[tor-bugs] #33290 [Core Tor/Tor]: Add diagnostics for confusing corruption issue #32564 in ewma

[Tor Bug Tracker & Wiki]

#33290: Add diagnostics for confusing corruption issue #32564 in ewma
----------------------------------+------------------------------------
 Reporter:  nickm                 |          Owner:  nickm
     Type:  defect                |         Status:  needs_revision
 Priority:  Medium                |      Milestone:  Tor: 0.4.3.x-final
Component:  Core Tor/Tor          |        Version:  Tor: 0.4.1.6
 Severity:  Normal                |     Resolution:
 Keywords:  043-should backport?  |  Actual Points:  .1
Parent ID:  #32564                |         Points:
 Reviewer:  dgoulet               |        Sponsor:
----------------------------------+------------------------------------
Changes (by dgoulet):

 * status:  needs_review => needs_revision
 * reviewer:   => dgoulet


Comment:

 Replying to [comment:2 cypherpunks]:
 > Replying to [comment:1 nickm]:
 > > Branches are ticket33290_{041,042,043,master}.  I am recommending that
 we try this out in 043 first, so here is an 043 PR:
 https://github.com/torproject/tor/pull/1729 .
 >
 > Is there a way to stop the compiler from optimizing away the dead store
 to `pol->base_.magic = 0xDEAD901C;` right before it gets freed?

 So yeah I think that is a very good point. And gcc uses `-O2` which will
 likely optimize it out. I did a quick test case here and it does optimize
 it out. And also looked at tor disassembled and it is not there in those
 free function.

 Maybe we could set the `magic` to `volatile` which in "theory", but not
 enforced, should leave the variable alive on most compilers.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/33290#comment:3>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online