Re: pgcrypto seeding problem when ssl=on

Andres Freund <andres@2ndquadrant.com> writes:
> On 2012-12-22 14:20:56 -0500, Tom Lane wrote:
>> I believe that we'd be better off doing something in postmaster.c to
>> positively ensure that each session has a distinct seed value.

> I am entirely unconvinced that adding in a gettimeofday() provides more
> entropy than what openssl gathers internally after a
> RAND_cleanup().

No, it doesn't provide more entropy, but what it does do is avoid
draining entropy from the kernel (see my reply to Marko just now).

> gettimeofday() will yield the same result in close
> enough forks on a significant number of systems whereas openssl should
> use stuff like /dev/urandom to initialize its PRNG after a cleanup.

Well, in the first place, that doesn't matter much because the PRNG
state will still change from repeated mix-ins of gettimeofday, even if
we obtain the same reading successive times.  In the second place, if we
add it where I suggest, it would be easy to also include the child
process PID in the added entropy -- we could xor it into the tv_sec
value for instance -- thus further reducing the chance of identical
values getting added in.
        regards, tom lane