linux-kernel - Re: x86/random: Speculation to the rescue

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAHk-=wgjC01UaoV35PZvGPnrQ812SRGPoV7Xp63BBFxAsJjvrg@mail.gmail.com>
Date:   Sat, 28 Sep 2019 16:53:52 -0700
From:   Linus Torvalds <torvalds@...ux-foundation.org>
To:     Thomas Gleixner <tglx@...utronix.de>,
        "Ahmed S. Darwish" <darwish.07@...il.com>
Cc:     LKML <linux-kernel@...r.kernel.org>,
        "Theodore Ts'o" <tytso@....edu>,
        Nicholas Mc Guire <hofrat@...ntech.at>,
        "the arch/x86 maintainers" <x86@...nel.org>,
        Andy Lutomirski <luto@...nel.org>,
        Kees Cook <keescook@...omium.org>
Subject: Re: x86/random: Speculation to the rescue

On Sat, Sep 28, 2019 at 3:24 PM Thomas Gleixner <tglx@...utronix.de> wrote:
>
> Nicholas presented the idea to (ab)use speculative execution for random
> number generation years ago at the Real-Time Linux Workshop:

What you describe is just a particularly simple version of the jitter
entropy. Not very reliable.

But hey, here's a made-up patch. It basically does jitter entropy, but
it uses a more complex load than the fibonacci LFSR folding: it calls
"schedule()" in a loop, and it sets up a timer to fire.

And then it mixes in the TSC in that loop.

And to be fairly conservative, it then credits one bit of entropy for
every timer tick. Not because the timer itself would be all that
unpredictable, but because the interaction between the timer and the
loop is going to be pretty damn unpredictable.

Ok, I'm handwaving. But I do claim it really is fairly conservative to
think that a cycle counter would give one bit of entropy when you time
over a timer actually happening. The way that loop is written, we do
guarantee that we'll mix in the TSC value both before and after the
timer actually happened. We never look at the difference of TSC
values, because the mixing makes that uninteresting, but the code does
start out with verifying that "yes, the TSC really is changing rapidly
enough to be meaningful".

So if we want to do jitter entropy, I'd much rather do something like
this that actually has a known fairly complex load with timers and
scheduling.

And even if absolutely no actual other process is running, the timer
itself is still going to cause perturbations. And the "schedule()"
call is more complicated than the LFSR is anyway.

It does wait for one second the old way before it starts doing this.

Whatever. I'm entirely convinced this won't make everybody happy
anyway, but it's _one_ approach to handle the issue.

Ahmed - would you be willing to test this on your problem case (with
the ext4 optimization re-enabled, of course)?

And Thomas - mind double-checking that I didn't do anything
questionable with the timer code..

And this goes without saying - this patch is ENTIRELY untested.  Apart
from making people upset for the lack of rigor, it might do
unspeakable crimes against your pets. You have been warned.

               Linus

View attachment "patch.diff" of type "text/x-patch" (2412 bytes)