Re: [Qemu-devel] [RFC PATCH v2.1 00/12] Core based CPU hotplug for PowerPC sPAPR

[Igor Mammedov]

On Tue, 5 Apr 2016 20:25:57 +0530
Bharata B Rao <address@hidden> wrote:

> On Mon, Apr 04, 2016 at 04:44:27PM +0200, Igor Mammedov wrote:
> > On Thu, 31 Mar 2016 14:09:09 +0530
> > Bharata B Rao <address@hidden> wrote:
> > 
> > > Hi,
> > > 
> > > This is v2.1 of "Core based CPU hotplug for PowerPC sPAPR". v2 was
> > > posted here:
> > > https://lists.gnu.org/archive/html/qemu-ppc/2016-03/msg00201.html
> > > 
> > > I am making this v2.1 instead of v3 as this introduces a change in
> > > device_add semantics and if we don't want to continue this, I
> > > will go back to the previous v2 semantics from v3 onwards. v2 had
> > > spapr-cpu-core device that served as CPU core device for all
> > > types of PowerPCCPU threads. In this version however,
> > > spapr-cpu-core becomes an abstract device and we will have
> > > different core devices for each of the PowerPCCPU type. So in
> > > this version, the hotplug semantics looks like this:
> > > 
> > > (qemu) device_add
> > > POWER8-spapr-cpu-core,id=core2,core=16[,threads=4]
> > > 
> > > cpu_model specification is gone as it becomes redundant with
> > > different core types for each CPU type. CPU core types are
> > > defined only for host, POWER7 and POWER8 models only (and not for
> > > their variants/aliases) yet.
> > > 
> > > I have mostly taken care of all the review comments that I got
> > > for v2.
> > Could you rebase on top of current master, pls?
> 
> Sure, I usually update, but missed it this time.
> 
> > 
> > and fix compile error:
> > hw/cpu/core.c: In function ‘core_prop_set_core’:
> > hw/cpu/core.c:31:9: error: implicit declaration of function
> > ‘error_propagate’ [-Werror=implicit-function-declaration]
> > error_propagate(errp, local_err);
> 
> I don't see this when compiling all the targets either on x86 or PPC
> systems. May be this is seen with latest master, in any case will
> check this out when spinning next version.
> 
> However do you have any specific comment to be addressed from this
> version before I post the next one ?
Sorry, I haven't had a spare time to review it yet.

But device_add example looks good and similar what x86 will do,
which is good from libvirt point of view as it would be able to
implement hotplug in a uniform way for both targets.

> 
> Regards,
> Bharata.
>