On Tue, 22 Mar 2011 15:55:56 +0300 Dmitry Konishchev <konishc...@gmail.com> wrote:
> Hi! I use QEMU via QMP and I've discovered that for some tasks there > is no proper way to do them via QMP. I've written few patches: I'm going to respond you without looking at the patches. > * One of them modifies "pci_add" command to return pci address of the > added device, when user hasn't specified it (to be able to delete it > via "pci_del" in the future). The pci_add command is not available in QMP because it's a bad interface, we use device_add (for the guest part) and additional commands like netdev_add for the host part. Unfortunately, block hot plug is still missing. This is going to be done by blockdev_add, which is under development. Any updates, Markus? You can use the human-monitor-command as a workaround if you want to, but please, be sure to read its documentation, specially the part that says that this is not a stable interface. > * The second one adds INCOMING_FINISHED QMP event which is emitted > when QEMU finished incoming migration (when started with -incoming > command line option). It is needed because now there is no way to > determine, whether it finished or not, and QMP "cont" command just > returns error. We had a long discussion about this and the conclusion was that adding a new event like this would be more a workaround than a proper solution, the proper solution being a truly asynchronous command which would not need additional events. However, we lack infrastructure to do this today, but this is likely to be solved by the QAPI (our brand new internal QMP machinery). The current workaround is to use query-migrate to pull the VM in order to know when the migration has finished. > > It will be awesome if you include this patches to the upstream (I've > attached them to this message). >
