On Tue, 04 Nov 2014 09:45:15 +0100 Paolo Bonzini <pbonz...@redhat.com> wrote:
> On 03/11/2014 20:11, Igor Mammedov wrote: > >> > Understood. But I think -m slots=too_much is not an important > >> > problem. What is problematic is filling those slots, which patch 1 > >> > will help with. > > From user's pov it's still error, i.e. he asked on CLI for one amount > > of hotpluggable slots and couldn't use all of them at runtime. > > But he cannot even with your patch, if he was off by just a couple > items. So I don't think the improvement is worth the extra complication > of the code and the small layering violation. With this patch user can launch QEMU with exact max amount of possible slots with exception that he won't hotplug other than pc-dimm devices that consume KVM's memory slots (not sure if such exists (hotpluggable)). but, sure we can drop this patch until someone complains. How about adding QMP interface to query amount of possible slots, so libvirt could get a hint about max possible value?
