On Fri, Mar 11, 2011 at 08:26:43AM -0700, Eric Blake wrote:
On 03/10/2011 03:53 PM, Eric Blake wrote:
>> Hmm, for other API additions, I have provided dummy wrappers for the new
>> function anywhere the old function existed, that merely require the same
>> flags as the old function would use. I'll probably propose that as a
>> followup patch.
>
> Like so.
>
> src/esx/esx_driver.c | 17 ++++++++++++++---
> src/lxc/lxc_driver.c | 15 +++++++++++++--
> src/test/test_driver.c | 19 ++++++++++++++++---
> src/uml/uml_driver.c | 15 +++++++++++++--
> src/vbox/vbox_tmpl.c | 15 +++++++++++++--
> src/xen/xen_driver.c | 15 ++++++++++++++-
Hmm, re-reading this patch exposes a problem (existing before this
patch, and this patch doesn't change the situation):
In 0.8.8, 'virsh setmem' on xen:/// meant:
if guest is live, then affect live (only); otherwise affect config
But with this patch, 'virsh setmem --live' will _still_ affect a config.
What we need is three flags instead of two:
VIR_DOMAIN_MEM_LIVE
VIR_DOMAIN_MEM_CONFIG
VIR_DOMAIN_MEM_CURRENT - use LIVE or CONFIG according to domain state
where we document that virDomainSetMemory is hypervisor-defined whether
it implies CURRENT (xen) or LIVE (qemu), and where
virDomainSetMemoryFlags properly honors all three flags for both
hypervisors.
Ideally VIR_DOMAIN_MEM_CURRENT would have value 0 and thus be
the default. So the new API default matches the old API
Daniel
--
|:
http://berrange.com -o-
http://www.flickr.com/photos/dberrange/ :|
|:
http://libvirt.org -o-
http://virt-manager.org :|
|:
http://autobuild.org -o-
http://search.cpan.org/~danberr/ :|
|:
http://entangle-photo.org -o-
http://live.gnome.org/gtk-vnc :|