On 06/06/2017 20:19, Roman Kagan wrote: > + > + if ((atomic_fetch_or(&flags[set_idx], set_mask) & set_mask) != set_mask) > { > + ret = kvm_hv_sint_route_set_sint(sint_route); > + memory_region_set_dirty(&synic->evt_page_mr, 0, > + sizeof(*synic->evt_page));
I would swap the kvm_hv_sint_route_set_sint and memory_region_set_dirty lines. Paolo > + } else { > + ret = 0; > + }