[v1,4/5] ima: support fs-verity file digest based signatures

Commit Message

Mimi Zohar Dec. 2, 2021, 9:55 p.m. UTC

Instead of calculating a file hash and verifying the signature stored
in the security.ima xattr against the calculated file hash, verify the
signature based on the fs-verity's file digest and other metadata.  The
fs-verity file digest is a hash that includes the Merkle tree root hash.

Signed-off-by: Mimi Zohar <[e-mail]>
---
Changelog v1:
- Based on Eric Bigger's review, instead of verifying the fsverity's file
digest directly, sign a hash of it with other file metadata.

 security/integrity/ima/ima_api.c      | 20 ++++++++++++
 security/integrity/ima/ima_appraise.c | 45 ++++++++++++++++++++++++++-
 2 files changed, 64 insertions(+), 1 deletion(-)

Comments

Eric Biggers Dec. 2, 2021, 10:07 p.m. UTC | #1

On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
>  	case IMA_VERITY_DIGSIG:
> -		fallthrough;
> +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
> +
> +		/*
> +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
> +		 * and the fs-verity file digest, not directly on the
> +		 * fs-verity file digest.  Both digests should probably be
> +		 * included in the IMA measurement list, but for now this
> +		 * digest is only used for verifying the IMA signature.
> +		 */
> +		verity_digest[0] = IMA_VERITY_DIGSIG;
> +		memcpy(verity_digest + 1, iint->ima_hash->digest,
> +		       iint->ima_hash->length);
> +
> +		hash.hdr.algo = iint->ima_hash->algo;
> +		hash.hdr.length = iint->ima_hash->length;

This is still wrong because the bytes being signed don't include the hash
algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
supports SHA-512 too, and it may support other hash algorithms in the future.

- Eric

Mimi Zohar Dec. 2, 2021, 10:13 p.m. UTC | #2

On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
> On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
> >  	case IMA_VERITY_DIGSIG:
> > -		fallthrough;
> > +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
> > +
> > +		/*
> > +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
> > +		 * and the fs-verity file digest, not directly on the
> > +		 * fs-verity file digest.  Both digests should probably be
> > +		 * included in the IMA measurement list, but for now this
> > +		 * digest is only used for verifying the IMA signature.
> > +		 */
> > +		verity_digest[0] = IMA_VERITY_DIGSIG;
> > +		memcpy(verity_digest + 1, iint->ima_hash->digest,
> > +		       iint->ima_hash->length);
> > +
> > +		hash.hdr.algo = iint->ima_hash->algo;
> > +		hash.hdr.length = iint->ima_hash->length;
> 
> This is still wrong because the bytes being signed don't include the hash
> algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
> supports SHA-512 too, and it may support other hash algorithms in the future.

The signature stored in security.ima is prefixed with a header
(signature_v2_hdr).

Mimi

Eric Biggers Dec. 2, 2021, 10:18 p.m. UTC | #3

On Thu, Dec 02, 2021 at 05:13:15PM -0500, Mimi Zohar wrote:
> On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
> > On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
> > >  	case IMA_VERITY_DIGSIG:
> > > -		fallthrough;
> > > +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
> > > +
> > > +		/*
> > > +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
> > > +		 * and the fs-verity file digest, not directly on the
> > > +		 * fs-verity file digest.  Both digests should probably be
> > > +		 * included in the IMA measurement list, but for now this
> > > +		 * digest is only used for verifying the IMA signature.
> > > +		 */
> > > +		verity_digest[0] = IMA_VERITY_DIGSIG;
> > > +		memcpy(verity_digest + 1, iint->ima_hash->digest,
> > > +		       iint->ima_hash->length);
> > > +
> > > +		hash.hdr.algo = iint->ima_hash->algo;
> > > +		hash.hdr.length = iint->ima_hash->length;
> > 
> > This is still wrong because the bytes being signed don't include the hash
> > algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
> > supports SHA-512 too, and it may support other hash algorithms in the future.
> 
> The signature stored in security.ima is prefixed with a header
> (signature_v2_hdr).

Yes, but the byte that identifies the hash algorithm is not included in the
bytes that are actually signed, as far as I can tell.

