> From: Paolo Bonzini [mailto:pbonz...@redhat.com] > Il 27/08/2014 12:30, Pavel Dovgaluk ha scritto: > >> From: Paolo Bonzini [mailto:pbonz...@redhat.com] > >> Il 27/08/2014 12:15, Pavel Dovgaluk ha scritto: > >>>>> Again, this is only needed in your record/replay system (and you haven't > >>>>> yet quite explained why the design has this limitation), so it should > >>>>> not be a part of this series. > >>> I see. Updating s->timer and s->TimerExpire in post_load will be enough? > >> > >> In theory it should be done already, I guess it's not deterministic > >> enough though. > > > > I split existing function into the two parts: one sets irq (and calls > > another). > > And the second part sets only timer and TimerExpire fields. The second > > function > > is also called from post_load. > > > >> Have you tried my patch to rewrite the timer stuff? > > > > What patch do you mean? > > This one: > > http://article.gmane.org/gmane.comp.emulators.qemu/288521
It will solve the problem, because it removes raising an irq from rtl8139_set_next_tctr_time function. Ok then, I'll remove my rtl8139 patch from the series. Pavel Dovgalyuk