Re: [PATCH 2/4] KEYS: trusted: allow trust sources to use kernel RNG for key material
From: Sumit Garg
Date: Tue Aug 10 2021 - 01:25:05 EST
On Mon, 9 Aug 2021 at 15:26, Jarkko Sakkinen <jarkko@xxxxxxxxxx> wrote:
>
> On Mon, Aug 09, 2021 at 09:52:20AM +0200, Ahmad Fatoum wrote:
> > Hello Sumit,
> >
> > On 22.07.21 08:31, Sumit Garg wrote:
> > > On Wed, 21 Jul 2021 at 22:19, Ahmad Fatoum <a.fatoum@xxxxxxxxxxxxxx> wrote:
> > >>
> > >> The two existing trusted key sources don't make use of the kernel RNG,
> > >> but instead let the hardware that does the sealing/unsealing also
> > >> generate the random key material. While a previous change offers users
> > >> the choice to use the kernel RNG instead for both, new trust sources
> > >> may want to unconditionally use the kernel RNG for generating key
> > >> material, like it's done elsewhere in the kernel.
> > >>
> > >> This is especially prudent for hardware that has proven-in-production
> > >> HWRNG drivers implemented, as otherwise code would have to be duplicated
> > >> only to arrive at a possibly worse result.
> > >>
> > >> Make this possible by turning struct trusted_key_ops::get_random
> > >> into an optional member. If a driver leaves it NULL, kernel RNG
> > >> will be used instead.
> > >>
> > >> Signed-off-by: Ahmad Fatoum <a.fatoum@xxxxxxxxxxxxxx>
> > >> ---
> > >> To: James Bottomley <jejb@xxxxxxxxxxxxx>
> > >> To: Jarkko Sakkinen <jarkko@xxxxxxxxxx>
> > >> To: Mimi Zohar <zohar@xxxxxxxxxxxxx>
> > >> To: David Howells <dhowells@xxxxxxxxxx>
> > >> Cc: James Morris <jmorris@xxxxxxxxx>
> > >> Cc: "Serge E. Hallyn" <serge@xxxxxxxxxx>
> > >> Cc: "Horia Geantă" <horia.geanta@xxxxxxx>
> > >> Cc: Aymen Sghaier <aymen.sghaier@xxxxxxx>
> > >> Cc: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx>
> > >> Cc: "David S. Miller" <davem@xxxxxxxxxxxxx>
> > >> Cc: Udit Agarwal <udit.agarwal@xxxxxxx>
> > >> Cc: Eric Biggers <ebiggers@xxxxxxxxxx>
> > >> Cc: Jan Luebbe <j.luebbe@xxxxxxxxxxxxxx>
> > >> Cc: David Gstir <david@xxxxxxxxxxxxx>
> > >> Cc: Richard Weinberger <richard@xxxxxx>
> > >> Cc: Franck LENORMAND <franck.lenormand@xxxxxxx>
> > >> Cc: Sumit Garg <sumit.garg@xxxxxxxxxx>
> > >> Cc: keyrings@xxxxxxxxxxxxxxx
> > >> Cc: linux-crypto@xxxxxxxxxxxxxxx
> > >> Cc: linux-integrity@xxxxxxxxxxxxxxx
> > >> Cc: linux-kernel@xxxxxxxxxxxxxxx
> > >> Cc: linux-security-module@xxxxxxxxxxxxxxx
> > >> ---
> > >> include/keys/trusted-type.h | 2 +-
> > >> security/keys/trusted-keys/trusted_core.c | 2 +-
> > >> 2 files changed, 2 insertions(+), 2 deletions(-)
> > >>
> > >> diff --git a/include/keys/trusted-type.h b/include/keys/trusted-type.h
> > >> index d89fa2579ac0..4eb64548a74f 100644
> > >> --- a/include/keys/trusted-type.h
> > >> +++ b/include/keys/trusted-type.h
> > >> @@ -64,7 +64,7 @@ struct trusted_key_ops {
> > >> /* Unseal a key. */
> > >> int (*unseal)(struct trusted_key_payload *p, char *datablob);
> > >>
> > >> - /* Get a randomized key. */
> > >> + /* Optional: Get a randomized key. */
> > >> int (*get_random)(unsigned char *key, size_t key_len);
> > >>
> > >> /* Exit key interface. */
> > >> diff --git a/security/keys/trusted-keys/trusted_core.c b/security/keys/trusted-keys/trusted_core.c
> > >> index 569af9af8df0..d2b7626cde8b 100644
> > >> --- a/security/keys/trusted-keys/trusted_core.c
> > >> +++ b/security/keys/trusted-keys/trusted_core.c
> > >> @@ -334,7 +334,7 @@ static int __init init_trusted(void)
> > >> continue;
> > >>
> > >> get_random = trusted_key_sources[i].ops->get_random;
> > >> - if (trusted_kernel_rng)
> > >> + if (trusted_kernel_rng || !get_random)
> > >> get_random = kernel_get_random;
> > >>
> > >
> > > For ease of understanding, I would prefer to write it as:
> > >
> > > get_random = trusted_key_sources[i].ops->get_random ?:
> > > kernel_get_random;
> > > if (trusted_kernel_rng)
> > > get_random = kernel_get_random;
> > >
> > > With that:
> > >
> > > Acked-by: Sumit Garg <sumit.garg@xxxxxxxxxx>
> >
> > I don't think it improves readability to split up the conditional.
> > At least I need to take a second pass over the code to understand
> > the second conditional.
>
> Ternary operators are pain to read, unless a super trivial case.
>
> I'd stick to what you did.
Fair enough, I am fine with the current patch.
-Sumit
>
> /Jarkko