Re: [Qemu-devel] [RFC PATCH v2 12/23] spapr: CPU hotplug support

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

On Wed, Mar 25, 2015 at 02:03:45PM +1100, David Gibson wrote:
> On Mon, Mar 23, 2015 at 07:05:53PM +0530, Bharata B Rao wrote:
> > Support CPU hotplug via device-add command. Set up device tree
> > entries for the hotplugged CPU core and use the exising EPOW event
> > infrastructure to send CPU hotplug notification to the guest.
> > 
> > Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
> > ---
> >  hw/ppc/spapr.c        | 75 +++++++++++++++++++++++++++++++++++++++++++++++++++
> >  hw/ppc/spapr_events.c |  8 +++---
> >  hw/ppc/spapr_rtas.c   | 11 ++++++++
> >  3 files changed, 91 insertions(+), 3 deletions(-)
> > 
> > diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> > index f52d38f..b48994b 100644
> > --- a/hw/ppc/spapr.c
> > +++ b/hw/ppc/spapr.c
> > @@ -33,6 +33,7 @@
> >  #include "sysemu/block-backend.h"
> >  #include "sysemu/cpus.h"
> >  #include "sysemu/kvm.h"
> > +#include "sysemu/device_tree.h"
> >  #include "kvm_ppc.h"
> >  #include "mmu-hash64.h"
> >  #include "qom/cpu.h"
> > @@ -660,6 +661,10 @@ static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset)
> >      QemuOpts *opts = qemu_opts_find(qemu_find_opts("smp-opts"), NULL);
> >      unsigned sockets = opts ? qemu_opt_get_number(opts, "sockets", 0) : 0;
> >      uint32_t cpus_per_socket = sockets ? (smp_cpus / sockets) : 1;
> > +    sPAPRDRConnector *drc =
> > +        spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_CPU, index);
> > +    sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
> > +    int drc_index = drck->get_index(drc);
> >  
> >      _FDT((fdt_setprop_cell(fdt, offset, "reg", index)));
> >      _FDT((fdt_setprop_string(fdt, offset, "device_type", "cpu")));
> > @@ -728,6 +733,8 @@ static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset)
> >  
> >      _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id",
> >                              cs->cpu_index / cpus_per_socket)));
> > +    _FDT((fdt_setprop_cell(fdt, offset, "ibm,my-drc-index", drc_index)));
> > +
> 
> What effect will this have when running with DR disabled?

I realize now that I shouldn't probably populate this at all in that case.
This routine is now common for both bootpath and hotplug path. Will take
care of this in next post.

