Received-SPF: pass (google.com: domain of linux-crypto-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 22 Nov 2022 11:58:16 +0100
From:   "Jason A. Donenfeld" <Jason@zx2c4.com>
To:     Jiri Kosina <jikos@kernel.org>
Cc:     herbert@gondor.apana.org.au, linux-crypto@vger.kernel.org,
        Andrej Shadura <andrew.shadura@collabora.co.uk>
Subject: Re: [PATCH] hwrng: u2fzero - account for high quality RNG
Message-ID: <Y3yrSFel+sK5Fvqx@zx2c4.com>
References: <20221119134259.2969204-1-Jason@zx2c4.com>
 <nycvar.YFH.7.76.2211221122220.6045@cbobk.fhfr.pm>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <nycvar.YFH.7.76.2211221122220.6045@cbobk.fhfr.pm>
Precedence: bulk

On Tue, Nov 22, 2022 at 11:22:37AM +0100, Jiri Kosina wrote:
> On Sat, 19 Nov 2022, Jason A. Donenfeld wrote:
> 
> > The U2F zero apparently has a real TRNG in it with maximum quality, not
> > one with quality of "1", which was likely a misinterpretation of the
> > field as a boolean. So remove the assignment entirely, so that we get
> > the default quality setting.
> > 
> > In the u2f-zero firmware, the 0x21 RNG command used by this driver is
> > handled as such [1]:
> > 
> >   case U2F_CUSTOM_GET_RNG:
> >     if (atecc_send_recv(ATECC_CMD_RNG,ATECC_RNG_P1,ATECC_RNG_P2,
> >       NULL, 0,
> >       appdata.tmp,
> >       sizeof(appdata.tmp), &res) == 0 )
> >     {
> >       memmove(msg->pkt.init.payload, res.buf, 32);
> >       U2FHID_SET_LEN(msg, 32);
> >       usb_write((uint8_t*)msg, 64);
> >     }
> >     else
> >     {
> >       U2FHID_SET_LEN(msg, 0);
> >       usb_write((uint8_t*)msg, 64);
> >     }
> > 
> > This same call to `atecc_send_recv(ATECC_CMD_RNG,ATECC_RNG_P1,
> > ATECC_RNG_P2,...)` is then also used in the token's cryptographically
> > critical "u2f_new_keypair" function, as its rather straightforward
> > source of random bytes [2]:
> > 
> >   int8_t u2f_new_keypair(uint8_t * handle, uint8_t * appid, uint8_t * pubkey)
> >   {
> >     struct atecc_response res;
> >     uint8_t private_key[36];
> >     int i;
> > 
> >     watchdog();
> > 
> >     if (atecc_send_recv(ATECC_CMD_RNG,ATECC_RNG_P1,ATECC_RNG_P2,
> >       NULL, 0,
> >       appdata.tmp,
> >       sizeof(appdata.tmp), &res) != 0 )
> >     {
> >       return -1;
> >     }
> > 
> > So it seems rather plain that the ATECC RNG is considered to provide
> > good random numbers.
> > 
> > [1] https://github.com/conorpp/u2f-zero/blob/master/firmware/src/custom.c
> > [2] https://github.com/conorpp/u2f-zero/blob/master/firmware/src/u2f_atecc.c
> 
> Good catch, thanks Jason. Applied.

This should probably go via Herbert's tree, because it depends on some
changed handling for the zero quality field.

Jason