[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]
Re: [tor-bugs] #7440 [Tor]: Path Bias code should count circuit destructions as failures
#7440: Path Bias code should count circuit destructions as failures
-----------------------------------------+----------------------------------
Reporter: mikeperry | Owner: mikeperry
Type: defect | Status: new
Priority: major | Milestone: Tor: 0.2.4.x-final
Component: Tor | Version:
Keywords: MikePerry201211, tor-client | Parent:
Points: | Actualpoints:
-----------------------------------------+----------------------------------
Comment(by mikeperry):
My plan for this is to use either circ->isolation_any_streams_attached or
circ->timestamp_dirty in combination with a new field to represent stream
success (something like circ->got_stream_reply) to do this check.
If a successfully opened circuit has had any stream attempts but none that
succeeded (circ->isolation_any_streams_attached &&
!circ->got_stream_reply) in circuit_mark_for_close(), then we subtract a
success and add a failure for that circuit for its guard. We'll want to
store the number of circs this happens on in a separate guard PathBias
field, for debugging.
Technically, this check should cover any attempts to either choke a
circuit or destroy it once streams are attached.
However, in case there are non-destructive tagging mechanisms that allow
the circuit to survive construction (such as reply-only tags, perhaps?)
but allow it to be destroyed *before* any streams are attempted, we should
keep a separate count for circuits closed prematurely due to any remote
reason. I think this can also be done in circuit_mark_for_close(). We
should also keep a separate tally of this count for informational
purposes.
--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/7440#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