[PATCH 1/2] KEYS: asymmetric: enforce that sig algo matches key algo

[PATCH 1/2] KEYS: asymmetric: enforce that sig algo matches key algo

[Eric Biggers]

From: Eric Biggers <ebiggers@google.com>

Most callers of public_key_verify_signature(), including most indirect
callers via verify_signature() as well as pkcs7_verify_sig_chain(),
don't check that public_key_signature::pkey_algo matches
public_key::pkey_algo.  These should always match.  However, a malicious
signature could intentionally declare an unintended algorithm.  It is
essential that such signatures be rejected outright, or that the
algorithm of the *key* be used -- not the algorithm of the signature as
that would allow attackers to choose the algorithm used.

Currently, public_key_verify_signature() correctly uses the key's
algorithm when deciding which akcipher to allocate.  That's good.
However, it uses the signature's algorithm when deciding whether to do
the first step of SM2, which is incorrect.  Also, v4.19 and older
kernels used the signature's algorithm for the entire process.

Prevent such errors by making public_key_verify_signature() enforce that
the signature's algorithm matches the key's algorithm.

Also remove two checks of this done by callers, which are now redundant.

Cc: stable@vger.kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
---
 crypto/asymmetric_keys/pkcs7_verify.c    |  6 ------
 crypto/asymmetric_keys/public_key.c      | 15 +++++++++++++++
 crypto/asymmetric_keys/x509_public_key.c |  6 ------
 3 files changed, 15 insertions(+), 12 deletions(-)

diff --git a/crypto/asymmetric_keys/pkcs7_verify.c b/crypto/asymmetric_keys/pkcs7_verify.c
index 0b4d07aa88111..f94a1d1ad3a6c 100644
--- a/crypto/asymmetric_keys/pkcs7_verify.c
+++ b/crypto/asymmetric_keys/pkcs7_verify.c
@@ -174,12 +174,6 @@ static int pkcs7_find_key(struct pkcs7_message *pkcs7,
 		pr_devel("Sig %u: Found cert serial match X.509[%u]\n",
 			 sinfo->index, certix);
 
-		if (strcmp(x509->pub->pkey_algo, sinfo->sig->pkey_algo) != 0) {
-			pr_warn("Sig %u: X.509 algo and PKCS#7 sig algo don't match\n",
-				sinfo->index);
-			continue;
-		}
-
 		sinfo->signer = x509;
 		return 0;
 	}
diff --git a/crypto/asymmetric_keys/public_key.c b/crypto/asymmetric_keys/public_key.c
index 4fefb219bfdc8..aba7113d86c76 100644
--- a/crypto/asymmetric_keys/public_key.c
+++ b/crypto/asymmetric_keys/public_key.c
@@ -325,6 +325,21 @@ int public_key_verify_signature(const struct public_key *pkey,
 	BUG_ON(!sig);
 	BUG_ON(!sig->s);
 
+	/*
+	 * The signature's claimed public key algorithm *must* match the key's
+	 * actual public key algorithm.
+	 *
+	 * Small exception: ECDSA signatures don't specify the curve, but ECDSA
+	 * keys do.  So the strings can mismatch slightly in that case:
+	 * "ecdsa-nist-*" for the key, but "ecdsa" for the signature.
+	 */
+	if (!sig->pkey_algo)
+		return -EINVAL;
+	if (strcmp(pkey->pkey_algo, sig->pkey_algo) != 0 &&
+	    (strncmp(pkey->pkey_algo, "ecdsa-", 6) != 0 ||
+	     strcmp(sig->pkey_algo, "ecdsa") != 0))
+		return -EKEYREJECTED;
+
 	ret = software_key_determine_akcipher(sig->encoding,
 					      sig->hash_algo,
 					      pkey, alg_name);
diff --git a/crypto/asymmetric_keys/x509_public_key.c b/crypto/asymmetric_keys/x509_public_key.c
index fe14cae115b51..71cc1738fbfd2 100644
--- a/crypto/asymmetric_keys/x509_public_key.c
+++ b/crypto/asymmetric_keys/x509_public_key.c
@@ -128,12 +128,6 @@ int x509_check_for_self_signed(struct x509_certificate *cert)
 			goto out;
 	}
 
-	ret = -EKEYREJECTED;
-	if (strcmp(cert->pub->pkey_algo, cert->sig->pkey_algo) != 0 &&
-	    (strncmp(cert->pub->pkey_algo, "ecdsa-", 6) != 0 ||
-	     strcmp(cert->sig->pkey_algo, "ecdsa") != 0))
-		goto out;
-
 	ret = public_key_verify_signature(cert->pub, cert->sig);
 	if (ret < 0) {
 		if (ret == -ENOPKG) {
-- 
2.35.1

[PATCH 2/2] KEYS: asymmetric: properly validate hash_algo and encoding

[Eric Biggers]

From: Eric Biggers <ebiggers@google.com>

It is insecure to allow arbitrary hash algorithms and signature
encodings to be used with arbitrary signature algorithms.  Notably,
ECDSA, ECRDSA, and SM2 all sign/verify raw hash values and don't
disambiguate between different hash algorithms like RSA PKCS#1 v1.5
padding does.  Therefore, they need to be restricted to certain sets of
hash algorithms (ideally just one, but in practice small sets are used).
Additionally, the encoding is an integral part of modern signature
algorithms, and is not supposed to vary.

