >
> > Define an ABI to publish the timer output level to userspace.
> >
> > Signed-off-by: Alexander Graf <agraf@suse.de>
> > Signed-off-by: Christoffer Dall <christoffer.dall@linaro.org>
> > Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
> > ---
> >  Documentation/virtual/kvm/api.txt | 29 +++++++++++++++++++++++++++++
> >  arch/arm/include/uapi/asm/kvm.h   |  2 ++
> >  arch/arm64/include/uapi/asm/kvm.h |  2 ++
> >  include/uapi/linux/kvm.h          |  6 ++++++
> >  4 files changed, 39 insertions(+)
> >
> > diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
> > index 739db9a..2adf600 100644
> > --- a/Documentation/virtual/kvm/api.txt
> > +++ b/Documentation/virtual/kvm/api.txt
> > @@ -3928,3 +3928,32 @@ In order to use SynIC, it has to be activated by setting this
> >  capability via KVM_ENABLE_CAP ioctl on the vcpu fd. Note that this
> >  will disable the use of APIC hardware virtualization even if supported
> >  by the CPU, as it's incompatible with SynIC auto-EOI behavior.
> > +
> > +8.3 KVM_CAP_ARM_TIMER
> > +
> > +Architectures: arm, arm64
> > +This capability, if KVM_CHECK_EXTENSION indicates that it is available, means
> > +that if userspace creates a VM without an in-kernel interrupt controller, it
> > +will be notified of changes to the output level of ARM architected timers
> > +presented to the VM.  For such VMs, on every return to userspace, the kernel
> > +updates the vcpu's run->s.regs.timer_irq_level field to represent the actual
> > +output level of the timers.
> > +
> > +Whenever kvm detects a change in the timer output level, kvm guarantees at
> > +least one return to userspace before running the VM.  This exit could either
> > +be a KVM_EXIT_INTR or any other exit event, like KVM_EXIT_MMIO. This way,
> > +userspace can always sample the timer output level and re-compute the state of
> > +the userspace interrupt controller.  Userspace should always check the state
> > +of run->s.regs.timer_irq_level on every kvm exit.  The value in
> > +run->s.regs.timer_irq_level should be considered a level triggered interrupt
> > +signal.
> > +
> > +The field run->s.regs.timer_irq_level is available independent of
> > +run->kvm_valid_regs or run->kvm_dirty_regs bits.
> > +
> > +Currently the following bits are defined for the timer_irq_level bitmap:
> > +
> > +    KVM_ARM_TIMER_VTIMER  -  virtual timer
> > +
> > +Future versions of kvm may implement additional timer events. These will get
> > +indicated by additional KVM_CAP extensions.
> 
> This API looks good to me generally. My only question is whether we
> want to name the struct fields so they're not specifically talking
> about timer interrupts. For instance we probably want to expose the
> vPMU interrupt line to userspace too. We could do that by adding another
> struct field pmu_irq_level, but we could equally just assign it a bit
> in the existing irq_level field.
> 
> Possible current and future outbound interrupt lines (some of these
> would only show up in some unlikely or lots-of-implementation-needed
> cases, I'm just trying to produce an exhaustive list):
>  * virtual timer
>  * physical timer
>  * hyp timer (nested virtualization case)
>  * secure timer (unlikely but maybe if EL3 is ever supported inside a VM)
>  * gic maintenance interrupt (nested virt again)
>  * PMU interrupt

Thanks for the list, that's good to have around for the future.

There's also the potential of the EL2 virtual timer for nested VHE
support, right?

> 
> The kernel doesn't know which interrupt number these would be wired
> up to, so they're all just arbitrary outputs, and you could put them
> in one field or split them up into multiple fields, it doesn't make
> much difference.
> 

So if we keep this we're kind of suggesting that we'll have a field per
device type later on.  Since this is a u8 and we are talking about up 5
5 timers already, there's not much waste currently, and we have plenty
of padding.  I suppose we an always add a 'other_devices' thing later,
so I prefer just sticking with this one for now.

Thanks,
-Christoffer