Re: [Qemu-devel] [RFC PATCH v1 3/5] spapr: Implement CPUClass::get_migration_id() for PowerPC CPUs

[Greg Kurz <groug@kaod.org>]

[-- Attachment #1: Type: text/plain, Size: 2467 bytes --]

On Thu, 7 Jul 2016 10:57:28 +1000
David Gibson <david@gibson.dropbear.id.au> wrote:

> On Wed, Jul 06, 2016 at 07:51:01PM +0530, Bharata B Rao wrote:
> > On Wed, Jul 06, 2016 at 02:01:14PM +0200, Igor Mammedov wrote:  
> > > On Wed,  6 Jul 2016 14:29:19 +0530
> > > Bharata B Rao <bharata@linux.vnet.ibm.com> wrote:
> > >   
> > > > cpu_index is used as migration_id by default. For machine type
> > > > versions that set use-migration-id property, cpu_dt_it is returned.
> > > > 
> > > > Signed-off-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
> > > > ---
> > > >  target-ppc/translate_init.c | 12 ++++++++++++
> > > >  1 file changed, 12 insertions(+)
> > > > 
> > > > diff --git a/target-ppc/translate_init.c b/target-ppc/translate_init.c
> > > > index efd6b88..9ca2f5e 100644
> > > > --- a/target-ppc/translate_init.c
> > > > +++ b/target-ppc/translate_init.c
> > > > @@ -10359,6 +10359,17 @@ static gchar *ppc_gdb_arch_name(CPUState *cs)
> > > >  #endif
> > > >  }
> > > >  
> > > > +static int ppc_cpu_get_migration_id(CPUState *cs)
> > > > +{
> > > > +    PowerPCCPU *cpu = POWERPC_CPU(cs);
> > > > +
> > > > +    if (cs->use_migration_id) {
> > > > +        return (int) cpu->cpu_dt_id;  
> > > Could cpu_dt_id have value bigger than 32bit int? If yes, it's not safe
> > > to do so, that's the reason why I'm going to use index in possible_cpus
> > > on ARM and for the sake of uniformity do the same for x86 (even though
> > > it's possible to use 32bit APIC ID).  
> > 
> > For the existing max_cpus limit, 32 bit should be fine, but obviously
> > that is not good and future safe.  
> 
> I think we should be OK.  The hardware IDs these are modelled on are
> 32-bit - I'm pretty sure #address-cells in /cpus is 1, not 2.
> 
> And even with a lot of wastage for alignment gaps of various sorts,
> we're a *long* way off from supporting 4 billion cpus...
> 

Indeed ! Even in the "worst" case, which is running a guest with 1 thread
per core on a POWER8 host with full cores (8 HW threads), the current maximum
value for cpu_dt_id is 254 * 8 == 2032 :)

> > I had a brief look at the implementation around possible_cpus in
> > x86, let me see if I can do something similar for generating stable
> > ids for PowerPC. I thought device tree ID would be our stable id, but
> > that being 64 bit and migration code requiring 32 bit value isn't
> > helping :(
> > 
> > Regards,
> > Bharata.
> >   
> 


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 181 bytes --]