Therefore, tighten the checks of hash_algo and encoding done by
software_key_determine_akcipher().

Also rearrange the parameters to software_key_determine_akcipher() to
put the public_key first, as this is the most important parameter and it
often determines everything else.

Fixes: 299f561a6693 ("x509: Add support for parsing x509 certs with ECDSA keys")
Fixes: 215525639631 ("X.509: support OSCCA SM2-with-SM3 certificate verification")
Fixes: 0d7a78643f69 ("crypto: ecrdsa - add EC-RDSA (GOST 34.10) algorithm")
Cc: stable@vger.kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
---
 crypto/asymmetric_keys/public_key.c | 111 +++++++++++++++++++---------
 1 file changed, 76 insertions(+), 35 deletions(-)

diff --git a/crypto/asymmetric_keys/public_key.c b/crypto/asymmetric_keys/public_key.c
index aba7113d86c76..a603ee8afdb8d 100644
--- a/crypto/asymmetric_keys/public_key.c
+++ b/crypto/asymmetric_keys/public_key.c
@@ -60,39 +60,83 @@ static void public_key_destroy(void *payload0, void *payload3)
 }
 
 /*
- * Determine the crypto algorithm name.
+ * Given a public_key, and an encoding and hash_algo to be used for signing
+ * and/or verification with that key, determine the name of the corresponding
+ * akcipher algorithm.  Also check that encoding and hash_algo are allowed.
  */
-static
-int software_key_determine_akcipher(const char *encoding,
-				    const char *hash_algo,
-				    const struct public_key *pkey,
-				    char alg_name[CRYPTO_MAX_ALG_NAME])
+static int
+software_key_determine_akcipher(const struct public_key *pkey,
+				const char *encoding, const char *hash_algo,
+				char alg_name[CRYPTO_MAX_ALG_NAME])
 {
 	int n;
 
-	if (strcmp(encoding, "pkcs1") == 0) {
-		/* The data wangled by the RSA algorithm is typically padded
-		 * and encoded in some manner, such as EMSA-PKCS1-1_5 [RFC3447
-		 * sec 8.2].
+	if (!encoding)
+		return -EINVAL;
+
+	if (strcmp(pkey->pkey_algo, "rsa") == 0) {
+		/*
+		 * RSA signatures usually use EMSA-PKCS1-1_5 [RFC3447 sec 8.2].
+		 */
+		if (strcmp(encoding, "pkcs1") == 0) {
+			if (!hash_algo)
+				n = snprintf(alg_name, CRYPTO_MAX_ALG_NAME,
+					     "pkcs1pad(%s)",
+					     pkey->pkey_algo);
+			else
+				n = snprintf(alg_name, CRYPTO_MAX_ALG_NAME,
+					     "pkcs1pad(%s,%s)",
+					     pkey->pkey_algo, hash_algo);
+			return n >= CRYPTO_MAX_ALG_NAME ? -EINVAL : 0;
+		}
+		if (strcmp(encoding, "raw") != 0)
+			return -EINVAL;
+		/*
+		 * Raw RSA cannot differentiate between different hash
+		 * algorithms.
+		 */
+		if (hash_algo)
+			return -EINVAL;
+	} else if (strncmp(pkey->pkey_algo, "ecdsa", 5) == 0) {
+		if (strcmp(encoding, "x962") != 0)
+			return -EINVAL;
+		/*
+		 * ECDSA signatures are taken over a raw hash, so they don't
+		 * differentiate between different hash algorithms.  That means
+		 * that the verifier should hard-code a specific hash algorithm.
+		 * Unfortunately, in practice ECDSA is used with multiple SHAs,
+		 * so we have to allow all of them and not just one.
 		 */
 		if (!hash_algo)
-			n = snprintf(alg_name, CRYPTO_MAX_ALG_NAME,
-				     "pkcs1pad(%s)",
-				     pkey->pkey_algo);
-		else
-			n = snprintf(alg_name, CRYPTO_MAX_ALG_NAME,
-				     "pkcs1pad(%s,%s)",
-				     pkey->pkey_algo, hash_algo);
-		return n >= CRYPTO_MAX_ALG_NAME ? -EINVAL : 0;
-	}
-
-	if (strcmp(encoding, "raw") == 0 ||
-	    strcmp(encoding, "x962") == 0) {
-		strcpy(alg_name, pkey->pkey_algo);
-		return 0;
+			return -EINVAL;
+		if (strcmp(hash_algo, "sha1") != 0 &&
+		    strcmp(hash_algo, "sha224") != 0 &&
+		    strcmp(hash_algo, "sha256") != 0 &&
+		    strcmp(hash_algo, "sha384") != 0 &&
+		    strcmp(hash_algo, "sha512") != 0)
+			return -EINVAL;
+	} else if (strcmp(pkey->pkey_algo, "sm2") == 0) {
+		if (strcmp(encoding, "raw") != 0)
+			return -EINVAL;
+		if (!hash_algo)
+			return -EINVAL;
+		if (strcmp(hash_algo, "sm3") != 0)
+			return -EINVAL;
+	} else if (strcmp(pkey->pkey_algo, "ecrdsa") == 0) {
+		if (strcmp(encoding, "raw") != 0)
+			return -EINVAL;
+		if (!hash_algo)
+			return -EINVAL;
+		if (strcmp(hash_algo, "streebog256") != 0 &&
+		    strcmp(hash_algo, "streebog512") != 0)
+			return -EINVAL;
+	} else {
+		/* Unknown public key algorithm */
+		return -ENOPKG;
 	}
-
-	return -ENOPKG;
+	if (strscpy(alg_name, pkey->pkey_algo, CRYPTO_MAX_ALG_NAME) < 0)
+		return -EINVAL;
+	return 0;
 }
 
