On 04/15/2013 11:04 AM, Daniel P. Berrange wrote:
On Mon, Apr 15, 2013 at 12:49:12PM -0400, John Ferlan wrote:
>
https://bugzilla.redhat.com/show_bug.cgi?id=906644
>
> Added checks to both virsh suspend and virsh resume for the domain to be
> in a the right state before trying the suspend/resume. Similar checks to
> examples/domsuspend/suspend.c.
IMHO this is just a pointless bug request. State checks don't
belong in virsh for a start, since that makes it inherantly
racey. While the drivers do check for whether the domain is
running, they explicitly chose not to raise an error if the
VM is already paused, when pause is executed & vica-verca
I agree that doing it in virsh is too racy. If anything, we would need
to implement new virDomainSuspendFlags() and virDomainResumeFlags() to
let the user pass in a flag that controls whether or not they want
libvirtd to reject a no-op state change (we can't change the default in
libvirtd, for fear of breaking existing clients, but the only way to do
a non-racy non-default behavior is to add a flag which requires adding API).
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library
http://libvirt.org