Re: [Qemu-devel] [PATCH] spapr_pci: fail gracefully with non-pseries machine types

[Greg Kurz <groug@kaod.org>]

[-- Attachment #1: Type: text/plain, Size: 3382 bytes --]

On Sat, 7 Oct 2017 16:19:25 +1100
David Gibson <david@gibson.dropbear.id.au> wrote:

> On Fri, Oct 06, 2017 at 05:09:04PM +0200, Greg Kurz wrote:
> > QEMU currently crashes when the user tries to add a spapr-pci-host-bridge
> > on a non-pseries machine:
> > 
> > $ qemu-system-ppc64 -M ppce500 -device spapr-pci-host-bridge,index=1
> > hw/ppc/spapr_pci.c:1535:spapr_phb_realize:
> > Object 0x1003dacae60 is not an instance of type spapr-machine
> > Aborted (core dumped)
> > 
> > The same thing happens with the deprecated but still available child type
> > spapr-pci-vfio-host-bridge.
> > 
> > Fix both by checking the machine type with object_dynamic_cast().
> > 
> > Signed-off-by: Greg Kurz <groug@kaod.org>  
> 
> 
> 
> > ---
> >  hw/ppc/spapr_pci.c |    8 +++++++-
> >  1 file changed, 7 insertions(+), 1 deletion(-)
> > 
> > diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
> > index 5049ced4e8b4..9e85106f51f8 100644
> > --- a/hw/ppc/spapr_pci.c
> > +++ b/hw/ppc/spapr_pci.c
> > @@ -1507,7 +1507,7 @@ static void spapr_pci_unplug_request(HotplugHandler *plug_handler,
> >  
> >  static void spapr_phb_realize(DeviceState *dev, Error **errp)
> >  {
> > -    sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
> > +    sPAPRMachineState *spapr;
> >      SysBusDevice *s = SYS_BUS_DEVICE(dev);
> >      sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(s);
> >      PCIHostState *phb = PCI_HOST_BRIDGE(s);
> > @@ -1519,6 +1519,12 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
> >      const unsigned windows_supported =
> >          sphb->ddw_enabled ? SPAPR_PCI_DMA_MAX_WINDOWS : 1;
> >  
> > +    spapr = (sPAPRMachineState *) qdev_get_machine();
> > +    if (!object_dynamic_cast((Object *) spapr, TYPE_SPAPR_MACHINE)) {
> > +        error_setg(errp, TYPE_SPAPR_PCI_HOST_BRIDGE " needs a pseries machine");
> > +        return;
> > +    }  
> 
> This is slightly clunky.  You could instead use OBJECT_CHECK in the
> initializer, then just if (!spapr) here.
> 

The negative review on v2 seem to indicate we have an agreement on using
object_dynamic_cast() explicitly. So we have two possibilities, either
the "clunky" v1, or possibly:

-    sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
+    /* We don't use SPAPR_MACHINE() in order to exit gracefully if the user
+     * tries to add a PHB to a non-pseries machine.
+     */
+    sPAPRMachineState *spapr =
+        (sPAPRMachineState *) object_dynamic_cast(qdev_get_machine(),
+                                                  TYPE_SPAPR_MACHINE);
     SysBusDevice *s = SYS_BUS_DEVICE(dev);
     sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(s);
     PCIHostState *phb = PCI_HOST_BRIDGE(s);
     char *namebuf;
     int i;
     PCIBus *bus;
     uint64_t msi_window_size = 4096;
     sPAPRTCETable *tcet;
     const unsigned windows_supported =
         sphb->ddw_enabled ? SPAPR_PCI_DMA_MAX_WINDOWS : 1;
 
+    if (!spapr) {
+        error_setg(errp, TYPE_SPAPR_PCI_HOST_BRIDGE " needs a pseries machine");
+        return;
+    }
+

Which one do you prefer ? I'll also change spapr_cpu_core_realize()
if needed.

> > +
> >      if (sphb->index != (uint32_t)-1) {
> >          sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
> >          Error *local_err = NULL;
> >   
> 
> 


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 195 bytes --]