[tor-bugs] #5786 [Tor Directory Authority]: A single authority can crash the other authorities

[Tor Bug Tracker & Wiki]

#5786: A single authority can crash the other authorities
-------------------------------------+--------------------------------------
 Reporter:  asn                      |          Owner:                    
     Type:  defect                   |         Status:  needs_review      
 Priority:  normal                   |      Milestone:  Tor: 0.2.3.x-final
Component:  Tor Directory Authority  |        Version:                    
 Keywords:                           |         Parent:                    
   Points:                           |   Actualpoints:                    
-------------------------------------+--------------------------------------

Comment(by mikeperry):

 Oh man. Rather embarrassing mistake to use INT32_MIN instead of a positive
 integer. Not sure if 1 is the best choice, but in addition to the assert,
 the v10+ consensus methods will also not include bandwidth-weight lines
 that do not actually solve the system of linear equations and/or if they
 somehow solve them with values outside of the 0-1.0 (scaled) range for any
 weight. See the use of networkstatus_check_weights in
 networkstatus_compute_bw_weights_v10 for those checks.

 Those checks should also prevent badness in the event of authorities
 choosing a ridiculously low bwweightscale, like 1 or 2 (even 10 is
 probably too low in practice).

 But overall, yes, the combination of bug5786_range_022 and your
 weight_scale commit should cause the code to function as intended (which
 is to fall back to BW_WEIGHT_SCALE for super crazy values of
 bwweightscale).

 I do not claim to understand rounterparse.c well enough to properly review
 your "prevent dirauths from voting more than once for the same parameter"
 commit, though.

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/5786#comment:5>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online