Re: FD_SET and FORTIFY_SOURCE

[Rich Felker <dalias at aerifal dot cx>]

On Thu, Feb 28, 2013 at 08:52:38AM +0100, Florian Weimer wrote:
> I thought we did bounds checking for FD_SET in FORTIFY_SOURCE mode.
> But it seems we don't.  Would this be reasonable to add?
> 
> (The advantages are rather minor because it's not immediately
> obvious how these out-of-bound writes could be exploited for code
> execution, considering that the index is not under full control by
> the attacker. So all that FORTIFY_SOURCE does is that it turns a
> potential crash into a definite crash.)

The problem is that a large amount of otherwise-correct software uses
malloc and out-of-bounds FD_SET arguments to work with sets larger
than FD_SETSIZE. I have written on the issue before that the "correct"
way to do this is allocate an array of fd_set objects and use / and %
to address the bit rather than passing an out-of-bounds argument to
FD_SET (which invokes UB), but some important software like libevent
(if I remember correctly) remains wrong in this regard.

So, any such change needs to consider this issue. Personally, I'm
still not opposed to it; I think it would get these broken programs to
fix the way they're doing things if they got bug reports that say "X
breaks with _FORTIFY_SOURCE because it's invoking UB". But others
might want to be more conservative.

Rich