> 
> >      _FDT(spapr_fixup_cpu_numa_smt_dt(fdt, offset, cs, spapr));
> >  }
> > @@ -1840,6 +1847,70 @@ static void spapr_nmi(NMIState *n, int cpu_index, Error **errp)
> >      }
> >  }
> >  
> > +static void spapr_cpu_hotplug_add(DeviceState *dev, CPUState *cs, Error **errp)
> > +{
> > +    PowerPCCPU *cpu = POWERPC_CPU(cs);
> > +    DeviceClass *dc = DEVICE_GET_CLASS(cs);
> > +    int id = ppc_get_vcpu_dt_id(cpu);
> > +    sPAPRDRConnector *drc =
> > +        spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_CPU, id);
> > +    sPAPRDRConnectorClass *drck = SPAPR_DR_CONNECTOR_GET_CLASS(drc);
> > +    void *fdt;
> > +    int offset, i, fdt_size;
> > +    char *nodename;
> > +
> > +    fdt = create_device_tree(&fdt_size);
> > +    nodename = g_strdup_printf("%s@%x", dc->fw_name, id);
> > +    offset = fdt_add_subnode(fdt, 0, nodename);
> > +
> > +    /* Set NUMA node for the added CPU core */
> > +    for (i = 0; i < nb_numa_nodes; i++) {
> > +        if (test_bit(cs->cpu_index, numa_info[i].node_cpu)) {
> > +            cs->numa_node = i;
> > +            break;
> > +        }
> > +    }
> > +
> > +    spapr_populate_cpu_dt(cs, fdt, offset);
> > +    g_free(nodename);
> > +
> > +    drck->attach(drc, dev, fdt, offset, !dev->hotplugged, errp);
> > +    if (*errp) {
> > +        g_free(fdt);
> > +    }
> > +}
> > +
> > +static void spapr_cpu_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
> > +                            Error **errp)
> > +{
> > +    CPUState *cs = CPU(dev);
> > +    PowerPCCPU *cpu = POWERPC_CPU(cs);
> > +    int id = ppc_get_vcpu_dt_id(cpu);
> > +    sPAPRDRConnector *drc =
> > +        spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_CPU, id);
> > +    int smt = kvmppc_smt_threads();
> > +    Error *local_err = NULL;
> > +
> > +    /*
> > +     * SMT threads return from here, only main thread (core) will
> > +     * continue and signal hotplug event to the guest.
> > +     */
> > +    if ((id % smt) != 0) {
> > +        return;
> > +    }
> > +
> > +    g_assert(drc);
> > +
> > +    spapr_cpu_hotplug_add(dev, cs, &local_err);
> > +    if (local_err) {
> > +        error_propagate(errp, local_err);
> > +        return;
> > +    }
> > +    spapr_hotplug_req_add_event(drc);
> > +
> > +    return;
> > +}
> > +
> >  static void spapr_machine_device_plug(HotplugHandler *hotplug_dev,
> >                                        DeviceState *dev, Error **errp)
> >  {
> > @@ -1848,6 +1919,10 @@ static void spapr_machine_device_plug(HotplugHandler *hotplug_dev,
> >          PowerPCCPU *cpu = POWERPC_CPU(cs);
> >  
> >          spapr_cpu_init(cpu);
> > +        spapr_cpu_reset(cpu);
> > +        if (dev->hotplugged && spapr->dr_cpu_enabled) {
> > +            spapr_cpu_plug(hotplug_dev, dev, errp);
> > +        }
> >      }
> >  }
> >  
> > diff --git a/hw/ppc/spapr_events.c b/hw/ppc/spapr_events.c
> > index be82815..4ae818a 100644
> > --- a/hw/ppc/spapr_events.c
> > +++ b/hw/ppc/spapr_events.c
> > @@ -421,14 +421,16 @@ static void spapr_hotplug_req_event(sPAPRDRConnector *drc, uint8_t hp_action)
> >      hp->hdr.section_length = cpu_to_be16(sizeof(*hp));
> >      hp->hdr.section_version = 1; /* includes extended modifier */
> >      hp->hotplug_action = hp_action;
> > -
> > +    hp->drc.index = cpu_to_be32(drck->get_index(drc));
> > +    hp->hotplug_identifier = RTAS_LOG_V6_HP_ID_DRC_INDEX;
> >  
> >      switch (drc_type) {
> >      case SPAPR_DR_CONNECTOR_TYPE_PCI:
> > -        hp->drc.index = cpu_to_be32(drck->get_index(drc));
> > -        hp->hotplug_identifier = RTAS_LOG_V6_HP_ID_DRC_INDEX;
> >          hp->hotplug_type = RTAS_LOG_V6_HP_TYPE_PCI;
> >          break;
> > +    case SPAPR_DR_CONNECTOR_TYPE_CPU:
> > +        hp->hotplug_type = RTAS_LOG_V6_HP_TYPE_CPU;
> > +        break;
> >      default:
> >          /* we shouldn't be signaling hotplug events for resources
> >           * that don't support them
> > diff --git a/hw/ppc/spapr_rtas.c b/hw/ppc/spapr_rtas.c
> > index 57ec97a..48aeb86 100644
> > --- a/hw/ppc/spapr_rtas.c
> > +++ b/hw/ppc/spapr_rtas.c
> > @@ -121,6 +121,16 @@ static void rtas_query_cpu_stopped_state(PowerPCCPU *cpu_,
> >      rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
> >  }
> >  
> > +static void spapr_cpu_set_endianness(PowerPCCPU *cpu)
> > +{
> > +    PowerPCCPU *fcpu = POWERPC_CPU(first_cpu);
> > +    PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(fcpu);
> 
> In some ways it might be nicer to store the global endian mode in
> sPAPRMachineState, updating it on H_SET_MODE, rather than copying the
> first cpu, although I guess that works.

I can give it a stab at that if you feel strongly about it.

Regards,
Bharata.