Add rudimentary Linux exploitability analysis

This change adds very simple exploitability analysis for Linux binaries. The
only signals currently considered are whether the process exited because a stack
overflow or buffer overflow was detected.

Add exploitability tests for stack smash and stack overflow crashes on Linux.

Rename EXPLOITABLITY_MEDIUM [sic, note the missing 'I'] to
EXPLOITABILITY_MEDIUM. Keep the former as an alias for the latter for
compatibility.

Remove some unused code from the exploitability tests and refactor the rest to
be shorter and easier to read.

[mattdr]

Hi folks. Could someone take a look at this review? Thanks.

[mattdr]

Ping.


On Thu, Sep 12, 2013 at 3:38 PM, <mattdr@google.com> wrote:

> Hi folks. Could someone take a look at this review? Thanks.
>
> https://breakpad.appspot.com/**622002/<https://breakpad.appspot.com/622002/>
>

[Cris Neckar]

I'm looking now. Sorry I almost never look at the breakpad codereview site :)

[Cris Neckar]

LGTM with nit

https://breakpad.appspot.com/622002/diff/10001/src/processor/exploitability_l...
File src/processor/exploitability_linux.cc (right):

https://breakpad.appspot.com/622002/diff/10001/src/processor/exploitability_l...
src/processor/exploitability_linux.cc:73: return EXPLOITABILITY_MEDIUM;
I might even argue for High here as these both seem like pretty clearly bad
scenarios. I was mostly using medium for things that were bad in the wrong
context but where context couldn't reliably be determined. In both of these
cases the crash seems like it would be very likely to be user controllable.