Re: [Qemu-devel] [RFC PATCH v1 2/5] cpu: Introduce CPUState::migration_id

[Bharata B Rao]

On Wed, Jul 06, 2016 at 01:34:59PM +0200, Igor Mammedov wrote:
> On Wed,  6 Jul 2016 14:29:18 +0530
> Bharata B Rao <address@hidden> wrote:
> 
> > Add CPUState::migration_id and use that as instance_id in
> > vmstate_register() call.
> > 
> > Introduce use-migration-id property that allows target machines to
> > optionally switch to using migration_id instead of cpu_index.
> > This will help allow successful migration in cases where holes are
> > introduced in cpu_index range after CPU hot removals.
> > 
> > The default implementation of CPUClass:get_migration_id() continues
> > to return cpu_index.
> > 
> > Suggested-by: Igor Mammedov <address@hidden>
> > Signed-off-by: Bharata B Rao <address@hidden>
> > ---
> >  exec.c            |  5 +++--
> >  include/qom/cpu.h |  6 ++++++
> >  qom/cpu.c         | 21 +++++++++++++++++++++
> >  3 files changed, 30 insertions(+), 2 deletions(-)
> > 
> > diff --git a/exec.c b/exec.c
> > index fb73910..a8afeda 100644
> > --- a/exec.c
> > +++ b/exec.c
> > @@ -619,12 +619,13 @@ static void cpu_release_index(CPUState *cpu)
> >  void cpu_vmstate_register(CPUState *cpu)
> >  {
> >      CPUClass *cc = CPU_GET_CLASS(cpu);
> > +    int instance_id = cc->get_migration_id(cpu);
> >  
> >      if (qdev_get_vmsd(DEVICE(cpu)) == NULL) {
> > -        vmstate_register(NULL, cpu->cpu_index, &vmstate_cpu_common,
> > cpu);
> > +        vmstate_register(NULL, instance_id, &vmstate_cpu_common,
> > cpu); }
> >      if (cc->vmsd != NULL) {
> > -        vmstate_register(NULL, cpu->cpu_index, cc->vmsd, cpu);
> > +        vmstate_register(NULL, instance_id, cc->vmsd, cpu);
> >      }
> >  }
> >  
> > diff --git a/include/qom/cpu.h b/include/qom/cpu.h
> > index 29ccf5c..6d00143 100644
> > --- a/include/qom/cpu.h
> > +++ b/include/qom/cpu.h
> > @@ -187,6 +187,7 @@ typedef struct CPUClass {
> >      bool (*cpu_exec_interrupt)(CPUState *cpu, int interrupt_request);
> >  
> >      void (*disas_set_info)(CPUState *cpu, disassemble_info *info);
> > +    int (*get_migration_id)(CPUState *cpu);
> >  } CPUClass;
> >  
> >  #ifdef HOST_WORDS_BIGENDIAN
> > @@ -273,6 +274,9 @@ struct qemu_work_item {
> >   * @kvm_fd: vCPU file descriptor for KVM.
> >   * @work_mutex: Lock to prevent multiple access to queued_work_*.
> >   * @queued_work_first: First asynchronous work pending.
> > + * @migration_id: Use as instance_id argument in cpu
> > vmstate_register calls
> with this field get_migration_id() callback is redundant.
> You can drop get_migration_id() and assign value to migration_id in
> machine specific code /spapr/ppc/x86/.../ before realize is called.
> That should simplify this patch and its users as well.

Let me give that a try and see how it looks.

> 
> > + * @use_migration_id: Set to enforce the use of
> > CPUClass.get_migration_id()
> > + *     over cpu_index during vmstate registration.
> >   *
> >   * State of one CPU core or thread.
> >   */
> > @@ -360,6 +364,8 @@ struct CPUState {
> >         (absolute value) offset as small as possible.  This reduces
> > code size, especially for hosts without large memory offsets.  */
> >      uint32_t tcg_exit_req;
> > +    int migration_id;
> > +    bool use_migration_id;
> >  };
> >  
> >  QTAILQ_HEAD(CPUTailQ, CPUState);
> > diff --git a/qom/cpu.c b/qom/cpu.c
> > index 488ecc6..01cf136 100644
> > --- a/qom/cpu.c
> > +++ b/qom/cpu.c
> > @@ -326,6 +326,17 @@ static void cpu_common_realizefn(DeviceState
> > *dev, Error **errp) }
> >  }
> >  
> > +static bool cpu_common_get_use_migration_id(Object *obj, Error
> > **errp) +{
> > +    return CPU(obj)->use_migration_id;
> > +}
> > +
> > +static void cpu_common_set_use_migration_id(Object *obj, bool value,
> > +                                            Error **err)
> > +{
> > +    CPU(obj)->use_migration_id = value;
> > +}
> It's possible to create dumb property in a more compact way,
> see DEFINE_PROP_BOOL()

I stated with it and defined dc->props, but then felt it is a bit too
much to define a list just for one property. But if that preferable
let me switch back to it.

Regards,
Bharata.