 static u8 *pkey_pack_u32(u8 *dst, u32 val)
@@ -113,9 +157,8 @@ static int software_key_query(const struct kernel_pkey_params *params,
 	u8 *key, *ptr;
 	int ret, len;
 
-	ret = software_key_determine_akcipher(params->encoding,
-					      params->hash_algo,
-					      pkey, alg_name);
+	ret = software_key_determine_akcipher(pkey, params->encoding,
+					      params->hash_algo, alg_name);
 	if (ret < 0)
 		return ret;
 
@@ -179,9 +222,8 @@ static int software_key_eds_op(struct kernel_pkey_params *params,
 
 	pr_devel("==>%s()\n", __func__);
 
-	ret = software_key_determine_akcipher(params->encoding,
-					      params->hash_algo,
-					      pkey, alg_name);
+	ret = software_key_determine_akcipher(pkey, params->encoding,
+					      params->hash_algo, alg_name);
 	if (ret < 0)
 		return ret;
 
@@ -340,9 +382,8 @@ int public_key_verify_signature(const struct public_key *pkey,
 	     strcmp(sig->pkey_algo, "ecdsa") != 0))
 		return -EKEYREJECTED;
 
-	ret = software_key_determine_akcipher(sig->encoding,
-					      sig->hash_algo,
-					      pkey, alg_name);
+	ret = software_key_determine_akcipher(pkey, sig->encoding,
+					      sig->hash_algo, alg_name);
 	if (ret < 0)
 		return ret;
 
-- 
2.35.1

Re: [PATCH 1/2] KEYS: asymmetric: enforce that sig algo matches key algo

[Vitaly Chikunov]

Eric,

On Mon, Jan 31, 2022 at 04:34:13PM -0800, Eric Biggers wrote:
> From: Eric Biggers <ebiggers@google.com>
> 
> Most callers of public_key_verify_signature(), including most indirect
> callers via verify_signature() as well as pkcs7_verify_sig_chain(),
> don't check that public_key_signature::pkey_algo matches
> public_key::pkey_algo.  These should always match.

Why should they match?

public_key_signature is the data prepared to verify the cert's
signature. The cert's signature algorithm could be different from the
public key algorithm defined in the cert itself. They should match only
for self-signed certs. For example, you should be able to sign RSA
public key with ECDSA signature and vice versa. Or 256-bit EC-RDSA with
512-bit EC-RDSA. This check will prevent this.


> However, a malicious
> signature could intentionally declare an unintended algorithm.  It is
> essential that such signatures be rejected outright, or that the
> algorithm of the *key* be used -- not the algorithm of the signature as
> that would allow attackers to choose the algorithm used.
> 
> Currently, public_key_verify_signature() correctly uses the key's
> algorithm when deciding which akcipher to allocate.  That's good.
> However, it uses the signature's algorithm when deciding whether to do
> the first step of SM2, which is incorrect.  Also, v4.19 and older
> kernels used the signature's algorithm for the entire process.
> 
> Prevent such errors by making public_key_verify_signature() enforce that
> the signature's algorithm matches the key's algorithm.
> 
> Also remove two checks of this done by callers, which are now redundant.
> 
> Cc: stable@vger.kernel.org
> Signed-off-by: Eric Biggers <ebiggers@google.com>
> ---
>  crypto/asymmetric_keys/pkcs7_verify.c    |  6 ------
>  crypto/asymmetric_keys/public_key.c      | 15 +++++++++++++++
>  crypto/asymmetric_keys/x509_public_key.c |  6 ------
>  3 files changed, 15 insertions(+), 12 deletions(-)
> 
> diff --git a/crypto/asymmetric_keys/pkcs7_verify.c b/crypto/asymmetric_keys/pkcs7_verify.c
> index 0b4d07aa88111..f94a1d1ad3a6c 100644
> --- a/crypto/asymmetric_keys/pkcs7_verify.c
> +++ b/crypto/asymmetric_keys/pkcs7_verify.c
> @@ -174,12 +174,6 @@ static int pkcs7_find_key(struct pkcs7_message *pkcs7,
>  		pr_devel("Sig %u: Found cert serial match X.509[%u]\n",
>  			 sinfo->index, certix);
>  
> -		if (strcmp(x509->pub->pkey_algo, sinfo->sig->pkey_algo) != 0) {
> -			pr_warn("Sig %u: X.509 algo and PKCS#7 sig algo don't match\n",
> -				sinfo->index);
> -			continue;
> -		}
> -
>  		sinfo->signer = x509;
>  		return 0;
>  	}
> diff --git a/crypto/asymmetric_keys/public_key.c b/crypto/asymmetric_keys/public_key.c
> index 4fefb219bfdc8..aba7113d86c76 100644
> --- a/crypto/asymmetric_keys/public_key.c
> +++ b/crypto/asymmetric_keys/public_key.c
> @@ -325,6 +325,21 @@ int public_key_verify_signature(const struct public_key *pkey,
>  	BUG_ON(!sig);
>  	BUG_ON(!sig->s);
>  
> +	/*
> +	 * The signature's claimed public key algorithm *must* match the key's
> +	 * actual public key algorithm.
> +	 *
> +	 * Small exception: ECDSA signatures don't specify the curve, but ECDSA
> +	 * keys do.  So the strings can mismatch slightly in that case:
> +	 * "ecdsa-nist-*" for the key, but "ecdsa" for the signature.
> +	 */
> +	if (!sig->pkey_algo)
> +		return -EINVAL;

