[tor-bugs] #14881 [Core Tor/Tor]: incorrect defaults when producing bandwidth-weights line in directory footer
Tor Bug Tracker & Wiki
blackhole at torproject.org
Wed Apr 27 01:23:46 UTC 2016
#14881: incorrect defaults when producing bandwidth-weights line in directory
footer
-------------------------------------------------+-------------------------
Reporter: robgjansen | Owner:
Type: defect | robgjansen
Priority: Medium | Status:
Component: Core Tor/Tor | needs_revision
Severity: Normal | Milestone: Tor:
Keywords: 027-triaged-1-in, 028-triaged, | 0.2.9.x-final
pre028-patch, TorCoreTeam201604, tor- | Version: Tor:
sponsorU-orphan | 0.2.7
Parent ID: | Resolution:
Reviewer: mikeperry | Actual Points:
| Points: medium
| Sponsor:
| SponsorU-can
-------------------------------------------------+-------------------------
Changes (by mikeperry):
* status: needs_review => needs_revision
Comment:
First thoughts: Does this run without printing out additional errors?
networkstatus_check_weights() verifies that the linear algebra has a
consistent solution and will cause a warn to be emitted if not, and the
weights will still be omitted. At minimum, for full correctness I think we
need to initialize T to 4 here, if we initialize all other weights to 1.
Technically we're also introducing a slight inaccuracy by this method of
initialization, as well (since we're giving every node class a tiny bit of
weight that isn't actually there).
It may also be the case that we were within a rounding error of
correctness, and so this ran OK.
I think this might mean that it would be simplest for this change to apply
to testing networks only. Perhaps we just have a testing network check at
the top of networkstatus_compute_bw_weights_v10() that sets weights to 1
only if it is a testing network and they are zero.
Give me a second and I will attach an alternate (untested) patch just for
testing networks.
--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/14881#comment:22>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online
More information about the tor-bugs
mailing list