On Fri, 03 Jun 2022 06:23:25 +0100,
Reiji Watanabe <reijiw@xxxxxxxxxx> wrote:
>
> Hi Marc,
>
> On Sat, May 28, 2022 at 4:38 AM Marc Zyngier <maz@xxxxxxxxxx> wrote:
> >
> > The vcpu KVM_ARM64_FP_FOREIGN_FPSTATE flag tracks the thread's own
> > TIF_FOREIGN_FPSTATE so that we can evaluate just before running
> > the vcpu whether it the FP regs contain something that is owned
> > by the vcpu or not by updating the rest of the FP flags.
> >
> > We do this in the hypervisor code in order to make sure we're
> > in a context where we are not interruptible. But we already
> > have a hook in the run loop to generate this flag. We may as
> > well update the FP flags directly and save the pointless flag
> > tracking.
> >
> > Whilst we're at it, rename update_fp_enabled() to guest_owns_fp_regs()
> > to indicate what the leftover of this helper actually do.
> >
> > Signed-off-by: Marc Zyngier <maz@xxxxxxxxxx>
>
> Reviewed-by: Reiji Watanabe <reijiw@xxxxxxxxxx>
>
>
> > --- a/arch/arm64/kvm/fpsimd.c
> > +++ b/arch/arm64/kvm/fpsimd.c
> > @@ -107,16 +107,19 @@ void kvm_arch_vcpu_load_fp(struct kvm_vcpu *vcpu)
> > }
> >
> > /*
> > - * Called just before entering the guest once we are no longer
> > - * preemptable. Syncs the host's TIF_FOREIGN_FPSTATE with the KVM
> > - * mirror of the flag used by the hypervisor.
> > + * Called just before entering the guest once we are no longer preemptable
> > + * and interrupts are disabled. If we have managed to run anything using
> > + * FP while we were preemptible (such as off the back of an interrupt),
> > + * then neither the host nor the guest own the FP hardware (and it was the
> > + * responsibility of the code that used FP to save the existing state).
> > + *
> > + * Note that not supporting FP is basically the same thing as far as the
> > + * hypervisor is concerned (nothing to save).
> > */
> > void kvm_arch_vcpu_ctxflush_fp(struct kvm_vcpu *vcpu)
> > {
> > - if (test_thread_flag(TIF_FOREIGN_FPSTATE))
> > - vcpu->arch.flags |= KVM_ARM64_FP_FOREIGN_FPSTATE;
> > - else
> > - vcpu->arch.flags &= ~KVM_ARM64_FP_FOREIGN_FPSTATE;
> > + if (!system_supports_fpsimd() || test_thread_flag(TIF_FOREIGN_FPSTATE))
> > + vcpu->arch.flags &= ~(KVM_ARM64_FP_ENABLED | KVM_ARM64_FP_HOST);
> > }
>
> Although kvm_arch_vcpu_load_fp() unconditionally sets KVM_ARM64_FP_HOST,
> perhaps having kvm_arch_vcpu_load_fp() set KVM_ARM64_FP_HOST only when
> FP is supported might be more consistent?
> Then, checking system_supports_fpsimd() is unnecessary here.
> (KVM_ARM64_FP_ENABLED is not set when FP is not supported)
That's indeed a possibility. But I'm trying not to change the logic
here, only to move it to a place that provides the same semantic
without the need for an extra flag.
I'm happy to stack an extra patch on top of this series though.
Thanks,
M.
--
Without deviation from the norm, progress is not possible.
