Re: [Qemu-devel] [PATCH v6 11/11] spapr: CPU hot unplug support

[Bharata B Rao <bharata@linux.vnet.ibm.com>]

On Wed, Jan 13, 2016 at 03:57:00PM +1100, David Gibson wrote:
> On Wed, Jan 13, 2016 at 09:40:54AM +0530, Bharata B Rao wrote:
> > On Tue, Jan 12, 2016 at 05:06:34PM +1100, David Gibson wrote:
> > > On Fri, Jan 08, 2016 at 12:25:19PM +0530, Bharata B Rao wrote:
> > > > Remove the CPU core device by removing the underlying CPU thread devices.
> > > > Support hot removal of CPU for sPAPR guests by sending the hot unplug
> > > > notification to the guest via EPOW interrupt. Release the vCPU object
> > > > after CPU hot unplug so that vCPU fd can be parked and reused.
> > > > 
> > > > Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
> > > > ---
> > > >  hw/ppc/spapr.c         | 113 +++++++++++++++++++++++++++++++++++++++++++++++++
> > > >  include/hw/ppc/spapr.h |   6 +++
> > > >  2 files changed, 119 insertions(+)
> > > > 
> > > > diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> > > > index c2af9ca..43cb726 100644
> > > > --- a/hw/ppc/spapr.c
> > > > +++ b/hw/ppc/spapr.c
> > > > @@ -93,6 +93,9 @@
> > > >  
> > > >  #define HTAB_SIZE(spapr)        (1ULL << ((spapr)->htab_shift))
> > > >  
> > > > +static QLIST_HEAD(cpu_unplug_list, sPAPRCPUUnplug) cpu_unplug_list
> > > > +                  = QLIST_HEAD_INITIALIZER(cpu_unplug_list);
> > > > +
> > > >  static XICSState *try_create_xics(const char *type, int nr_servers,
> > > >                                    int nr_irqs, Error **errp)
> > > >  {
> > > > @@ -2432,11 +2435,121 @@ static void spapr_machine_device_plug(HotplugHandler *hotplug_dev,
> > > >      }
> > > >  }
> > > >  
> > > > +static void spapr_cpu_destroy(PowerPCCPU *cpu)
> > > > +{
> > > > +    sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
> > > > +
> > > > +    xics_cpu_destroy(spapr->icp, cpu);
> > > > +    qemu_unregister_reset(spapr_cpu_reset, cpu);
> > > > +}
> > > > +
> > > > +static void spapr_cpu_core_cleanup(void)
> > > > +{
> > > > +    sPAPRCPUUnplug *unplug, *next;
> > > > +    Object *cpu;
> > > > +
> > > > +    QLIST_FOREACH_SAFE(unplug, &cpu_unplug_list, node, next) {
> > > > +        cpu = unplug->cpu;
> > > > +        object_unparent(cpu);
> > > > +        QLIST_REMOVE(unplug, node);
> > > > +        g_free(unplug);
> > > > +    }
> > > > +}
> > > > +
> > > > +static void spapr_add_cpu_to_unplug_list(Object *cpu)
> > > > +{
> > > > +    sPAPRCPUUnplug *unplug = g_malloc(sizeof(*unplug));
> > > > +
> > > > +    unplug->cpu = cpu;
> > > > +    QLIST_INSERT_HEAD(&cpu_unplug_list, unplug, node);
> > > > +}
> > > > +
> > > > +static int spapr_cpu_release(Object *obj, void *opaque)
> > > > +{
> > > > +    DeviceState *dev = DEVICE(obj);
> > > > +    CPUState *cs = CPU(dev);
> > > > +    PowerPCCPU *cpu = POWERPC_CPU(cs);
> > > > +
> > > > +    spapr_cpu_destroy(cpu);
> > > > +    cpu_remove_sync(cs);
> > > > +
> > > > +    /*
> > > > +     * We are still walking the core object's children list, and
> > > > +     * hence can't cleanup this CPU thread object just yet. Put
> > > > +     * it on a list for later removal.
> > > > +     */
> > > > +    spapr_add_cpu_to_unplug_list(obj);
> > > > +    return 0;
> > > > +}
> > > > +
> > > > +static void spapr_core_release(DeviceState *dev, void *opaque)
> > > > +{
> > > > +    object_child_foreach(OBJECT(dev), spapr_cpu_release, opaque);
> > > > +    spapr_cpu_core_cleanup();
> > > > +    object_unparent(OBJECT(dev));
> > > 
> > > I'd prefer to see the unplug_list as a local to this function (passed
> > > to spapr_cpu_release via opaque) rather than using a new global.
> > 
> > Will try that in the next iteration.
> > 
> > > 
> > > > +}
> > > > +
> > > > +static int spapr_core_detach(Object *obj, void *opaque)
> > > > +{
> > > > +    sPAPRCoreState *core = opaque;
> > > > +    DeviceState *dev = DEVICE(obj);
> > > > +    CPUState *cs = CPU(dev);
> > > > +    PowerPCCPU *cpu = POWERPC_CPU(cs);
> > > > +    int id = ppc_get_vcpu_dt_id(cpu);
> > > > +    int smt = kvmppc_smt_threads();
> > > > +    sPAPRDRConnector *drc =
> > > > +        spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_CPU, id);
> > > > +    sPAPRDRConnectorClass *drck;
> > > > +    Error *local_err = NULL;
> > > > +
> > > > +    /*
> > > > +     * SMT threads return from here, only main thread (thread 0) will
> > > > +     * continue and signal hot unplug event to the guest.
> > > > +     */
> > > 
> > > This seems silly.  spapr_core_unplug() iterates through all the
> > > children only to have all of them except thread 0 ignore the call..
> > > 
> > > Can't you just grab the first thread and do this logic directly from
> > > spapr_core_unplug()?
> > 
> > If I purely rely on the children list of the core object like I am doing
> > here, I don't see how to grab the first thread object from the list w/o
> > doing what I am doing here. However I can store the first thread object
> > in the core object during the core object's instance_init and use it here.
> > Will give it a try in the next iteration.
> 
> It should be accessible by name as "thread[0]" no?

Yes, like this AFAICS:

ObjectProperty *prop = object_property_find(OBJECT(core), "thread[0]", &local_err);
Object *thread = prop->opaque;

Storing the object pointer of the main thread in the core looks simpler to
me.

Regards,
Bharata.