This seem incorrect too, as sig->pkey_algo could be NULL for direct
signature verification calls. For example, for keyctl pkey_verify.

(Side note: keyctl pkey_verify will not work for non-RSA signatures,
though, due to other bug - because signature size is twice key size for
them, but akcipher_alg::max_size cannot distinguish this, and
max_data_size, key_size, and max_sig_size are set from it).

> +	if (strcmp(pkey->pkey_algo, sig->pkey_algo) != 0 &&
> +	    (strncmp(pkey->pkey_algo, "ecdsa-", 6) != 0 ||
> +	     strcmp(sig->pkey_algo, "ecdsa") != 0))

This seems to be taken from x509_check_for_self_signed, that's why this
should not work for non-self-signed certs.

Thanks,

> +		return -EKEYREJECTED;
> +
>  	ret = software_key_determine_akcipher(sig->encoding,
>  					      sig->hash_algo,
>  					      pkey, alg_name);
> diff --git a/crypto/asymmetric_keys/x509_public_key.c b/crypto/asymmetric_keys/x509_public_key.c
> index fe14cae115b51..71cc1738fbfd2 100644
> --- a/crypto/asymmetric_keys/x509_public_key.c
> +++ b/crypto/asymmetric_keys/x509_public_key.c
> @@ -128,12 +128,6 @@ int x509_check_for_self_signed(struct x509_certificate *cert)
>  			goto out;
>  	}
>  
> -	ret = -EKEYREJECTED;
> -	if (strcmp(cert->pub->pkey_algo, cert->sig->pkey_algo) != 0 &&
> -	    (strncmp(cert->pub->pkey_algo, "ecdsa-", 6) != 0 ||
> -	     strcmp(cert->sig->pkey_algo, "ecdsa") != 0))
> -		goto out;
> -
>  	ret = public_key_verify_signature(cert->pub, cert->sig);
>  	if (ret < 0) {
>  		if (ret == -ENOPKG) {
> -- 
> 2.35.1

Re: [PATCH 1/2] KEYS: asymmetric: enforce that sig algo matches key algo

[Eric Biggers]

On Wed, Feb 02, 2022 at 05:52:30AM +0300, Vitaly Chikunov wrote:
> Eric,
> 
> On Mon, Jan 31, 2022 at 04:34:13PM -0800, Eric Biggers wrote:
> > From: Eric Biggers <ebiggers@google.com>
> > 
> > Most callers of public_key_verify_signature(), including most indirect
> > callers via verify_signature() as well as pkcs7_verify_sig_chain(),
> > don't check that public_key_signature::pkey_algo matches
> > public_key::pkey_algo.  These should always match.
> 
> Why should they match?

For the reasons I explain in the rest of the commit message.  To summarize: to
have a valid signature verification scheme the algorithm must be fixed by the
key, and not attacker-controlled.

> 
> public_key_signature is the data prepared to verify the cert's
> signature. The cert's signature algorithm could be different from the
> public key algorithm defined in the cert itself. They should match only
> for self-signed certs. For example, you should be able to sign RSA
> public key with ECDSA signature and vice versa. Or 256-bit EC-RDSA with
> 512-bit EC-RDSA. This check will prevent this.

That has nothing to do with this patch, as this patch is only dealing with the
signature.  A cert's public key algorithm can be different, and that is fine.

> > diff --git a/crypto/asymmetric_keys/public_key.c b/crypto/asymmetric_keys/public_key.c
> > index 4fefb219bfdc8..aba7113d86c76 100644
> > --- a/crypto/asymmetric_keys/public_key.c
> > +++ b/crypto/asymmetric_keys/public_key.c
> > @@ -325,6 +325,21 @@ int public_key_verify_signature(const struct public_key *pkey,
> >  	BUG_ON(!sig);
> >  	BUG_ON(!sig->s);
> >  
> > +	/*
> > +	 * The signature's claimed public key algorithm *must* match the key's
> > +	 * actual public key algorithm.
> > +	 *
> > +	 * Small exception: ECDSA signatures don't specify the curve, but ECDSA
> > +	 * keys do.  So the strings can mismatch slightly in that case:
> > +	 * "ecdsa-nist-*" for the key, but "ecdsa" for the signature.
> > +	 */
> > +	if (!sig->pkey_algo)
> > +		return -EINVAL;
> 
> This seem incorrect too, as sig->pkey_algo could be NULL for direct
> signature verification calls. For example, for keyctl pkey_verify.

