[Libvir] [PATCH] check the maximum of virtual CPU

[Daniel P. Berrange]

On Mon, Mar 05, 2007 at 03:42:35PM +0900, Masayuki Sunou wrote:
> I removed a needless part from the former patch.
> 
> Add check the maxmum of virtual CPU.

I was thinking about how to make use of this new API in virt-manager 
when I came up with a further complication :-) There are three scenarios
where I'd like to be able to find out the maximum VCPUs possible for
a guest VM:

 1. When initially creating a VM
 2. When changing the config of an inactive VM
 3. When changing the config of a running VM

Thus far in this thread, we've focused on addressing the last point,
but I think its worth trying address all 3 while we're looking at
this area.

In the first case, we do not have a virDomainPtr object at all yet,
all we have is a virConnectPtr and info about the type of domain
we're planning to create - Xen, QEMU, KVM, KQEMU. When creating a
VM we need to apply some domain type specific limit.

In the case of an inactive domain, we have a virDomainPtr object
and from that the internal drivers can determine the domain type
or Xen, QEMU, KVM, etc, and again apply some domain type specific
limit.

In the 3rd case of a running VM though, the hypervisor limit is
typically not neccessarily the one that is relevant. While the
hypervisor will still has its own limit of VCPUs, the effective
limit is likely lower - ie fixed at the number of virtual CPUs
which were allocated when the guest OS booted. So if you boot a
guest with 4 cpus, you can hotplug remove & add CPUs, but you
can never go above 4 until the guest is shutdown.

So anyway, I think we need to add 2 apis to address this whole
issue:

  -  int virConnectGetMaxVcpus(virConnectPtr conn, 
                               const char *type);

      Returns the maximum number of virtual CPUs supported for
      a guest VM of a specific type. Thje 'type' parameter here
      corresponds to the 'type' attribute in the <domain>
      element of the XML.

  -  int virDomainGetMaxVcpus(virDomainPtr dom);

      Returns the maximum number of virual CPUs supported for
      the guest VM. If the guest is inactive, this is basically
      the same as virConnectGetMaxVcpus. If the guest is running
      this will reflect the maximum number of VCPus the guest
      was booted with.

In terms of the Xen implementation, the virConnectGetMaxVcpus
method can simply return  MAX_VIRT_VCPUS. Likewie virDomainGetMaxVcpus
can do the same for inactive guests, but if the guest is running then
we will need to call out to XenD to extract the max vcpus info.

Any other thoughts from people on the list ???

Dan.
-- 
|=- Red Hat, Engineering, Emerging Technologies, Boston.  +1 978 392 2496 -=|
|=-           Perl modules: http://search.cpan.org/~danberr/              -=|
|=-               Projects: http://freshmeat.net/~danielpb/               -=|
|=-  GnuPG: 7D3B9505   F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505  -=|