Re: [Qemu-devel] [PATCH 6/6] spapr: fix migration of ICP objects from/to older QEMU

[David Gibson <david@gibson.dropbear.id.au>]

[-- Attachment #1: Type: text/plain, Size: 8319 bytes --]

On Mon, May 15, 2017 at 02:22:32PM +0200, Cédric Le Goater wrote:
> On 05/15/2017 01:40 PM, Greg Kurz wrote:
> > Commit 5bc8d26de20c ("spapr: allocate the ICPState object from under
> > sPAPRCPUCore") moved ICP objects from the machine to CPU cores. This
> > is an improvement since we no longer allocate ICP objects that will
> > never be used. But it has the side-effect of breaking migration of
> > older machine types from older QEMU versions.
> > 
> > This patch introduces a compat flag in the sPAPR machine class so
> > that all pseries machine up to 2.9 go on with the previous behavior
> > of pre-allocating ICP objects.
> 
> I think this is a quite elegant way to a handle the migration 
> regression. Thanks for taking care of it.
> 
> Have you tried to simply reparent the ICPs objects to OBJECT(spapr) 
> instead of the OBJECT(cpu)  ? 

I actually kind of hate changing the QOM tree structure based on
machine type compatibility.  Unfortunately, since we're matching up
the migration state based (essentially) on QOM path, I don't see any
easy alternative.  I really wish there was a mechanism for defining
"alias paths" or something to handle this kind of migration
compatibility shim.

> See some minor comments below.
> 
> > While here, we also ensure that object_property_add_child() errors cause
> > QEMU to abort for newer machines.
> > 
> > Signed-off-by: Greg Kurz <groug@kaod.org>
> > ---
> >  hw/ppc/spapr.c          |   36 ++++++++++++++++++++++++++++++++++++
> >  hw/ppc/spapr_cpu_core.c |   28 ++++++++++++++++++++--------
> >  include/hw/ppc/spapr.h  |    2 ++
> >  3 files changed, 58 insertions(+), 8 deletions(-)
> > 
> > diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> > index c53989bb10b1..ab3683bcd677 100644
> > --- a/hw/ppc/spapr.c
> > +++ b/hw/ppc/spapr.c
> > @@ -126,6 +126,7 @@ error:
> >  static void xics_system_init(MachineState *machine, int nr_irqs, Error **errp)
> >  {
> >      sPAPRMachineState *spapr = SPAPR_MACHINE(machine);
> > +    sPAPRMachineClass *smc = SPAPR_MACHINE_GET_CLASS(spapr);
> >      Error *local_err = NULL;
> >  
> >      if (kvm_enabled()) {
> > @@ -151,6 +152,38 @@ static void xics_system_init(MachineState *machine, int nr_irqs, Error **errp)
> >                                        &local_err);
> >      }
> >  
> > +    if (!spapr->ics) {
> > +        goto out;
> > +    }
> > +
> > +    if (smc->must_pre_allocate_icps) {
> 
> I am not sure I like 'must', I think 'pre_allocate_icps' should be enough ? 
> or simply 'allocate_legacy_icps' ?

I'd actually prefer to make it explicit that this is a migration
compatibility shim and call it something like
'pre_2_10_icp_allocation'.

> > +        int smt = kvmppc_smt_threads();
> > +        int nr_servers = DIV_ROUND_UP(max_cpus * smt, smp_threads);
> 
> may be we should reintroduce nr_servers at the machine level ? 
> 
> > +        int i;
> > +
> > +        spapr->legacy_icps = g_malloc0(nr_servers * sizeof(ICPState));

This isn't technically safe, although you'll probably get away with
it.  spapr->icp_type is parameterized, which means it could be a
sub-class with a larger state structure than base ICPState.

