[Gc] security issue with libgc ?

Hans Boehm Hans.Boehm at hp.com
Mon Mar 19 16:07:59 PST 2007

On Fri, 16 Mar 2007, Florian Weimer wrote:

> * Andrew Haley:
> > But injecting code or crashing the program?  Not unless your program
> > has some other bug that is triggered by running out of memory.
> Isn't there always some risk that a live pointer is missed due to
> legitimate compiler optimizations?  This could have security
> implications (but the risk of generally broken optimizers is probably
> higher).
I agree that this is a small concern.  But it seems to happen so rarely in
practice that it's hard to convince compiler writers to address it.
Aince the gcc back-end is also used to compile Java, it really should
not produce such code.

If it did happen, I expect it would be more likely to result in
crashes than an exploitable security hole.  I think you would need an
object used for sensitive data to be accidentally reclaimed and
reallocated to an object that could be externally read or overwritten,
without crashing the process before that.


More information about the Gc mailing list