[tor-bugs] #23061 [Core Tor/Tor]: crypto_rand_double() should produce all possible outputs on platforms with 32-bit int

[Tor Bug Tracker & Wiki]

#23061: crypto_rand_double() should produce all possible outputs on platforms with
32-bit int
-------------------------------------------------+-------------------------
 Reporter:  teor                                 |          Owner:  nickm
     Type:  defect                               |         Status:
                                                 |  needs_revision
 Priority:  Medium                               |      Milestone:  Tor:
                                                 |  0.3.2.x-final
Component:  Core Tor/Tor                         |        Version:  Tor:
                                                 |  0.2.2.14-alpha
 Severity:  Normal                               |     Resolution:
 Keywords:  tor-relay, security-low, privcount,  |  Actual Points:  0.5
  031-backport, 030-backport, 029-backport, 028  |
  -backport-maybe, 027-backport-maybe, 026       |
  -backport-maybe                                |
Parent ID:                                       |         Points:  0.1
 Reviewer:                                       |        Sponsor:
                                                 |  SponsorQ
-------------------------------------------------+-------------------------

Comment (by teor):

 Replying to [comment:10 catalyst]:
 > Replying to [comment:9 teor]:
 > > It might be ok to only have 48 bits of entropy. Or we could extend
 that scheme to give us 52 (or >52 with subnormals).
 >
 > Would we have to do something different to make sure the subnormal
 values are uniform with the others?

 If the architecture supports subnormals, ldexp() generates them, and +
 correctly handles them, then this code will work:
 {{{
 unsigned short rseed[4];
 crypto_rand(rseed, sizeof rseed);
 return ldexp((double) rseed[0], -64) +
        ldexp((double) rseed[1], -48) +
        ldexp((double) rseed[2], -32) +
        ldexp((double) rseed[3], -16);
 }}}

 It's much better than bit-stuffing, because it doesn't depend on how
 subnormals are implemented.

 Modified from:
 ​https://github.com/libressl-
 portable/openbsd/blob/master/src/lib/libc/stdlib/drand48.c

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/23061#comment:11>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online