We can make it optional if some callers aren't providing it.  Of course, such
callers wouldn't be able to verify ECDSA signatures.

- Eric

Re: [PATCH 1/2] KEYS: asymmetric: enforce that sig algo matches key algo

[Eric Biggers]

On Tue, Feb 01, 2022 at 07:10:33PM -0800, Eric Biggers wrote:
> > This seem incorrect too, as sig->pkey_algo could be NULL for direct
> > signature verification calls. For example, for keyctl pkey_verify.
> 
> We can make it optional if some callers aren't providing it.  Of course, such
> callers wouldn't be able to verify ECDSA signatures.

Sorry, I got that backwards.  ECDSA signatures don't specify the curve, but the
keys do (as I noted in a comment).  So ECDSA wouldn't require sig->pkey_algo.

Since it appears that KEYCTL_PKEY_VERIFY does in fact have no way to specify a
pkey_algo, I'll allow NULL pkey_algo in v2.

Note that SM2 isn't implemented correctly when sig->pkey_algo is NULL, as the
following code incorrectly uses the signature's pkey_algo rather than the key's:

        if (sig->pkey_algo && strcmp(sig->pkey_algo, "sm2") == 0 &&
            sig->data_size) {
                ret = cert_sig_digest_update(sig, tfm);
                if (ret)
                        goto error_free_key;
        }

I'm not sure whether I should even bother fixing that, given how broken the SM2
stuff is anyway.

- Eric

Re: [PATCH 1/2] KEYS: asymmetric: enforce that sig algo matches key algo

[Vitaly Chikunov]

On Tue, Feb 01, 2022 at 07:10:33PM -0800, Eric Biggers wrote:
> On Wed, Feb 02, 2022 at 05:52:30AM +0300, Vitaly Chikunov wrote:
> > Eric,
> > 
> > On Mon, Jan 31, 2022 at 04:34:13PM -0800, Eric Biggers wrote:
> > > From: Eric Biggers <ebiggers@google.com>
> > > 
> > > Most callers of public_key_verify_signature(), including most indirect
> > > callers via verify_signature() as well as pkcs7_verify_sig_chain(),
> > > don't check that public_key_signature::pkey_algo matches
> > > public_key::pkey_algo.  These should always match.
> > 
> > Why should they match?
> 
> For the reasons I explain in the rest of the commit message.  To summarize: to
> have a valid signature verification scheme the algorithm must be fixed by the
> key, and not attacker-controlled.
> 
> > 
> > public_key_signature is the data prepared to verify the cert's
> > signature. The cert's signature algorithm could be different from the
> > public key algorithm defined in the cert itself. They should match only
> > for self-signed certs. For example, you should be able to sign RSA
> > public key with ECDSA signature and vice versa. Or 256-bit EC-RDSA with
> > 512-bit EC-RDSA. This check will prevent this.
> 
> That has nothing to do with this patch, as this patch is only dealing with the
> signature.  A cert's public key algorithm can be different, and that is fine.

You are right and I was mistaken about that (obscured by keyctl
pkey_verify error and self-signed keys verification). Then it's all
good!

I also tested these patches to work well with rsa-ecdsa and ecrdsa
certificates using keyctl restrict_keyring.

Thanks,

> 
> > > diff --git a/crypto/asymmetric_keys/public_key.c b/crypto/asymmetric_keys/public_key.c
> > > index 4fefb219bfdc8..aba7113d86c76 100644
> > > --- a/crypto/asymmetric_keys/public_key.c
> > > +++ b/crypto/asymmetric_keys/public_key.c
> > > @@ -325,6 +325,21 @@ int public_key_verify_signature(const struct public_key *pkey,
> > >  	BUG_ON(!sig);
> > >  	BUG_ON(!sig->s);
> > >  
> > > +	/*
> > > +	 * The signature's claimed public key algorithm *must* match the key's
> > > +	 * actual public key algorithm.
> > > +	 *
> > > +	 * Small exception: ECDSA signatures don't specify the curve, but ECDSA
> > > +	 * keys do.  So the strings can mismatch slightly in that case:
> > > +	 * "ecdsa-nist-*" for the key, but "ecdsa" for the signature.
> > > +	 */
> > > +	if (!sig->pkey_algo)
> > > +		return -EINVAL;
> > 
> > This seem incorrect too, as sig->pkey_algo could be NULL for direct
> > signature verification calls. For example, for keyctl pkey_verify.
> 
> We can make it optional if some callers aren't providing it.  Of course, such
> callers wouldn't be able to verify ECDSA signatures.
> 
> - Eric

Re: [PATCH 1/2] KEYS: asymmetric: enforce that sig algo matches key algo

[Jarkko Sakkinen]