- Eric

Mimi Zohar Dec. 31, 2021, 3:35 p.m. UTC | #4

Hi Eric,

On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
> On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
> >  	case IMA_VERITY_DIGSIG:
> > -		fallthrough;
> > +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
> > +
> > +		/*
> > +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
> > +		 * and the fs-verity file digest, not directly on the
> > +		 * fs-verity file digest.  Both digests should probably be
> > +		 * included in the IMA measurement list, but for now this
> > +		 * digest is only used for verifying the IMA signature.
> > +		 */
> > +		verity_digest[0] = IMA_VERITY_DIGSIG;
> > +		memcpy(verity_digest + 1, iint->ima_hash->digest,
> > +		       iint->ima_hash->length);
> > +
> > +		hash.hdr.algo = iint->ima_hash->algo;
> > +		hash.hdr.length = iint->ima_hash->length;
> 
> This is still wrong because the bytes being signed don't include the hash
> algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
> supports SHA-512 too, and it may support other hash algorithms in the future.

IMA assumes that the file hash algorithm and the signature algorithm
are the same.   If they're not the same, for whatever reason, the
signature verification would simply fail.

Based on the v2 signature header 'type' field, IMA can differentiate
between regular IMA file hash based signatures and fs-verity file
digest based signatures.  The digest field (d-ng) in the IMA
meausrement list prefixes the digest with the hash algorithm. I'm
missing the reason for needing to hash fs-verity's file digest with
other metadata, and sign that hash rather than fs-verity's file digest
directly.

thanks,

Mimi

Eric Biggers Jan. 5, 2022, 11:37 p.m. UTC | #5

On Fri, Dec 31, 2021 at 10:35:00AM -0500, Mimi Zohar wrote:
> Hi Eric,
> 
> On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
> > On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
> > >  	case IMA_VERITY_DIGSIG:
> > > -		fallthrough;
> > > +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
> > > +
> > > +		/*
> > > +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
> > > +		 * and the fs-verity file digest, not directly on the
> > > +		 * fs-verity file digest.  Both digests should probably be
> > > +		 * included in the IMA measurement list, but for now this
> > > +		 * digest is only used for verifying the IMA signature.
> > > +		 */
> > > +		verity_digest[0] = IMA_VERITY_DIGSIG;
> > > +		memcpy(verity_digest + 1, iint->ima_hash->digest,
> > > +		       iint->ima_hash->length);
> > > +
> > > +		hash.hdr.algo = iint->ima_hash->algo;
> > > +		hash.hdr.length = iint->ima_hash->length;
> > 
> > This is still wrong because the bytes being signed don't include the hash
> > algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
> > supports SHA-512 too, and it may support other hash algorithms in the future.
> 
> IMA assumes that the file hash algorithm and the signature algorithm
> are the same.   If they're not the same, for whatever reason, the
> signature verification would simply fail.
> 
> Based on the v2 signature header 'type' field, IMA can differentiate
> between regular IMA file hash based signatures and fs-verity file
> digest based signatures.  The digest field (d-ng) in the IMA
> meausrement list prefixes the digest with the hash algorithm. I'm
> missing the reason for needing to hash fs-verity's file digest with
> other metadata, and sign that hash rather than fs-verity's file digest
> directly.

Because if someone signs a raw hash, then they also implicitly sign the same
hash value for all supported hash algorithms that produce the same length hash.
Signing a raw hash is only appropriate when there is only 1 supported algorithm.

All the other stuff you mentioned is irrelevant.

