[tor-bugs] #19969 [Core Tor/Tor]: tor client does not immediately open new circuits after standby

[Tor Bug Tracker & Wiki]

#19969: tor client does not immediately open new circuits after standby
-------------------------------------------------+-------------------------
 Reporter:  weasel                               |          Owner:
     Type:  defect                               |         Status:
                                                 |  needs_review
 Priority:  High                                 |      Milestone:  Tor:
                                                 |  0.2.9.x-final
Component:  Core Tor/Tor                         |        Version:  Tor:
                                                 |  0.2.8.6
 Severity:  Normal                               |     Resolution:
 Keywords:  regression 029-backport              |  Actual Points:
  028-backport                                   |
Parent ID:                                       |         Points:  2
 Reviewer:                                       |        Sponsor:
-------------------------------------------------+-------------------------

Comment (by arma):

 Replying to [comment:24 nickm]:
 > > My {{{bug19969_028_alt}}} branch implements this option. I found it
 useful to write to make sure I understood what was going on, so I will be
 totally fine if you pick your branch as being the less messy / more safe
 one. :)
 >
 > That's actually a little bad: calling event_base_loopexit() over and
 over like that is not idempotent.  Instead, it creates "exit the loop"
 events proportional to the number of times you call it.  That's not
 ''super'' harmful, but it does mean allocate more RAM and waste more CPU
 than one might want.

 Ok, I have added a new commit to {{{bug19969_028_alt}}} that tries to
 avoid calling event_base_loopexit() when we shouldn't.

 Also, I made a {{{bug19969_029_alt}}} branch that cherry-picks these three
 commits onto maint-0.2.9, since we are considering not putting this
 version of the patch into 0.2.8.

 This is getting messier as I look at it more though. Now when I do a socks
 request while my Tor doesn't have enough dir info, I actually get k
 "optimistically trying directory fetches again" lines, where k is the
 number of streams that are pending each time a new pending one gets added.
 That's surprising, but I haven't sorted out whether I think it's harmful
 -- maybe not, because each of them just resets directory fetching stuff,
 or maybe so, because if they reset the directory fetching stuff, they will
 clobber any new efforts we made to mark down fallback dirs as actually
 down? Hm. There is a little part of me that is tempted to try to back out
 commits b1d56fc589, 84b3350c83, and 58edf926 and call that the 028
 solution. But they don't revert cleanly, and I haven't explored further.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/19969#comment:27>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online