Password theft from memory?

[RW]

On Thu, 28 Apr 2011 13:17:41 -0400
Bob Hall <rjhjr0 at gmail.com> wrote:

> On Thu, Apr 28, 2011 at 12:10:10AM +0100, RW wrote:
> > I'm not saying that anonymous mappings used by malloc aren't
> > zero-filled, just that it's not mentioned anywhere in the mmap man
> > page. I think it's just taken as read.
> 
> I just got what you're trying to say. Unfortunately, your quotes
> mislead me about what you were concerned about. You're right, the man
> page doesn't explicitly state whether anonymous mappings are zero
> filled or not. Since man pages prioritize concision, I would expect
> the page to explain how anonymous mappings are different from other
> mappings, but not how they are the same.

but they aren't the same - that's what the quotes were about. 

If the zero-filled extensions included whole pages that have the same
status as the rest of the mapping then it would be reasonable to infer
that anonymous mappings are similarly filled with zeroed-pages.

What that man page is talking about is a small padding region at the
end of a file-backed mapping that isn't really intended for use and
presumably could get re-zeroed at any moment since it has no backing
store. It seems to me to be unreasonable to infer anything about
anonymous mappings from this.