- Eric

Vitaly Chikunov Jan. 9, 2022, 8:45 p.m. UTC | #6

Eric,

On Wed, Jan 05, 2022 at 03:37:39PM -0800, Eric Biggers wrote:
> On Fri, Dec 31, 2021 at 10:35:00AM -0500, Mimi Zohar wrote:
> > On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
> > > On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
> > > >  	case IMA_VERITY_DIGSIG:
> > > > -		fallthrough;
> > > > +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
> > > > +
> > > > +		/*
> > > > +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
> > > > +		 * and the fs-verity file digest, not directly on the
> > > > +		 * fs-verity file digest.  Both digests should probably be
> > > > +		 * included in the IMA measurement list, but for now this
> > > > +		 * digest is only used for verifying the IMA signature.
> > > > +		 */
> > > > +		verity_digest[0] = IMA_VERITY_DIGSIG;
> > > > +		memcpy(verity_digest + 1, iint->ima_hash->digest,
> > > > +		       iint->ima_hash->length);
> > > > +
> > > > +		hash.hdr.algo = iint->ima_hash->algo;
> > > > +		hash.hdr.length = iint->ima_hash->length;
> > > 
> > > This is still wrong because the bytes being signed don't include the hash
> > > algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
> > > supports SHA-512 too, and it may support other hash algorithms in the future.
> > 
> > IMA assumes that the file hash algorithm and the signature algorithm
> > are the same.   If they're not the same, for whatever reason, the
> > signature verification would simply fail.
> > 
> > Based on the v2 signature header 'type' field, IMA can differentiate
> > between regular IMA file hash based signatures and fs-verity file
> > digest based signatures.  The digest field (d-ng) in the IMA
> > meausrement list prefixes the digest with the hash algorithm. I'm
> > missing the reason for needing to hash fs-verity's file digest with
> > other metadata, and sign that hash rather than fs-verity's file digest
> > directly.
> 
> Because if someone signs a raw hash, then they also implicitly sign the same
> hash value for all supported hash algorithms that produce the same length hash.

Unless there is broken hash algorithm allowing for preimage attacks this
is irrelevant. If there is two broken algorithms allowing for collisions,
colliding hashes could be prepared even if algo id is hashed too.

Thanks,

> Signing a raw hash is only appropriate when there is only 1 supported algorithm.
> 
> All the other stuff you mentioned is irrelevant.
> 
> - Eric

Eric Biggers Jan. 9, 2022, 9:07 p.m. UTC | #7

On Sun, Jan 09, 2022 at 11:45:37PM +0300, Vitaly Chikunov wrote:
> Eric,
> 
> On Wed, Jan 05, 2022 at 03:37:39PM -0800, Eric Biggers wrote:
> > On Fri, Dec 31, 2021 at 10:35:00AM -0500, Mimi Zohar wrote:
> > > On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
> > > > On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
> > > > >  	case IMA_VERITY_DIGSIG:
> > > > > -		fallthrough;
> > > > > +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
> > > > > +
> > > > > +		/*
> > > > > +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
> > > > > +		 * and the fs-verity file digest, not directly on the
> > > > > +		 * fs-verity file digest.  Both digests should probably be
> > > > > +		 * included in the IMA measurement list, but for now this
> > > > > +		 * digest is only used for verifying the IMA signature.
> > > > > +		 */
> > > > > +		verity_digest[0] = IMA_VERITY_DIGSIG;
> > > > > +		memcpy(verity_digest + 1, iint->ima_hash->digest,
> > > > > +		       iint->ima_hash->length);
> > > > > +
> > > > > +		hash.hdr.algo = iint->ima_hash->algo;
> > > > > +		hash.hdr.length = iint->ima_hash->length;
> > > > 
> > > > This is still wrong because the bytes being signed don't include the hash
> > > > algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
> > > > supports SHA-512 too, and it may support other hash algorithms in the future.
> > > 
> > > IMA assumes that the file hash algorithm and the signature algorithm
> > > are the same.   If they're not the same, for whatever reason, the
> > > signature verification would simply fail.
> > > 
> > > Based on the v2 signature header 'type' field, IMA can differentiate
> > > between regular IMA file hash based signatures and fs-verity file
> > > digest based signatures.  The digest field (d-ng) in the IMA
> > > meausrement list prefixes the digest with the hash algorithm. I'm
> > > missing the reason for needing to hash fs-verity's file digest with
> > > other metadata, and sign that hash rather than fs-verity's file digest
> > > directly.
> > 
> > Because if someone signs a raw hash, then they also implicitly sign the same
> > hash value for all supported hash algorithms that produce the same length hash.
> 
> Unless there is broken hash algorithm allowing for preimage attacks this
> is irrelevant. If there is two broken algorithms allowing for collisions,
> colliding hashes could be prepared even if algo id is hashed too.
> 

Only one algorithm needs to be broken.  For example, SM3 has the same hash
length as SHA-256.  If SM3 support were to be added to fs-verity, and if someone
were to find a way to find an input that has a specific SM3 digest, then they
could also make it match a specific SHA-256 digest.  Someone might intend to
sign a SHA-256 digest, but if they are only signing the raw 32 bytes of the
digest, then they would also be signing the corresponding SM3 digest.  That's
why the digest that is signed *must* also include the algorithm used in the
digest (not the algorithm(s) used in the signature, which is different).