> > +        for (i = 0; i < nr_servers; i++) {
> > +            void* obj = &spapr->legacy_icps[i];
> 
> 'void *'
> 
> > +
> > +            object_initialize(obj, sizeof(ICPState), spapr->icp_type);
> > +            object_property_add_child(OBJECT(spapr), "icp[*]", obj,
> > +                                      &error_abort);
> 
> David does not like the "icp[*]" syntax.
> 
> > +            object_unref(obj);
> > +            object_property_add_const_link(obj, "xics", OBJECT(spapr),
> > +                                           &error_abort);
> > +            object_property_set_bool(obj, true, "realized", &local_err);
> > +            if (local_err) {
> > +                while (i--) {
> > +                    object_unparent(obj);
> > +                }
> > +                g_free(spapr->legacy_icps);
> > +                break;
> > +            }
> > +        }
> > +    }
> > +
> > +out:
> >      error_propagate(errp, local_err);
> >  }
> >  
> > @@ -3256,8 +3289,11 @@ static void spapr_machine_2_9_instance_options(MachineState *machine)
> >  
> >  static void spapr_machine_2_9_class_options(MachineClass *mc)
> >  {
> > +    sPAPRMachineClass *smc = SPAPR_MACHINE_CLASS(mc);
> > +
> >      spapr_machine_2_10_class_options(mc);
> >      SET_MACHINE_COMPAT(mc, SPAPR_COMPAT_2_9);
> > +    smc->must_pre_allocate_icps = true;
> >  }
> >  
> >  DEFINE_SPAPR_MACHINE(2_9, "2.9", false);
> > diff --git a/hw/ppc/spapr_cpu_core.c b/hw/ppc/spapr_cpu_core.c
> > index 63d160f7e010..5476647efa06 100644
> > --- a/hw/ppc/spapr_cpu_core.c
> > +++ b/hw/ppc/spapr_cpu_core.c
> > @@ -119,6 +119,7 @@ static void spapr_cpu_core_unrealizefn(DeviceState *dev, Error **errp)
> >      size_t size = object_type_get_instance_size(typename);
> >      CPUCore *cc = CPU_CORE(dev);
> >      int i;
> > +    sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
> >  
> >      for (i = 0; i < cc->nr_threads; i++) {
> >          void *obj = sc->threads + i * size;
> > @@ -127,7 +128,9 @@ static void spapr_cpu_core_unrealizefn(DeviceState *dev, Error **errp)
> >          PowerPCCPU *cpu = POWERPC_CPU(cs);
> >  
> >          spapr_cpu_destroy(cpu);
> > -        object_unparent(cpu->intc);
> > +        if (!spapr->legacy_icps) {
> > +            object_unparent(cpu->intc);
> > +        }
> >          cpu_remove_sync(cs);
> >          object_unparent(obj);
> >      }
> > @@ -142,12 +145,19 @@ static void spapr_cpu_core_realize_child(Object *child, Error **errp)
> >      PowerPCCPU *cpu = POWERPC_CPU(cs);
> >      Object *obj;
> >  
> > -    obj = object_new(spapr->icp_type);
> > -    object_property_add_child(OBJECT(cpu), "icp", obj, NULL);
> > -    object_property_add_const_link(obj, "xics", OBJECT(spapr), &error_abort);
> > -    object_property_set_bool(obj, true, "realized", &local_err);
> > -    if (local_err) {
> > -        goto error;
> > +    if (spapr->legacy_icps) {
> > +        int index = cpu->parent_obj.cpu_index;
> > +
> > +        obj = OBJECT(&spapr->legacy_icps[index]);
> > +    } else {
> > +        obj = object_new(spapr->icp_type);
> > +        object_property_add_child(OBJECT(cpu), "icp", obj, &error_abort);
> > +        object_property_add_const_link(obj, "xics", OBJECT(spapr),
> > +                                       &error_abort);
> > +        object_property_set_bool(obj, true, "realized", &local_err);
> > +        if (local_err) {
> > +            goto error;
> > +        }
> >      }
> >  
> >      object_property_set_bool(child, true, "realized", &local_err);
> > @@ -164,7 +174,9 @@ static void spapr_cpu_core_realize_child(Object *child, Error **errp)
> >      return;
> >  
> >  error:
> > -    object_unparent(obj);
> > +    if (!spapr->legacy_icps) {
> > +        object_unparent(obj);
> > +    }
> >      error_propagate(errp, local_err);
> >  }
> >  
> > diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
> > index 5802f888c39d..72cd5af2679b 100644
> > --- a/include/hw/ppc/spapr.h
> > +++ b/include/hw/ppc/spapr.h
> > @@ -53,6 +53,7 @@ struct sPAPRMachineClass {
> >      bool dr_lmb_enabled;       /* enable dynamic-reconfig/hotplug of LMBs */
> >      bool use_ohci_by_default;  /* use USB-OHCI instead of XHCI */
> >      const char *tcg_default_cpu; /* which (TCG) CPU to simulate by default */
> > +    bool must_pre_allocate_icps; /* only for pseries-2.9 and older */
> >      void (*phb_placement)(sPAPRMachineState *spapr, uint32_t index,
> >                            uint64_t *buid, hwaddr *pio, 
> >                            hwaddr *mmio32, hwaddr *mmio64,
> > @@ -109,6 +110,7 @@ struct sPAPRMachineState {
> >      MemoryHotplugState hotplug_memory;
> >  
> >      const char *icp_type;
> > +    ICPState *legacy_icps;
> >  };
> >  
> >  #define H_SUCCESS         0
> > 
> 

-- 
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

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]