[tor-bugs] #2746 [Tor Relay]: libevent2 problems on CCC tor servers

[Tor Bug Tracker & Wiki]

#2746: libevent2 problems on CCC tor servers
-----------------------+----------------------------------------------------
 Reporter:  kaner      |          Owner:                     
     Type:  defect     |         Status:  new                
 Priority:  normal     |      Milestone:  Tor: 0.2.2.x-final 
Component:  Tor Relay  |        Version:  Tor: 0.2.2.22-alpha
 Keywords:             |         Parent:                     
   Points:             |   Actualpoints:                     
-----------------------+----------------------------------------------------

Comment(by nickm):

 Hm.  We should think about next steps on this.  First thing to do is write
 test code and see if we can make kevent() report EBADF in the cases we
 think it should (and not in the cases where we think it shouldn't).  We
 should also have a quick look at the kernel in question to see if there
 are other circumstances where kevent() can return EBADF that appear from
 reading the source.

 Once *that's* done, next step is to make absolutely sure that libevent is
 reporting this bug right.  We should look at the code, and make sure it's
 obviously correct.  Also we should have a quick gander at which patches,
 if any, freebsd applies to libevent.

 And once *that's* done, if we want to go on the theory that something is
 closing the kqueue fd, let's see if we can instrument close() in libevent
 and tor such that every time it's about to close, it asserts that the fd
 is not the kqueue fd.  That'd be a hack thatwe would never want to merge,
 but if the bug *does* turn out to be a bogus close, we'll get a stack
 trace to tell us where it is.

 Progress!

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/2746#comment:8>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online