Re: pgcrypto seeding problem when ssl=on

On Sun, Dec 23, 2012 at 02:49:08PM -0500, Tom Lane wrote:
> Noah Misch <noah(at)leadboat(dot)com> writes:
> > On Sat, Dec 22, 2012 at 02:20:56PM -0500, Tom Lane wrote:
> >> #ifdef USE_SSL
> >> if (EnableSSL)
> >> {
> >> struct timeval tv;
> >>
> >> gettimeofday(&tv, NULL);
> >> RAND_add(&tv, sizeof(tv), 0);
> >> }
> >> #endif
>
> > Take the caution one step further and make it independent of EnableSSL. In a
> > stock installation, a !EnableSSL postmaster will never seed its PRNG, and
> > there's no vulnerability. Add a shared_preload_libraries module that uses the
> > OpenSSL PRNG in its _PG_init(), and suddenly you're vulnerable again.
>
> Meh. In a postmaster that wasn't built with SSL support at all, such
> a module is still dangerous (and I'm not convinced anybody would build
> such a module anyway). I think we should confine our ambitions to
> preventing security issues caused by our own code.

You're adding lines of code to prematurely micro-optimize the backend fork
cycle. If code introduced into the postmaster, by us or others, ever violates
the assumption behind that micro-optimization, certain users get a precipitous
loss of security with no clear alarm bells. I don't like that trade.

nm