- Eric

Vitaly Chikunov Jan. 15, 2022, 5:31 a.m. UTC | #8

Eric,

On Sun, Jan 09, 2022 at 01:07:18PM -0800, Eric Biggers wrote:
> On Sun, Jan 09, 2022 at 11:45:37PM +0300, Vitaly Chikunov wrote:
> > On Wed, Jan 05, 2022 at 03:37:39PM -0800, Eric Biggers wrote:
> > > On Fri, Dec 31, 2021 at 10:35:00AM -0500, Mimi Zohar wrote:
> > > > On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
> > > > > On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
> > > > > >  	case IMA_VERITY_DIGSIG:
> > > > > > -		fallthrough;
> > > > > > +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
> > > > > > +
> > > > > > +		/*
> > > > > > +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
> > > > > > +		 * and the fs-verity file digest, not directly on the
> > > > > > +		 * fs-verity file digest.  Both digests should probably be
> > > > > > +		 * included in the IMA measurement list, but for now this
> > > > > > +		 * digest is only used for verifying the IMA signature.
> > > > > > +		 */
> > > > > > +		verity_digest[0] = IMA_VERITY_DIGSIG;
> > > > > > +		memcpy(verity_digest + 1, iint->ima_hash->digest,
> > > > > > +		       iint->ima_hash->length);
> > > > > > +
> > > > > > +		hash.hdr.algo = iint->ima_hash->algo;
> > > > > > +		hash.hdr.length = iint->ima_hash->length;
> > > > > 
> > > > > This is still wrong because the bytes being signed don't include the hash
> > > > > algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
> > > > > supports SHA-512 too, and it may support other hash algorithms in the future.
> > > > 
> > > > IMA assumes that the file hash algorithm and the signature algorithm
> > > > are the same.   If they're not the same, for whatever reason, the
> > > > signature verification would simply fail.
> > > > 
> > > > Based on the v2 signature header 'type' field, IMA can differentiate
> > > > between regular IMA file hash based signatures and fs-verity file
> > > > digest based signatures.  The digest field (d-ng) in the IMA
> > > > meausrement list prefixes the digest with the hash algorithm. I'm
> > > > missing the reason for needing to hash fs-verity's file digest with
> > > > other metadata, and sign that hash rather than fs-verity's file digest
> > > > directly.
> > > 
> > > Because if someone signs a raw hash, then they also implicitly sign the same
> > > hash value for all supported hash algorithms that produce the same length hash.
> > 
> > Unless there is broken hash algorithm allowing for preimage attacks this
> > is irrelevant. If there is two broken algorithms allowing for collisions,
> > colliding hashes could be prepared even if algo id is hashed too.
> > 
> 
> Only one algorithm needs to be broken.  For example, SM3 has the same hash
> length as SHA-256.  If SM3 support were to be added to fs-verity, and if someone
> were to find a way to find an input that has a specific SM3 digest, then they
> could also make it match a specific SHA-256 digest.  Someone might intend to
> sign a SHA-256 digest, but if they are only signing the raw 32 bytes of the
> digest, then they would also be signing the corresponding SM3 digest.  That's
> why the digest that is signed *must* also include the algorithm used in the
> digest (not the algorithm(s) used in the signature, which is different).

I think it will be beneficial if we pass hash algo id to the
akcipher_alg::verify. In fact, ecrdsa should only be used with streebog.
And perhaps, sm2 with sm3, pkcs1 with md/sha/sm3, and ecdsa with sha family
hashes.

