On Thursday 10 February 2011 11:42:50 Linus Lüssing wrote: > > Would speak anything against re-using the gw_list_lock ? > > ... as we are usually changing the gw_list more often than the > curr_gw, so it's not really necessary to let a gw_list change for > another node wait for a curr_gw_node reassignment to finish.
Well, we only need the list lock when we are adding/deleting items from the list which does not happen that often nor is it time critical. > What is speaking for using the same lock for both, just having > less spinlocks in total in the code? Yes. Regards, Marek
