On Fri, 2014-09-26 at 20:57 +0200, Nikos Mavrogiannopoulos wrote: > > > That doesn't matter much for ocserv as there are no multiple threads. It > > > was added as it looked reasonable for other OSes which may not behave as > > > Linux. > > > > I'm not convinced there are any. I know of no other userspace which will > > retry a close() call. And it's hard to imagine how it could ever be > > necessary. The *flush* might fail, but closing the file descriptor is > > always easy enough because it just involves throwing away local state. > > > > Also, we close all file descriptors when a process exits. If close() can > > fail, that makes a horrid mess of the task exit code path. I dread to > > think how this could ever be handled. I just don't think it can happen. > > Niels, could you try this patch then (against master). If close() indeed > fails you should see "close failed:" in syslog. Otherwise we should > figure the real reason behind that issue. An additional way to obtain more information is lsof -p PID. Where PID is the PID of the main ocserv process (should be listed as ocserv-main in ps -e), as well as to the worker process that corresponds to the connection that caused the issue (the PID is the same as the ID in occtl users). regards, Nikos