On Mon, Jan 31, 2022 at 04:34:13PM -0800, Eric Biggers wrote:
> From: Eric Biggers <ebiggers@google.com>
> 
> Most callers of public_key_verify_signature(), including most indirect
> callers via verify_signature() as well as pkcs7_verify_sig_chain(),
> don't check that public_key_signature::pkey_algo matches
> public_key::pkey_algo.  These should always match.  However, a malicious
> signature could intentionally declare an unintended algorithm.  It is
> essential that such signatures be rejected outright, or that the
> algorithm of the *key* be used -- not the algorithm of the signature as
> that would allow attackers to choose the algorithm used.
> 
> Currently, public_key_verify_signature() correctly uses the key's
> algorithm when deciding which akcipher to allocate.  That's good.
> However, it uses the signature's algorithm when deciding whether to do
> the first step of SM2, which is incorrect.  Also, v4.19 and older
> kernels used the signature's algorithm for the entire process.
> 
> Prevent such errors by making public_key_verify_signature() enforce that
> the signature's algorithm matches the key's algorithm.
> 
> Also remove two checks of this done by callers, which are now redundant.
> 
> Cc: stable@vger.kernel.org
> Signed-off-by: Eric Biggers <ebiggers@google.com>
> ---
>  crypto/asymmetric_keys/pkcs7_verify.c    |  6 ------
>  crypto/asymmetric_keys/public_key.c      | 15 +++++++++++++++
>  crypto/asymmetric_keys/x509_public_key.c |  6 ------
>  3 files changed, 15 insertions(+), 12 deletions(-)
> 
> diff --git a/crypto/asymmetric_keys/pkcs7_verify.c b/crypto/asymmetric_keys/pkcs7_verify.c
> index 0b4d07aa88111..f94a1d1ad3a6c 100644
> --- a/crypto/asymmetric_keys/pkcs7_verify.c
> +++ b/crypto/asymmetric_keys/pkcs7_verify.c
> @@ -174,12 +174,6 @@ static int pkcs7_find_key(struct pkcs7_message *pkcs7,
>  		pr_devel("Sig %u: Found cert serial match X.509[%u]\n",
>  			 sinfo->index, certix);
>  
> -		if (strcmp(x509->pub->pkey_algo, sinfo->sig->pkey_algo) != 0) {
> -			pr_warn("Sig %u: X.509 algo and PKCS#7 sig algo don't match\n",
> -				sinfo->index);
> -			continue;
> -		}
> -
>  		sinfo->signer = x509;
>  		return 0;
>  	}
> diff --git a/crypto/asymmetric_keys/public_key.c b/crypto/asymmetric_keys/public_key.c
> index 4fefb219bfdc8..aba7113d86c76 100644
> --- a/crypto/asymmetric_keys/public_key.c
> +++ b/crypto/asymmetric_keys/public_key.c
> @@ -325,6 +325,21 @@ int public_key_verify_signature(const struct public_key *pkey,
>  	BUG_ON(!sig);
>  	BUG_ON(!sig->s);
>  
> +	/*
> +	 * The signature's claimed public key algorithm *must* match the key's
> +	 * actual public key algorithm.
> +	 *
> +	 * Small exception: ECDSA signatures don't specify the curve, but ECDSA
> +	 * keys do.  So the strings can mismatch slightly in that case:
> +	 * "ecdsa-nist-*" for the key, but "ecdsa" for the signature.
> +	 */
> +	if (!sig->pkey_algo)
> +		return -EINVAL;
> +	if (strcmp(pkey->pkey_algo, sig->pkey_algo) != 0 &&
> +	    (strncmp(pkey->pkey_algo, "ecdsa-", 6) != 0 ||
> +	     strcmp(sig->pkey_algo, "ecdsa") != 0))
> +		return -EKEYREJECTED;
> +
>  	ret = software_key_determine_akcipher(sig->encoding,
>  					      sig->hash_algo,
>  					      pkey, alg_name);
> diff --git a/crypto/asymmetric_keys/x509_public_key.c b/crypto/asymmetric_keys/x509_public_key.c
> index fe14cae115b51..71cc1738fbfd2 100644
> --- a/crypto/asymmetric_keys/x509_public_key.c
> +++ b/crypto/asymmetric_keys/x509_public_key.c
> @@ -128,12 +128,6 @@ int x509_check_for_self_signed(struct x509_certificate *cert)
>  			goto out;
>  	}
>  
> -	ret = -EKEYREJECTED;
> -	if (strcmp(cert->pub->pkey_algo, cert->sig->pkey_algo) != 0 &&
> -	    (strncmp(cert->pub->pkey_algo, "ecdsa-", 6) != 0 ||
> -	     strcmp(cert->sig->pkey_algo, "ecdsa") != 0))
> -		goto out;
> -
>  	ret = public_key_verify_signature(cert->pub, cert->sig);
>  	if (ret < 0) {
>  		if (ret == -ENOPKG) {
> -- 
> 2.35.1
> 

Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>

David, do you want to pick this?

