[Cryptography] WIPEONFORK in Linux 4.14

[Florian Weimer]

* Nico Williams:

> On Mon, Nov 27, 2017 at 08:02:11PM +0100, Florian Weimer wrote:
>> * Nico Williams:
>> > pthread_atfork() suffices for fork-safety for userland PRNGs,
>> 
>> This is not true on Linux once you have code that calls fork/clone
>> system calls directly (which is traditional for namespace/container
>> setup, although it shouldn't strictly be necessary anymore).
>> pthread_atfork won't be called in these cases because it's strictly a
>> userspace thing.
>
> Right.  One should not, however, call syscall(2) to avoid libc stubs.
> It's very dangerous for precisely this sort of reason.

Well, historically, we have done a very poor job exposing clone(2)
functionality in a more ready-to-use manner.  What we provide is
geared towards creation of threads, rather than the fork/vfork-like
interface many applications want.

>> What should work as an MADV_WIPEONFORK replacement is a MAP_SHARED
>> mapping and two counters, [...]
>
> Clever.  You can also just check that (my_saved_pid == getpid()), which
> if you have a fast getpid() via a vdso, is cheap.

Unfortunately, PIDs get reused too quickly for that:

<https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=0d1ecd6300191a450978ca2fcd12bbbb7c5e65e6>

OpenSSL started to mix in time as well since then, but it's still very
hackish (and I don't blame them—the MAP_SHARED trick is unlikely to
work on all platforms they need to support, and we don't know yet if
it will withstand the test of time).