Re: [Qemu-devel] [RFC 0/5] Allow object-add on X86CPU subclasses, for CPU model probing

[Igor Mammedov]

On Thu, 15 May 2014 11:03:49 -0300
Eduardo Habkost <address@hidden> wrote:

> On Thu, May 15, 2014 at 03:48:16PM +0200, Igor Mammedov wrote:
> > On Thu, 15 May 2014 10:07:51 -0300
> > Eduardo Habkost <address@hidden> wrote:
> > 
> > > On Thu, May 15, 2014 at 02:35:01PM +0200, Igor Mammedov wrote:
> > > > On Tue, 06 May 2014 22:29:24 +0200
> > > > Andreas Färber <address@hidden> wrote:
> > > > 
> > > > > Am 06.05.2014 22:19, schrieb Eduardo Habkost:
> > > > > > On Tue, May 06, 2014 at 10:01:11PM +0200, Igor Mammedov wrote:
> > > > > >> On Tue, 6 May 2014 11:42:56 -0300
> > > > > >> Eduardo Habkost <address@hidden> wrote:
> > > > > >>> On Tue, May 06, 2014 at 09:22:38AM +0200, Igor Mammedov wrote:
> > > > > >>>> On Fri, 2 May 2014 11:43:05 -0300
> > > > > >>>> Eduardo Habkost <address@hidden> wrote:
> > > > > >>>>> On Fri, May 02, 2014 at 03:45:03PM +0200, Igor Mammedov wrote:
> > > > > >>>>>> On Wed, 30 Apr 2014 17:29:28 -0300
> > > > > >>>>>> Eduardo Habkost <address@hidden> wrote:
> > > > > >>>>>>> This series allows management code to use object-add on 
> > > > > >>>>>>> X86CPU subclasses, so it
> > > > > >>>>>> Is there any reason why "device-add" couldn't be used?
> > > > > >>>>> It needs to work with "-machine none", device_add requires a 
> > > > > >>>>> bus to
> > > > > >>>>> exist, and there is no icc-bus on machine_none.
> > > > > >>>> The thing is that CPUID is a function of machine so using
> > > > > >>>> "-machine none" will provide only approximately accurate data.
> > > > > >>>> I'm not sure that retrieved possibly not accurate data are useful
> > > > > >>>> for libvirt.
> > > > > >>> "-cpu host" doesn't depend on machine, and is the most important 
> > > > > >>> thing
> > > > > >>> right now (because libvirt's checks for host QEMU/kernel/CPU
> > > > > >>> capabilities is completely broken).
> > > > > >> true, but device-add/-cpu host could be used right now to get the
> > > > > >> same CPUID data wile using any machine type or default one without
> > > > > >> any of this patches.
> > > > > > 
> > > > > > device_add can't be used with "-machine none".
> > > > > 
> > > > > I see no reason why we couldn't *make* CPUs work on -machine none. The
> > > > > ICC bus and bridge were a hack to make APIC(?) hot-add work in face of
> > > > > SysBus; if that prohibits other valid uses now, then evaluating Igor's
> > > > > memory work for CPU might be an option.
> > > > Yep, if CPU is hot-plugged as bus-less device.
> > > > There is a little concern of APIC device if we go that direction since
> > > > in addition to hotpluggable BUS, BUS provides address-space for APIC 
> > > > MMIO.
> > > > With that resolved, x86-cpu shouldn't depend on any bus and if there 
> > > > isn't
> > > > any current user that uses QOM path to CPU for introspecting (Eduardo's
> > > > ABI concern), then it could be done in time for 2.1.
> > > 
> > > Maybe there are no users of the current QOM path, but we do need a
> > > stable path to allow management to locate the CPU objects. Do we have
> > > one, already?
> > > 
> > 
> > Can't we add query-cpus QMP command or something like this to hide path
> > from user.
> 
> That would work, too. But why is a dedicated "query-cpus" command better
> than something like "qom-list path=/machine/cpus" (that could simply
> return a list of links to the actual CPU objects)?
So that not to stall the work on deciding on
 - if exposing not yet stables QOM paths as stable ABI is a good thing, I
   recall Andreas objecting to using QOM paths with device hotplug
 - what paths to CPUs should be wrt stalled topology discussion

> 
> -- 
> Eduardo
> 


-- 
Regards,
  Igor