BR, Jarkko

Re: [PATCH 2/2] KEYS: asymmetric: properly validate hash_algo and encoding

[Jarkko Sakkinen]

On Mon, Jan 31, 2022 at 04:34:14PM -0800, Eric Biggers wrote:
> From: Eric Biggers <ebiggers@google.com>
> 
> It is insecure to allow arbitrary hash algorithms and signature
> encodings to be used with arbitrary signature algorithms.  Notably,
> ECDSA, ECRDSA, and SM2 all sign/verify raw hash values and don't
> disambiguate between different hash algorithms like RSA PKCS#1 v1.5
> padding does.  Therefore, they need to be restricted to certain sets of
> hash algorithms (ideally just one, but in practice small sets are used).
> Additionally, the encoding is an integral part of modern signature
> algorithms, and is not supposed to vary.
> 
> Therefore, tighten the checks of hash_algo and encoding done by
> software_key_determine_akcipher().
> 
> Also rearrange the parameters to software_key_determine_akcipher() to
> put the public_key first, as this is the most important parameter and it
> often determines everything else.
> 
> Fixes: 299f561a6693 ("x509: Add support for parsing x509 certs with ECDSA keys")
> Fixes: 215525639631 ("X.509: support OSCCA SM2-with-SM3 certificate verification")
> Fixes: 0d7a78643f69 ("crypto: ecrdsa - add EC-RDSA (GOST 34.10) algorithm")
> Cc: stable@vger.kernel.org
> Signed-off-by: Eric Biggers <ebiggers@google.com>
> ---
>  crypto/asymmetric_keys/public_key.c | 111 +++++++++++++++++++---------
>  1 file changed, 76 insertions(+), 35 deletions(-)
> 
> diff --git a/crypto/asymmetric_keys/public_key.c b/crypto/asymmetric_keys/public_key.c
> index aba7113d86c76..a603ee8afdb8d 100644
> --- a/crypto/asymmetric_keys/public_key.c
> +++ b/crypto/asymmetric_keys/public_key.c
> @@ -60,39 +60,83 @@ static void public_key_destroy(void *payload0, void *payload3)
>  }
>  
>  /*
> - * Determine the crypto algorithm name.
> + * Given a public_key, and an encoding and hash_algo to be used for signing
> + * and/or verification with that key, determine the name of the corresponding
> + * akcipher algorithm.  Also check that encoding and hash_algo are allowed.
>   */
> -static
> -int software_key_determine_akcipher(const char *encoding,
> -				    const char *hash_algo,
> -				    const struct public_key *pkey,
> -				    char alg_name[CRYPTO_MAX_ALG_NAME])
> +static int
> +software_key_determine_akcipher(const struct public_key *pkey,
> +				const char *encoding, const char *hash_algo,
> +				char alg_name[CRYPTO_MAX_ALG_NAME])

Why is changing parameter order necessary?

BR, Jarkko

Re: [PATCH 2/2] KEYS: asymmetric: properly validate hash_algo and encoding

[Eric Biggers]

On Mon, Feb 21, 2022 at 02:46:26AM +0100, Jarkko Sakkinen wrote:
> On Mon, Jan 31, 2022 at 04:34:14PM -0800, Eric Biggers wrote:
> > From: Eric Biggers <ebiggers@google.com>
> > 
> > It is insecure to allow arbitrary hash algorithms and signature
> > encodings to be used with arbitrary signature algorithms.  Notably,
> > ECDSA, ECRDSA, and SM2 all sign/verify raw hash values and don't
> > disambiguate between different hash algorithms like RSA PKCS#1 v1.5
> > padding does.  Therefore, they need to be restricted to certain sets of
> > hash algorithms (ideally just one, but in practice small sets are used).
> > Additionally, the encoding is an integral part of modern signature
> > algorithms, and is not supposed to vary.
> > 
> > Therefore, tighten the checks of hash_algo and encoding done by
> > software_key_determine_akcipher().
> > 
> > Also rearrange the parameters to software_key_determine_akcipher() to
> > put the public_key first, as this is the most important parameter and it
> > often determines everything else.
> > 
> > Fixes: 299f561a6693 ("x509: Add support for parsing x509 certs with ECDSA keys")
> > Fixes: 215525639631 ("X.509: support OSCCA SM2-with-SM3 certificate verification")
> > Fixes: 0d7a78643f69 ("crypto: ecrdsa - add EC-RDSA (GOST 34.10) algorithm")
> > Cc: stable@vger.kernel.org
> > Signed-off-by: Eric Biggers <ebiggers@google.com>
> > ---
> >  crypto/asymmetric_keys/public_key.c | 111 +++++++++++++++++++---------
> >  1 file changed, 76 insertions(+), 35 deletions(-)
> > 
> > diff --git a/crypto/asymmetric_keys/public_key.c b/crypto/asymmetric_keys/public_key.c
> > index aba7113d86c76..a603ee8afdb8d 100644
> > --- a/crypto/asymmetric_keys/public_key.c
> > +++ b/crypto/asymmetric_keys/public_key.c
> > @@ -60,39 +60,83 @@ static void public_key_destroy(void *payload0, void *payload3)
> >  }
> >  
> >  /*
> > - * Determine the crypto algorithm name.
> > + * Given a public_key, and an encoding and hash_algo to be used for signing
> > + * and/or verification with that key, determine the name of the corresponding
> > + * akcipher algorithm.  Also check that encoding and hash_algo are allowed.
> >   */
> > -static
> > -int software_key_determine_akcipher(const char *encoding,
> > -				    const char *hash_algo,
> > -				    const struct public_key *pkey,
> > -				    char alg_name[CRYPTO_MAX_ALG_NAME])
> > +static int
> > +software_key_determine_akcipher(const struct public_key *pkey,
> > +				const char *encoding, const char *hash_algo,
> > +				char alg_name[CRYPTO_MAX_ALG_NAME])
> 
> Why is changing parameter order necessary?
> 

It's mentioned in the commit message.  It's obviously not necessary but this way
makes much more sense IMO.

