On 03/21/2011 01:58 AM, Daniel Veillard wrote:
Rather than add a new API, now that we have
virDomainSetMemoryFlags()
I don't see why we could not use and extra flag on that API,
VIR_DOMAIN_MAX_MEM and use the same entry point.
Agreed, and this was how I handled the vcpu issue (we had separate calls
for max vcpus vs. current vcpus, but the new vcpuFlag api covers both
cases).
virDomainSetMaxMemory and virDomainSetMemory are separate because we
didn't add a flag for them, that was a mistake, but now we can
tunnel maximum memory changes though virDomainSetMemoryFlags()
We would then allow
VIR_DOMAIN_MEM_LIVE -> change live memory target
VIR_DOMAIN_MEM_CONFIG -> change memory target on config i.e. next reboot
VIR_DOMAIN_MEM_LIVE | VIR_DOMAIN_MAX_MEM -> change live max memory (if feasible)
This would imply memory hot-plugging, which is in itself an interesting
topic not currently supported by qemu, but which might be worth adding
someday in the future. Of course, the API would fail where it is not
supported, but I agree that this flag combination might become useful in
the future.
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library
http://libvirt.org