[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: [tor-bugs] #9093 [Tor]: Better, fairer circuit OOM handling



#9093: Better, fairer circuit OOM handling
-----------------------------+----------------------------------------
     Reporter:  nickm        |      Owner:
         Type:  enhancement  |     Status:  needs_review
     Priority:  major        |  Milestone:  Tor: 0.2.4.x-final
    Component:  Tor          |    Version:
   Resolution:               |   Keywords:  tor-relay 023-backport oom
Actual Points:               |  Parent ID:
       Points:               |
-----------------------------+----------------------------------------

Comment (by arma):

 Patch looks fine. The only part that popped out as weird is where
 circuit_max_queued_cell_age() looks at both the age in the n_conn and also
 the age in the p_conn.

 We've been mostly phrasing the attack as an attack on guards, so in that
 case only p_conn matters. But I guess we could imagine this attack where
 you extend your circuit to a colluding relay (maybe even not one in the
 consensus), and then do the attack in reverse. Sounds good.

 The other variant of the attack, where we upload a large file to a
 colluding webserver which stops reading, in hopes of filling the ram at
 the exit relay... I guess the exit will stop sending sendmes when the
 webserver stops accepting bytes. Which means the attack would be a
 parallel one, where you queue your 500 cells (250KB) and then move on to a
 new connection to do it again. Is that variant handled by this fix (do
 most of those cells hang out on an edge outbuf, or in a circuit queue?),
 or is it a new ticket?

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/9093#comment:3>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online
_______________________________________________
tor-bugs mailing list
tor-bugs@xxxxxxxxxxxxxxxxxxxx
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-bugs