- Eric

Re: [PATCH 2/2] KEYS: asymmetric: properly validate hash_algo and encoding

[Jarkko Sakkinen]

On Sun, Feb 20, 2022 at 06:21:36PM -0800, Eric Biggers wrote:
> On Mon, Feb 21, 2022 at 02:46:26AM +0100, Jarkko Sakkinen wrote:
> > On Mon, Jan 31, 2022 at 04:34:14PM -0800, Eric Biggers wrote:
> > > From: Eric Biggers <ebiggers@google.com>
> > > 
> > > It is insecure to allow arbitrary hash algorithms and signature
> > > encodings to be used with arbitrary signature algorithms.  Notably,
> > > ECDSA, ECRDSA, and SM2 all sign/verify raw hash values and don't
> > > disambiguate between different hash algorithms like RSA PKCS#1 v1.5
> > > padding does.  Therefore, they need to be restricted to certain sets of
> > > hash algorithms (ideally just one, but in practice small sets are used).
> > > Additionally, the encoding is an integral part of modern signature
> > > algorithms, and is not supposed to vary.
> > > 
> > > Therefore, tighten the checks of hash_algo and encoding done by
> > > software_key_determine_akcipher().
> > > 
> > > Also rearrange the parameters to software_key_determine_akcipher() to
> > > put the public_key first, as this is the most important parameter and it
> > > often determines everything else.
> > > 
> > > Fixes: 299f561a6693 ("x509: Add support for parsing x509 certs with ECDSA keys")
> > > Fixes: 215525639631 ("X.509: support OSCCA SM2-with-SM3 certificate verification")
> > > Fixes: 0d7a78643f69 ("crypto: ecrdsa - add EC-RDSA (GOST 34.10) algorithm")
> > > Cc: stable@vger.kernel.org
> > > Signed-off-by: Eric Biggers <ebiggers@google.com>
> > > ---
> > >  crypto/asymmetric_keys/public_key.c | 111 +++++++++++++++++++---------
> > >  1 file changed, 76 insertions(+), 35 deletions(-)
> > > 
> > > diff --git a/crypto/asymmetric_keys/public_key.c b/crypto/asymmetric_keys/public_key.c
> > > index aba7113d86c76..a603ee8afdb8d 100644
> > > --- a/crypto/asymmetric_keys/public_key.c
> > > +++ b/crypto/asymmetric_keys/public_key.c
> > > @@ -60,39 +60,83 @@ static void public_key_destroy(void *payload0, void *payload3)
> > >  }
> > >  
> > >  /*
> > > - * Determine the crypto algorithm name.
> > > + * Given a public_key, and an encoding and hash_algo to be used for signing
> > > + * and/or verification with that key, determine the name of the corresponding
> > > + * akcipher algorithm.  Also check that encoding and hash_algo are allowed.
> > >   */
> > > -static
> > > -int software_key_determine_akcipher(const char *encoding,
> > > -				    const char *hash_algo,
> > > -				    const struct public_key *pkey,
> > > -				    char alg_name[CRYPTO_MAX_ALG_NAME])
> > > +static int
> > > +software_key_determine_akcipher(const struct public_key *pkey,
> > > +				const char *encoding, const char *hash_algo,
> > > +				char alg_name[CRYPTO_MAX_ALG_NAME])
> > 
> > Why is changing parameter order necessary?
> > 
> 
> It's mentioned in the commit message.  It's obviously not necessary but this way
> makes much more sense IMO.

Ah, so it is.

Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>

BR, Jarkko

Re: [PATCH 1/2] KEYS: asymmetric: enforce that sig algo matches key algo

[Eric Biggers]

On Mon, Feb 21, 2022 at 02:43:21AM +0100, Jarkko Sakkinen wrote:
> 
> Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>
> 
> David, do you want to pick this?
> 

No response from David, so I think you need to take these.

- Eric

Re: [PATCH 1/2] KEYS: asymmetric: enforce that sig algo matches key algo

[Jarkko Sakkinen]

On Fri, Mar 04, 2022 at 11:26:10AM -0800, Eric Biggers wrote:
> On Mon, Feb 21, 2022 at 02:43:21AM +0100, Jarkko Sakkinen wrote:
> > 
> > Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>
> > 
> > David, do you want to pick this?
> > 
> 
> No response from David, so I think you need to take these.
> 
> - Eric

I did, they're applied now, thank you.

BR, Jarkko