Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220926213130.1508261-1-Jason@zx2c4.com> <YzKZnkwCi0UwY/4Q@owl.dominikbrodowski.net>
In-Reply-To: <YzKZnkwCi0UwY/4Q@owl.dominikbrodowski.net>
From:   "Jason A. Donenfeld" <Jason@zx2c4.com>
Date:   Tue, 27 Sep 2022 10:28:11 +0200
Message-ID: <CAHmME9oGkjAxvoBvWMBRSjFmKLzOdzfcQAB4q3P869BsySSfNg@mail.gmail.com>
Subject: Re: [PATCH v2 1/2] random: split initialization into early step and
 later step
To:     linux@dominikbrodowski.net
Cc:     linux-kernel@vger.kernel.org, linux-mm@kvack.org,
        kasan-dev@googlegroups.com, Kees Cook <keescook@chromium.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        stable@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, Sep 27, 2022 at 8:35 AM Dominik Brodowski
<linux@dominikbrodowski.net> wrote:
> >  #if defined(LATENT_ENTROPY_PLUGIN)
> >       static const u8 compiletime_seed[BLAKE2S_BLOCK_SIZE] __initconst __latent_entropy;
> > @@ -803,34 +798,46 @@ int __init random_init(const char *command_line)
> >                       i += longs;
> >                       continue;
> >               }
> > -             entropy[0] = random_get_entropy();
> > -             _mix_pool_bytes(entropy, sizeof(*entropy));
> >               arch_bits -= sizeof(*entropy) * 8;
> >               ++i;
> >       }
>
>
> Previously, random_get_entropy() was mixed into the pool ARRAY_SIZE(entropy)
> times.
>
> > +/*
> > + * This is called a little bit after the prior function, and now there is
> > + * access to timestamps counters. Interrupts are not yet enabled.
> > + */
> > +void __init random_init(void)
> > +{
> > +     unsigned long entropy = random_get_entropy();
> > +     ktime_t now = ktime_get_real();
> > +
> > +     _mix_pool_bytes(utsname(), sizeof(*(utsname())));
>
> But now, it's only mixed into the pool once. Is this change on purpose?

Yea, it is. I don't think it's really doing much of use. Before we did
it because it was convenient -- because we simply could. But in
reality mostly what we care about is capturing when it gets to that
point in the execution. For jitter, the actual jitter function
(try_to_generate_entropy()) is better here.

However, before feeling too sad about it, remember that
extract_entropy() is still filling a block with rdtsc when rdrand
fails, the same way as this function was. So it's still in there
anyway.

Jason