Received-SPF: softfail (google.com: domain of transitioning linux-kernel-owner@vger.kernel.org does not designate 23.128.96.19 as permitted sender) client-ip=23.128.96.19;
MIME-Version: 1.0
References: <20220607111514.755009-1-Jason@zx2c4.com> <CAMj1kXFDYX3fAdO6hxH9DTFP7+LNYz0fL9Dy8eKsH_xGwXxatQ@mail.gmail.com>
 <Yp8+6Y+bEcmR1LS0@zx2c4.com> <CAMj1kXFPxP3y6Ma1AonMSiW5DP0veB+NAj+ggfvrWvAARsOmgA@mail.gmail.com>
 <Yp9C4xY0+CguTd7l@zx2c4.com> <CAMj1kXH4ifsytLNpaVTHAumadMjj8rNEZbsn-8t=hH51ucG11A@mail.gmail.com>
 <Yp9lLkYXCo9BZxpk@zx2c4.com>
In-Reply-To: <Yp9lLkYXCo9BZxpk@zx2c4.com>
From:   Ard Biesheuvel <ardb@kernel.org>
Date:   Tue, 7 Jun 2022 16:51:43 +0200
Message-ID: <CAMj1kXGY7tPOazLZ0VKagPDcmqURO2Y4MytPMmVjRTz=Ho=2jg@mail.gmail.com>
Subject: Re: [PATCH] random: defer use of bootloader randomness to random_init()
To:     "Jason A. Donenfeld" <Jason@zx2c4.com>
Cc:     Catalin Marinas <catalin.marinas@arm.com>,
        linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org,
        Stephen Boyd <swboyd@chromium.org>,
        Russell King <linux@armlinux.org.uk>,
        Arnd Bergmann <arnd@arndb.de>,
        Phil Elwell <phil@raspberrypi.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, 7 Jun 2022 at 16:48, Jason A. Donenfeld <Jason@zx2c4.com> wrote:
>
> Hi Ard,
>
> On Tue, Jun 07, 2022 at 04:19:26PM +0200, Ard Biesheuvel wrote:
> > #include <stdio.h>
> > #include <stdlib.h>
> > #include <sys/random.h>
> >
> > static unsigned char buf[16];
> >
> > int main(void)
> > {
> >   for (int i = 0; i < 1000000; i++) {
> >     if (getrandom(buf, sizeof(buf),
> >         GRND_RANDOM | GRND_NONBLOCK) < sizeof(buf)) {
> >           fprintf(stderr, "getrandom() error!\n");
> >           exit(-1);
> >     }
> >   }
> >   return 0;
> > }
>
> I'm actually more worried about the random input flow than the random
> output flow and branch misprediction. But more generally, I'd just like
> to keep that code as cold as possible after crng init. It's code that's
> only used in that one phase and then never again. It can be entirely
> disabled.
>
> Anyway, we've got a few solutions now to pick from on the random.c side
> of things. I'm going to investigate the arm32 situation next. And then
> we'll see what it all looks like.
>

Sure.

It would be helpful if some other folks could chime in as well?