Re: Race between vmtruncate and mapped areas?

[Rik van Riel]

On Wed, 14 May 2003, Rik van Riel wrote:
> On Wed, 14 May 2003, Andrew Morton wrote:
> 
> > It would be nice to make them go away - they cause problems.
> 
> Not to mention they could end up being outside of any VMA,
> meaning there's no sane way to deal with them.

I hate to follow up to my own email, but the fact that
they're not in any VMA could mean we leak these pages
at exit() time.

Which means a security bug, as well as the potential to
end up with bad pointers in kernel space, eg. think about
the rmap code jumping to a no longer existing mm_struct.

The more I think about it, the more I agree with Andrew
that it would be really really nice to get rid of them ;)