On Thu, Jun 23, 2016 at 3:22 PM, Willem Jan Withagen <wjw@xxxxxxxxxxx> wrote: > On 23-6-2016 00:16, Willem Jan Withagen wrote: >> On 22-6-2016 16:36, Haomai Wang wrote: >>> Oh, sorry. I still realize you are testing on kqueue event backend. >>> >>> I submit a pr to fix this. plz help to verify whether it works for you >>> since I don't have bsd handy... >>> >>> https://github.com/ceph/ceph/pull/9869 >> >> I think add_event needs about the same treatment. >> Trying, Testing ATM.... > > errgh, not quite... > > It also generates errors when trying to delete EVFILT_READ (mask=2) from > an eventfilter that has both READ and WRITE set (mask=3). > Next to that the ms_async_messenger threads seem to be busy_waiting > looping and loading a full CPU core per thread. > > So I think I need some testing code to see what the requirements of > kqueue actually are, compared to what async_messenger does. > > Could be that if we want to go from (READ|WRITE) to either READ or WRITE > the event needs to be deleted first and than added anew. Hmm, I think I need to reread kqueue man page to figure out the problem... > > --WjW > -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
