Re: [PATCH 18/21] target/arm/kvm: Init cap_has_inject_serror_esr in kvm_arch_init

[Richard Henderson]

On 12/11/23 10:43, Philippe Mathieu-Daudé wrote:
> On 11/12/23 18:09, Richard Henderson wrote:
> > On 11/24/23 03:54, Philippe Mathieu-Daudé wrote:
> > > On 23/11/23 05:42, Richard Henderson wrote:
> > > > There is no need to do this in kvm_arch_init_vcpu per vcpu.
> > > > Inline kvm_arm_init_serror_injection rather than keep separate.
> > > >
> > > > Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> > > > ---
> > > >   target/arm/kvm_arm.h |  8 --------
> > > >   target/arm/kvm.c     | 13 ++++---------
> > > >   2 files changed, 4 insertions(+), 17 deletions(-)
> > >
> > >
> > > > @@ -562,6 +556,10 @@ int kvm_arch_init(MachineState *ms, KVMState *s)
> > > >       cap_has_mp_state = kvm_check_extension(s, KVM_CAP_MP_STATE);
> > > > +    /* Check whether user space can specify guest syndrome value */
> > > > +    cap_has_inject_serror_esr =
> > > > +        kvm_check_extension(s, KVM_CAP_ARM_INJECT_SERROR_ESR);
> > > > +
> > > >       if (ms->smp.cpus > 256 &&
> > > >           !kvm_check_extension(s, KVM_CAP_ARM_IRQ_LINE_LAYOUT_2)) {
> > > >           error_report("Using more than 256 vcpus requires a host kernel "
> > > > @@ -1948,9 +1946,6 @@ int kvm_arch_init_vcpu(CPUState *cs)
> > > >       }
> > > >       cpu->mp_affinity = mpidr & ARM64_AFFINITY_MASK;
> > > > -    /* Check whether user space can specify guest syndrome value */
> > > > -    kvm_arm_init_serror_injection(cs);
> > > > -
> > > >       return kvm_arm_init_cpreg_list(cpu);
> > > >   }
> > >
> > >
> > > Just checking, in a heterogeneous setup we still want to keep
> > > these 2 calls per-vCPU, right?
> >
> > There is no hetrogeneous kvm -- every vcpu must match the host cpu.
>
> So big.LITTLE will never be a KVM thing?

Not as far as I'm aware.  There are even issues *running* on big.LITTLE hosts -- you must 
use taskset to limit qemu to one cpu type, either big or little.


r~