kvm: check KVM_CAP_NR_VCPUS with kvm_vm_check_extension()

kvm: check KVM_CAP_NR_VCPUS with kvm_vm_check_extension()

On a modern server-class ppc host with the following CPU topology:

Architecture:          ppc64le
Byte Order:            Little Endian
CPU(s):                32
On-line CPU(s) list:   0,8,16,24
Off-line CPU(s) list:  1-7,9-15,17-23,25-31
Thread(s) per core:    1

If both KVM PR and KVM HV loaded and we pass:

        -machine pseries,accel=kvm,kvm-type=PR -smp 8

We expect QEMU to warn that this exceeds the number of online CPUs:

Warning: Number of SMP cpus requested (8) exceeds the recommended
 cpus supported by KVM (4)
Warning: Number of hotpluggable cpus requested (8) exceeds the
 recommended cpus supported by KVM (4)

but nothing is printed...

This happens because on ppc the KVM_CAP_NR_VCPUS capability is VM
specific  ndreally depends on the KVM type, but we currently use it
as a global capability. And KVM returns a fallback value based on
KVM HV being present. Maybe KVM on POWER shouldn't presume anything
as long as it doesn't have a VM, but in all cases, we should call
KVM_CREATE_VM first and use KVM_CAP_NR_VCPUS as a VM capability.

This patch hence changes kvm_recommended_vcpus() accordingly and
moves the sanity checking of smp_cpus after the VM creation.

It is okay for the other archs that also implement KVM_CAP_NR_VCPUS,
ie, mips, s390, x86 and arm, because they don't depend on the VM
being created or not.

Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <150600966286.30533.10909862523552370889.stgit@bahia.lan>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>