Re: [PATCH v5 05/15] KVM: s390: unify pending_irqs() and pending_irqs_no_gisa()

[Michael Mueller]

On 20.12.18 13:21, Cornelia Huck wrote:
> On Thu, 20 Dec 2018 12:49:56 +0100
> Michael Mueller <mimu@xxxxxxxxxxxxx> wrote:
>
> > On 20.12.18 12:06, Cornelia Huck wrote:
> > > On Wed, 19 Dec 2018 20:17:46 +0100
> > > Michael Mueller <mimu@xxxxxxxxxxxxx> wrote:
> > >    
> > > > Use a single function with parameter irq_flags to differentiate
> > > > between cases.
> > > >
> > > > New irq flag:
> > > >      IRQ_FLAG_LOCAL: include vcpu local interruptions pending
> > > >      IRQ_FLAG_FLOATING: include vcpu floating interruptions pending
> > > >      IRQ_FLAG_GISA: include GISA interruptions pending in IPM
> > >
> > > I presume that means that irqs may be in more than one set? Or are gisa
> > > irqs not considered floating irqs, because they use a different
> > > mechanism?
> >
> > Currently, the interruptions managed in GISA are floating only. But
> > that might change in future. The idea is not to subsume IRQ_FLAG_GISA
> > in IRQ_FLAG_FLOATING but to be able to address the right set of
> > procedures to determine the irq pending set for a given subset of irq
> > types that have different implementations.
> >
> > There might be a better name for IRQ_FLAG_FLOATING then?
>
> So the split is:
>
> - local interrupts that are pending via kvm structures;
> - floating interrupts that are pending via kvm structures;
> - interrupts that are pending via gisa?
>
> If so, what about
> - IRQ_FLAG_KVM_LOCAL
> - IRQ_FLAG_KVM_FLOATING
> - IRQ_FLAG_GISA (or maybe IRQ_FLAG_GISA_FLOATING, if you need to
>    distinguish those later on?)


yes, that's the split and I will go with:

IRQ_FLAG_KVM_LOCAL
IRQ_FLAG_KVM_FLOATING
IRQ_FLAG_GISA

initially.

The floating and local selection can be done by related masks

IRQ_MASK_LOCAL
IRQ_MASK_FLOATING

if required.

Thanks!

> > >    
> > > > New irq masks:
> > > >      IRQ_MASK_ALL: include all types
> > > >      IRQ_MASK_NO_GISA: include all types but GISA
> > > >
> > > > Examples:
> > > >      pending_irqs(vcpu, IRQ_MASK_ALL)
> > > >      pending_irqs(vcpu, IRQ_MASK_NO_GISA)
> > > >
> > > > There will be more irq flags with upcoming patches.
> > > >
> > > > Signed-off-by: Michael Mueller <mimu@xxxxxxxxxxxxx>
> > > > ---
> > > >    arch/s390/kvm/interrupt.c | 33 +++++++++++++++++++++------------
> > > >    1 file changed, 21 insertions(+), 12 deletions(-)
> > > >
> > > > diff --git a/arch/s390/kvm/interrupt.c b/arch/s390/kvm/interrupt.c
> > > > index 093b568b6356..4ab20d2eb180 100644
> > > > --- a/arch/s390/kvm/interrupt.c
> > > > +++ b/arch/s390/kvm/interrupt.c
> > > > @@ -31,6 +31,13 @@
> > > >    #define PFAULT_DONE 0x0680
> > > >    #define VIRTIO_PARAM 0x0d00
> > > >    
> > > > +#define IRQ_FLAG_LOCAL    0x8000 /* include local interruption pending mask */
> > > > +#define IRQ_FLAG_FLOATING 0x4000 /* include float interruption pending mask */
> > > > +#define IRQ_FLAG_GISA     0x2000 /* include GISA interruption pending mask */
> > > > +
> > > > +#define IRQ_MASK_ALL      (IRQ_FLAG_LOCAL | IRQ_FLAG_FLOATING | IRQ_FLAG_GISA)
> > > > +#define IRQ_MASK_NO_GISA  (IRQ_MASK_ALL & ~IRQ_FLAG_GISA)
> > > > +
> > > >    /* handle external calls via sigp interpretation facility */
> > > >    static int sca_ext_call_pending(struct kvm_vcpu *vcpu, int *src_id)
> > > >    {
> > > > @@ -237,16 +244,18 @@ static inline int kvm_s390_gisa_tac_ipm_gisc(struct kvm_s390_gisa *gisa, u32 gis
> > > >    	return test_and_clear_bit_inv(IPM_BIT_OFFSET + gisc, (unsigned long *) gisa);
> > > >    }
> > > >    
> > > > -static inline unsigned long pending_irqs_no_gisa(struct kvm_vcpu *vcpu)
> > > > +static inline unsigned long pending_irqs(struct kvm_vcpu *vcpu, u16 irq_flags)
> > >
> > > Any deeper reason why this is a u16? 16 bits should be enough for
> > > everyone? :)
> >
> > I want to use the 8 bits for the IRQ type and the other 8 for additional
> > controls, see: "KVM: s390: restore IAM in get_ipm() when IPM is clean"
>
> Still need to look at that patch, but my question mainly was "why only
> 16 bits"? I would think making this local variable larger is cheap.

I will enlarge the flag mask to u32 with 16 bits for the IRQ types then.

> > >    
> > > >    {
> > > > -	return vcpu->kvm->arch.float_int.pending_irqs |
> > > > -		vcpu->arch.local_int.pending_irqs;
> > > > -}

--
Mit freundlichen GrÃÃen / Kind regards
Michael MÃller

IBM Deutschland Research & Development GmbH
Vorsitzender des Aufsichtsrats: Matthias Hartmann
GeschÃftsfÃhrung: Dirk Wittkopp
Sitz der Gesellschaft: BÃblingen
Registergericht: Amtsgericht Stuttgart, HRB 243294