At Thu, 02 May 2013 11:33:18 +0800, Liu Yuan wrote: > > On 05/02/2013 10:51 AM, MORITA Kazutaka wrote: > > I'm not sure if I understand your suggestion correctly, but from the > > point of view that we have to set false to .in_use and close the > > timeout fd in the same lock context, adding another handler doesn't > > look a good idea to me. > > I meant a new helper function sheep_del_sockfd_timeout() that only close > the timeout-fd instead of trying to close all the fds and remove the nid > for the timeout condition.
Makes sense, looks better than modifying sockfd_cache_del() to me. I've noticed that we call sheep_del_sockfd() only where timeout can happen. I think of modifying the function so that it always close only the specified sockfd. Thanks, Kazutaka -- sheepdog mailing list sheepdog@lists.wpkg.org http://lists.wpkg.org/mailman/listinfo/sheepdog
