[tor-bugs] #3122 [Tor Client]: Write and use constant-time comparison functions

[Tor Bug Tracker & Wiki]

#3122: Write and use constant-time comparison functions
-------------------------+--------------------------------------------------
 Reporter:  rransom      |          Owner:  ioerror           
     Type:  enhancement  |         Status:  new               
 Priority:  major        |      Milestone:  Tor: 0.2.1.x-final
Component:  Tor Client   |        Version:                    
 Keywords:               |         Parent:                    
   Points:               |   Actualpoints:                    
-------------------------+--------------------------------------------------

Comment(by rransom):

 Replying to [comment:18 nickm]:

 > Suggestive datum: The reported T values are high even for the function
 do_nothing_at_all, which doesn't actually do anything but add all the
 bytes in its inputs and return the results.  This suggests a bug in the
 harness or in the analysis.

 The harness is broken.  The purpose of using a processor's cycle counter
 to time a function is to get a useful time value from a ''single''
 execution of the function.  If you repeat a function hundreds of times,
 you dramatically increase the likelihood that the OS will interrupt the
 process, thereby ruining your timings.

 (That ruination is much worse if you use a cycle counter for timing on a
 multi-processor system -- since only one of the processors started when
 the computer was turned on, and the rest were activated later as the OS
 kernel initialized its hardware drivers, the cycle counters for different
 processors (even if they live on the same chip) will be skewed by ''many''
 cycles.  You can see this effect on some of DJB's SUPERCOP graphs.)

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/3122#comment:19>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online