Re: [Qemu-ppc] [PATCH] hw/ppc/spapr_cpu_core: Add a proper check for spapr machine

[Greg Kurz]

On Sat, 12 Aug 2017 12:16:51 +0200
Greg Kurz <address@hidden> wrote:

> On Sat, 12 Aug 2017 10:38:10 +0200
> Thomas Huth <address@hidden> wrote:
> 
> > QEMU currently crashes when the user tries to add a spapr-cpu-core
> > on a non-pseries machine:
> > 
> > $ qemu-system-ppc64 -S -machine ppce500,accel=tcg \
> >                     -device POWER5+_v2.1-spapr-cpu-core
> > hw/ppc/spapr_cpu_core.c:178:spapr_cpu_core_realize_child:
> > Object 0x55cee1f55160 is not an instance of type spapr-machine
> > Aborted (core dumped)
> > 
> > So let's add a proper check for the correct machine time with
> > a more friendly error message here.
> > 
> > Reported-by: Eduardo Habkost <address@hidden>
> > Signed-off-by: Thomas Huth <address@hidden>
> > ---
> >  hw/ppc/spapr_cpu_core.c   | 9 ++++++++-
> >  scripts/device-crash-test | 1 +
> >  2 files changed, 9 insertions(+), 1 deletion(-)
> > 
> > diff --git a/hw/ppc/spapr_cpu_core.c b/hw/ppc/spapr_cpu_core.c
> > index ea278ce..0f3d653 100644
> > --- a/hw/ppc/spapr_cpu_core.c
> > +++ b/hw/ppc/spapr_cpu_core.c
> > @@ -175,11 +175,18 @@ static void spapr_cpu_core_unrealizefn(DeviceState 
> > *dev, Error **errp)
> >  static void spapr_cpu_core_realize_child(Object *child, Error **errp)
> >  {
> >      Error *local_err = NULL;
> > -    sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
> > +    sPAPRMachineState *spapr;
> >      CPUState *cs = CPU(child);
> >      PowerPCCPU *cpu = POWERPC_CPU(cs);
> >      Object *obj;
> >  
> > +    spapr = (sPAPRMachineState *)object_dynamic_cast(qdev_get_machine(),
> > +                                                     TYPE_SPAPR_MACHINE);
> > +    if (!spapr) {
> > +        error_setg(errp, "spapr-cpu-core needs a pseries machine");
> > +        return;
> > +    }
> > +  
> 
> This is the realize function for individual threads. Maybe this sanity check
> should be performed earlier at the core level in spapr_cpu_core_realize() ?
> 
> Also, spapr_cpu_core_realize_child() seem to only have spapr to pass it down
> to spapr_cpu_init()... ie, not even sure spapr_cpu_core_realize_child()
> actually needs a spapr variable.
> 

And spapr_cpu_init() doesn't even need the spapr machine object itself: it
only needs the TYPE_PPC_VIRTUAL_HYPERVISOR object. I guess the right fix
would be for spapr cpu core to have a "ppc-virtual-hypervisor property set
in spapr_cpu_init() and tested in spapr_cpu_core_realize(). If the change
is considered to be too invasive during hard freeze, I guess a sanity check
in spapr_cpu_core_realize() is enough for now as a temporary bug fix.

> >      object_property_set_bool(child, true, "realized", &local_err);
> >      if (local_err) {
> >          goto error;
> > diff --git a/scripts/device-crash-test b/scripts/device-crash-test
> > index e77b693..8eb2d02 100755
> > --- a/scripts/device-crash-test
> > +++ b/scripts/device-crash-test
> > @@ -200,6 +200,7 @@ ERROR_WHITELIST = [
> >      {'log':r"Multiple VT220 operator consoles are not supported"},
> >      {'log':r"core 0 already populated"},
> >      {'log':r"could not find stage1 bootloader"},
> > +    {'log':r"spapr-cpu-core needs a pseries machine"},
> >  
> >      # other exitcode=1 failures not listed above will just generate INFO 
> > messages:
> >      {'exitcode':1, 'loglevel':logging.INFO},  
>

pgpfjIKp00D8j.pgp
Description: OpenPGP digital signature