On 10/04/2011 03:49 PM, Luiz Capitulino wrote:
There's a semantic change which I'm not completely sure it won't generate unexpected side-effects: today vm_stop() will only carry any action if the machine is running, otherwise it's no-op. This patch changes that.
More or less, yes. I tried to limit the semantic change by not running notifiers, which again could be better or worse.
I don't think adding a new transition is a good solution, because you'll have to add a transition from PAUSED to anything that uses runstate_set instead of vm_stop.
However, you could change all vm_stop() to vm_stop(RSTATE_PAUSED) followed by runstate_set(), adjust the transition table consequently and possibly drop the argument to vm_stop. I tried to get the smallest patch, but I did need to follow-up with changes to the transition table.
In any case, can I assume this to be in your hands now? :) Paolo
