Re: [Qemu-ppc] [RFC for-2.12 7/8] spapr: Handle VMX/VSX presence as an spapr capability flag

[David Gibson]

On Tue, Dec 12, 2017 at 01:54:36PM +0100, Greg Kurz wrote:
> On Mon, 11 Dec 2017 18:08:07 +1100
> David Gibson <address@hidden> wrote:
> 
> > We currently have some conditionals in the spapr device tree code to decide
> > whether or not to advertise the availability of the VMX (aka Altivec) and
> > VSX vector extensions to the guest, based on whether the guest cpu has
> > those features.
> > 
> > This can lead to confusion and subtle failures on migration, since it makes
> > a guest visible change based only on host capabilities.  We now have a
> > better mechanism for this, in spapr capabilities flags, which explicitly
> > depend on user options rather than host capabilities.
> > 
> > Rework the advertisement of VSX and VMX based on a new VSX capability.  We
> > no longer bother with a conditional for VMX support, because every CPU
> > that's ever been supported by the pseries machine type supports VMX.
> > 
> > NOTE: Some userspace distributions (e.g. RHEL7.4) already rely on
> > availability of VSX in libc, so using cap-vsx=off may lead to a fatal
> > SIGILL in init.
> > 
> > Signed-off-by: David Gibson <address@hidden>
> > ---
> 
> Just one nit, see below, but anyway:
> 
> Reviewed-by: Greg Kurz <address@hidden>
> 
> 
> >  hw/ppc/spapr.c         | 18 ++++++++++--------
> >  hw/ppc/spapr_caps.c    | 24 ++++++++++++++++++++++++
> >  include/hw/ppc/spapr.h |  3 +++
> >  3 files changed, 37 insertions(+), 8 deletions(-)
> > 
> > diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> > index e206825ed9..6b37511e8f 100644
> > --- a/hw/ppc/spapr.c
> > +++ b/hw/ppc/spapr.c
> > @@ -556,14 +556,16 @@ static void spapr_populate_cpu_dt(CPUState *cs, void 
> > *fdt, int offset,
> >                            segs, sizeof(segs))));
> >      }
> >  
> > -    /* Advertise VMX/VSX (vector extensions) if available
> > -     *   0 / no property == no vector extensions
> > +    /* Advertise VSX (vector extensions) if available
> >       *   1               == VMX / Altivec available
> > -     *   2               == VSX available */
> > -    if (env->insns_flags & PPC_ALTIVEC) {
> > -        uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
> > -
> > -        _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", vmx)));
> > +     *   2               == VSX available
> > +     *
> > +     * Only CPUs for which we create core types in spapr_cpu_core.c
> > +     * are possible, and all of those have VMX */
> > +    if (spapr_has_cap(spapr, SPAPR_CAP_VSX)) {
> > +        _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", 2)));
> 
> Maybe a good opportunity to drop the extra parens since the _FDT() macro
> already adds them around its argument.
> 
> #define _FDT(exp)                                                  \
>     do {                                                           \
>         int ret = (exp);                                           \

Uh, yeah, that won't work.  The extra parens are so that the macro
considers the whole string inside as a single argument, even if it
includes ',' characters, which it usually will.

> 
> > +    } else {
> > +        _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", 1)));
> >      }
> >  
> >      /* Advertise DFP (Decimal Floating Point) if available
> > @@ -3829,7 +3831,7 @@ static void spapr_machine_class_init(ObjectClass *oc, 
> > void *data)
> >       */
> >      mc->numa_mem_align_shift = 28;
> >  
> > -    smc->default_caps = spapr_caps(SPAPR_CAP_HTM);
> > +    smc->default_caps = spapr_caps(SPAPR_CAP_HTM | SPAPR_CAP_VSX);
> >      spapr_capability_properties(smc, &error_abort);
> >  }
> >  
> > diff --git a/hw/ppc/spapr_caps.c b/hw/ppc/spapr_caps.c
> > index 6bba04d60e..b19a9f4417 100644
> > --- a/hw/ppc/spapr_caps.c
> > +++ b/hw/ppc/spapr_caps.c
> > @@ -44,6 +44,11 @@ static sPAPRCapabilityInfo capability_table[] = {
> >          .description = "Allow Hardware Transactional Memory (HTM)",
> >          .bit = SPAPR_CAP_HTM,
> >      },
> > +    {
> > +        .name = "vsx",
> > +        .description = "Allow Vector Scalar Extensions (VSX)",
> > +        .bit = SPAPR_CAP_VSX,
> > +    },
> >  };
> >  
> >  static sPAPRCapabilities default_caps_with_cpu(sPAPRMachineState *spapr, 
> > CPUState *cs)
> > @@ -59,6 +64,11 @@ static sPAPRCapabilities 
> > default_caps_with_cpu(sPAPRMachineState *spapr, CPUStat
> >          caps.mask &= ~SPAPR_CAP_HTM;
> >      }
> >  
> > +    if (!ppc_check_compat(cpu, CPU_POWERPC_LOGICAL_2_06,
> > +                          0, spapr->max_compat_pvr)) {
> > +        caps.mask &= ~SPAPR_CAP_VSX;
> > +    }
> > +
> >      return caps;
> >  }
> >  
> > @@ -149,6 +159,8 @@ const VMStateDescription vmstate_spapr_caps = {
> >  
> >  static void spapr_allow_caps(sPAPRMachineState *spapr, Error **errp)
> >  {
> > +    PowerPCCPU *cpu = POWERPC_CPU(first_cpu);
> > +    CPUPPCState *env = &cpu->env;
> >      Error *err = NULL;
> >  
> >      if (spapr_has_cap(spapr, SPAPR_CAP_HTM)) {
> > @@ -164,6 +176,14 @@ static void spapr_allow_caps(sPAPRMachineState *spapr, 
> > Error **errp)
> >          }
> >      }
> >  
> > +    g_assert(env->insns_flags & PPC_ALTIVEC);
> > +    if (spapr_has_cap(spapr, SPAPR_CAP_VSX)) {
> > +        if (!(env->insns_flags2 & PPC2_VSX)) {
> > +            error_setg(&err, "VSX support not available, try cap-vsx=off");
> > +            goto out;
> > +        }
> > +    }
> > +
> >  out:
> >      if (err) {
> >          error_propagate(errp, err);
> > @@ -175,6 +195,10 @@ static void spapr_enforce_caps(sPAPRMachineState 
> > *spapr, Error **errp)
> >      if (!spapr_has_cap(spapr, SPAPR_CAP_HTM)) {
> >          /* TODO: Tell KVM not to allow HTM instructions */
> >      }
> > +
> > +    if (!spapr_has_cap(spapr, SPAPR_CAP_VSX)) {
> > +        /* TODO: Any way to disable VSX? */
> > +    }
> >  }
> >  
> >  void spapr_caps_reset(sPAPRMachineState *spapr)
> > diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
> > index 9b80fe72ed..4c173cb40c 100644
> > --- a/include/hw/ppc/spapr.h
> > +++ b/include/hw/ppc/spapr.h
> > @@ -59,6 +59,9 @@ typedef enum {
> >  /* Hardware Transactional Memory */
> >  #define SPAPR_CAP_HTM               0x0000000000000001ULL
> >  
> > +/* Vector Scalar Extensions */
> > +#define SPAPR_CAP_VSX               0x0000000000000002ULL
> > +
> >  typedef struct sPAPRCapabilities sPAPRCapabilities;
> >  struct sPAPRCapabilities {
> >      uint64_t mask;
> 

-- 
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson

signature.asc
Description: PGP signature