[tor-bugs] #29698 [Core Tor/Tor]: Edge case that causes improper circuit prioritization for one scheduling run

[Tor Bug Tracker & Wiki]

#29698: Edge case that causes improper circuit prioritization for one scheduling
run
-------------------------------------------------+-------------------------
 Reporter:  pastly                               |          Owner:  dgoulet
     Type:  defect                               |         Status:
                                                 |  needs_revision
 Priority:  Medium                               |      Milestone:  Tor:
                                                 |  0.4.2.x-final
Component:  Core Tor/Tor                         |        Version:  Tor:
                                                 |  0.3.2.1-alpha
 Severity:  Normal                               |     Resolution:
 Keywords:  tor-sched, tor-cmux, kist,           |  Actual Points:  0.2
  041-deferred-20190530, 042-should,             |
  BugSmashFund                                   |
Parent ID:                                       |         Points:  0.2
 Reviewer:  nickm                                |        Sponsor:
-------------------------------------------------+-------------------------

Comment (by dgoulet):

 Replying to [comment:8 nickm]:
 > Looks reasonable; I've left some comments on the patch.
 >
 > One more thing I'd request: Could we please have a regression test on
 this?  That is, a test that fails without this patch, and passes with it,
 to demonstrate that we really fixed the problem.

 I've spent a bit of time trying to come up with a way to test this. And I
 sorta fail in a reasonable time frame. It would hitting so many code path
 in the cmux/ewma/channel that it would require quite a bit of work to set
 that test setup.

 We have virtually _no_ EWMA or cmux unit tests so everything needs to be
 done in mocking and what not to be able to do something proper. Also, the
 cmux and EWMA code is opaque, even to the tests. Thus more work there to
 expose what we need to test.

 All in all, I can do something there but there is a potential for a much
 larger patch to support everything needed for testing. And this will turn
 the ticket into a much larger Point size.

 Still OK with this?

 The other approach I guess would be to allocate myself some cycles and add
 considerable amount of testing over the entire cmux/EWMA code. I would
 expect that to be around 2 or 3 days of work for something proper which
 would include this `touch()` interface.

 > Also, is this a backport candidate?

 Nope. It is not critical enough to justify a backport imo.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/29698#comment:9>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online