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:   Mon, 21 Feb 2022 00:29:13 +0100
From:   Jarkko Sakkinen <jarkko@kernel.org>
To:     Stefan Berger <stefanb@linux.ibm.com>
Cc:     Vitaly Chikunov <vt@altlinux.org>, keyrings@vger.kernel.org,
        David Howells <dhowells@redhat.com>,
        linux-crypto@vger.kernel.org, linux-integrity@vger.kernel.org,
        Eric Biggers <ebiggers@kernel.org>
Subject: Re: [RFC PATCH] KEYS: Double max_size to make keyctl pkey_verify work
Message-ID: <YhLOyUgSihcdoNJc@iki.fi>
References: <20220202065906.2598366-1-vt@altlinux.org>
 <7c9d973f-847e-e8bc-95fb-6c98a98a02e6@linux.ibm.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <7c9d973f-847e-e8bc-95fb-6c98a98a02e6@linux.ibm.com>
Precedence: bulk

On Wed, Feb 02, 2022 at 10:15:24PM -0500, Stefan Berger wrote:
> 
> On 2/2/22 01:59, Vitaly Chikunov wrote:
> > Rarely used `keyctl pkey_verify' can verify raw signatures, but was
> > failing, because ECDSA/EC-RDSA signature sizes are twice key sizes which
> > does not pass in/out sizes check in keyctl_pkey_params_get_2.
> > This in turn because these values cannot be distinguished by a single
> > `max_size' callback return value.
> > Also, `keyctl pkey_query` displays incorrect `max_sig_size' about these
> > algorithms.
> > 
> > Signed-off-by: Vitaly Chikunov <vt@altlinux.org>
> > ---
> >   crypto/asymmetric_keys/public_key.c | 15 +++++++++++++--
> >   1 file changed, 13 insertions(+), 2 deletions(-)
> > 
> > diff --git a/crypto/asymmetric_keys/public_key.c b/crypto/asymmetric_keys/public_key.c
> > index 4fefb219bfdc..3ffbab07ed2a 100644
> > --- a/crypto/asymmetric_keys/public_key.c
> > +++ b/crypto/asymmetric_keys/public_key.c
> > @@ -143,8 +143,19 @@ static int software_key_query(const struct kernel_pkey_params *params,
> >   	len = crypto_akcipher_maxsize(tfm);
> >   	info->key_size = len * 8;
> > -	info->max_data_size = len;
> > -	info->max_sig_size = len;
> > +	if (strcmp(alg_name, "ecrdsa") == 0 ||
> > +	    strncmp(alg_name, "ecdsa-", 6) == 0) {
> > +		/*
> > +		 * For these algos sig size is twice key size.
> > +		 * keyctl uses max_sig_size as minimum input size, and
> > +		 * max_data_size as minimum output size for a signature.
> > +		 */
> > +		info->max_data_size = len * 2;
> > +		info->max_sig_size = len * 2;
> I don't know about the data size but following my tests this is not enough
> for ECDSA signature size. In ECDSA case the r and s components of the
> signature are encode in asn.1, not 'raw'. So there are 2 bytes at the
> beginning for sequence identifier , 2 bytes asn.1 for the r component, 1
> additional 0-byte to make the r component always a positive number, then the
> r component, then 2 bytes asn.1 for the s component, 1 addition 0-byte to
> make the s component a positive number, then the s component. Phew.
> 
> info->max_sig_size = 2 + (2 + 1 + len) * 2;
> 
> so for NIST P384 it's: 2 + (2+1+48) * 2 = 104
> 
> Then it works for me as well.

Thank you for the trouble of providing this great explanation. This
reasoning should be included to the commit message for future reference.

It would be also nice to encapsulate this calculation to an inline
function.

/Jarkko