Re: [PATCH 24/32] KVM: s390: intercept the rpcit instruction

[Matthew Rosato]

On 12/14/21 12:04 PM, Pierre Morel wrote:
> On 12/7/21 21:57, Matthew Rosato wrote:
> > For faster handling of PCI translation refreshes, intercept in KVM
> > and call the associated handler.
> >
> > Signed-off-by: Matthew Rosato <mjrosato@xxxxxxxxxxxxx>
> > ---
> >   arch/s390/kvm/pci.h  |  4 ++++
> >   arch/s390/kvm/priv.c | 41 +++++++++++++++++++++++++++++++++++++++++
> >   2 files changed, 45 insertions(+)
> >
> > diff --git a/arch/s390/kvm/pci.h b/arch/s390/kvm/pci.h
> > index d252a631b693..3f96eff432aa 100644
> > --- a/arch/s390/kvm/pci.h
> > +++ b/arch/s390/kvm/pci.h
> > @@ -18,6 +18,10 @@
> >   #define KVM_S390_PCI_DTSM_MASK 0x40
> > +#define KVM_S390_RPCIT_STAT_MASK 0xffffffff00ffffffUL
> > +#define KVM_S390_RPCIT_INS_RES (0x10 << 24)
> > +#define KVM_S390_RPCIT_ERR (0x28 << 24)
>
> I
>
> > +
> >   struct zpci_gaite {
> >       unsigned int gisa;
> >       u8 gisc;
> > diff --git a/arch/s390/kvm/priv.c b/arch/s390/kvm/priv.c
> > index 417154b314a6..768ae92ecc59 100644
> > --- a/arch/s390/kvm/priv.c
> > +++ b/arch/s390/kvm/priv.c
> > @@ -29,6 +29,7 @@
> >   #include <asm/ap.h>
> >   #include "gaccess.h"
> >   #include "kvm-s390.h"
> > +#include "pci.h"
> >   #include "trace.h"
> >   static int handle_ri(struct kvm_vcpu *vcpu)
> > @@ -335,6 +336,44 @@ static int handle_rrbe(struct kvm_vcpu *vcpu)
> >       return 0;
> >   }
> > +static int handle_rpcit(struct kvm_vcpu *vcpu)
> > +{
> > +    int reg1, reg2;
> > +    int rc;
> > +
> > +    if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
> > +        return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
> > +
> > +    kvm_s390_get_regs_rre(vcpu, &reg1, &reg2);
> > +
>
> I would prefer to take care of the interception immediately here
>
>          fh = vcpu->run->s.regs.gprs[reg1] >> 32;
>          if ((fh & aift.mdd) != 0)
>                  return -EOPNOTSUP
>
> instead of doing it inside kvm_s390_pci_refresh_trans.
> It would simplify in my opinion.

OK

> > +    rc = kvm_s390_pci_refresh_trans(vcpu, vcpu->run->s.regs.gprs[reg1],
> > +                    vcpu->run->s.regs.gprs[reg2],
> > +                    vcpu->run->s.regs.gprs[reg2+1]);
> > +
>
>
> > +    switch (rc) {
> > +    case 0:
> > +        kvm_s390_set_psw_cc(vcpu, 0);
> > +        break;
> > +    case -EOPNOTSUPP:
> > +        return -EOPNOTSUPP;
> > +    case -EINVAL:
> > +        kvm_s390_set_psw_cc(vcpu, 3);
> > +        break;
> > +    case -ENOMEM:
> > +        vcpu->run->s.regs.gprs[reg1] &= KVM_S390_RPCIT_STAT_MASK;
> > +        vcpu->run->s.regs.gprs[reg1] |= KVM_S390_RPCIT_INS_RES;
> > +        kvm_s390_set_psw_cc(vcpu, 1);
> > +        break;
> > +    default:
> > +        vcpu->run->s.regs.gprs[reg1] &= KVM_S390_RPCIT_STAT_MASK;
> > +        vcpu->run->s.regs.gprs[reg1] |= KVM_S390_RPCIT_ERR;
>
> I think you should use the status reported by the hardware, reporting 
> "Error recovery in progress" what ever the hardware error was does not 
> seem right.

OK, this ties into your other comment about calling __rpcit() directly 
so we have a status to look at -- will look into it

> > +        kvm_s390_set_psw_cc(vcpu, 1);
> > +        break;
> > +    }
>
> NIT: This switch above could be much more simple if you set CC after the 
> switch.

We are setting 3 different CCs over 4 cases, so there's only 1 
duplication in the switch, so I'm not sure how much simpler?

But anyway this might not be relevant if I change to call __rpcit() 
directly.

> > +
> > +    return 0;
> > +}
> > +
> >   #define SSKE_NQ 0x8
> >   #define SSKE_MR 0x4
> >   #define SSKE_MC 0x2
> > @@ -1275,6 +1314,8 @@ int kvm_s390_handle_b9(struct kvm_vcpu *vcpu)
> >           return handle_essa(vcpu);
> >       case 0xaf:
> >           return handle_pfmf(vcpu);
> > +    case 0xd3:
> > +        return handle_rpcit(vcpu);
> >       default:
> >           return -EOPNOTSUPP;
> >       }