On Aug 02 2013, liu ping fan wrote: > On Thu, Aug 1, 2013 at 10:28 PM, Paolo Bonzini <pbonz...@redhat.com> wrote: > > > > > > So actually there is another problem with this patch (both the > > > > > > condvar and the event approach are equally buggy). If a timer > > > > > > on clock X disables clock X, qemu_clock_enable will deadlock. > > > > Checking the code for callers of qemu_clock_enable() it seems like there > > shouldn't be any deadlocks. So it should work if qemu_clock_enable > > Do you mean the caller of qemu_clock_enable(foo,false) can NOT be > timer cb?
Yes. > As for this deadlock issue, making > qemu_clock_enable(foo,false) ASYNC, and forcing the caller to sync > explicitly ? Right now the callers of qemu_clock_enable(foo, false) are safe. So the problem can be "fixed" just by adequate documentation. > > But we should document the expectations since they are different from > > the current code. > > > > Paolo