Thanks,


> 
> - Eric

Eric Biggers Jan. 15, 2022, 6:21 a.m. UTC | #9

On Sat, Jan 15, 2022 at 08:31:01AM +0300, Vitaly Chikunov wrote:
> Eric,
> 
> On Sun, Jan 09, 2022 at 01:07:18PM -0800, Eric Biggers wrote:
> > On Sun, Jan 09, 2022 at 11:45:37PM +0300, Vitaly Chikunov wrote:
> > > On Wed, Jan 05, 2022 at 03:37:39PM -0800, Eric Biggers wrote:
> > > > On Fri, Dec 31, 2021 at 10:35:00AM -0500, Mimi Zohar wrote:
> > > > > On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
> > > > > > On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
> > > > > > >  	case IMA_VERITY_DIGSIG:
> > > > > > > -		fallthrough;
> > > > > > > +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
> > > > > > > +
> > > > > > > +		/*
> > > > > > > +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
> > > > > > > +		 * and the fs-verity file digest, not directly on the
> > > > > > > +		 * fs-verity file digest.  Both digests should probably be
> > > > > > > +		 * included in the IMA measurement list, but for now this
> > > > > > > +		 * digest is only used for verifying the IMA signature.
> > > > > > > +		 */
> > > > > > > +		verity_digest[0] = IMA_VERITY_DIGSIG;
> > > > > > > +		memcpy(verity_digest + 1, iint->ima_hash->digest,
> > > > > > > +		       iint->ima_hash->length);
> > > > > > > +
> > > > > > > +		hash.hdr.algo = iint->ima_hash->algo;
> > > > > > > +		hash.hdr.length = iint->ima_hash->length;
> > > > > > 
> > > > > > This is still wrong because the bytes being signed don't include the hash
> > > > > > algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
> > > > > > supports SHA-512 too, and it may support other hash algorithms in the future.
> > > > > 
> > > > > IMA assumes that the file hash algorithm and the signature algorithm
> > > > > are the same.   If they're not the same, for whatever reason, the
> > > > > signature verification would simply fail.
> > > > > 
> > > > > Based on the v2 signature header 'type' field, IMA can differentiate
> > > > > between regular IMA file hash based signatures and fs-verity file
> > > > > digest based signatures.  The digest field (d-ng) in the IMA
> > > > > meausrement list prefixes the digest with the hash algorithm. I'm
> > > > > missing the reason for needing to hash fs-verity's file digest with
> > > > > other metadata, and sign that hash rather than fs-verity's file digest
> > > > > directly.
> > > > 
> > > > Because if someone signs a raw hash, then they also implicitly sign the same
> > > > hash value for all supported hash algorithms that produce the same length hash.
> > > 
> > > Unless there is broken hash algorithm allowing for preimage attacks this
> > > is irrelevant. If there is two broken algorithms allowing for collisions,
> > > colliding hashes could be prepared even if algo id is hashed too.
> > > 
> > 
> > Only one algorithm needs to be broken.  For example, SM3 has the same hash
> > length as SHA-256.  If SM3 support were to be added to fs-verity, and if someone
> > were to find a way to find an input that has a specific SM3 digest, then they
> > could also make it match a specific SHA-256 digest.  Someone might intend to
> > sign a SHA-256 digest, but if they are only signing the raw 32 bytes of the
> > digest, then they would also be signing the corresponding SM3 digest.  That's
> > why the digest that is signed *must* also include the algorithm used in the
> > digest (not the algorithm(s) used in the signature, which is different).
> 
> I think it will be beneficial if we pass hash algo id to the
> akcipher_alg::verify. In fact, ecrdsa should only be used with streebog.
> And perhaps, sm2 with sm3, pkcs1 with md/sha/sm3, and ecdsa with sha family
> hashes.
> 

I was going to reply to this thread again, but I got a bit distracted by
everything else being broken.  Yes, the kernel needs to be restricting which
hash algorithms can be used with each public key algorithm, along the lines of
what you said.  I asked the BoringSSL maintainers for advice, and they confirmed
that ECDSA just signs/verifies a raw hash, and in fact it *must* be a raw hash
for it to be secure.  This is a design flaw in ECDSA, which was fixed in newer
algorithms such as EdDSA and SM2 as those have a hash built-in to the signature
scheme.  To mitigate it, the allowed hash algorithms must be restricted; in the
case of ECDSA, that means to the SHA family (preferably excluding SHA-1).

akcipher_alg::verify doesn't actually know which hash algorithm is used, except
in the case of rsa-pkcs1pad where it is built into the name of the algorithm.
So it can't check the hash algorithm.  I believe it needs to happen in
public_key_verify_signature() (and I'm working on a patch for that).

Now, SM2 is different from ECDSA and ECRDSA in that it uses the modern design
that includes the hash into the signature algorithm.  This means that it must be
used to sign/verify *data*, not a hash.  (Well, you can sign/verify a hash, but
SM2 will hash it again internally.)  Currently, public_key_verify_signature()
allows SM2 to be used to sign/verify a hash, skipping the SM2 internal hash, and
IMA uses this.  This is broken and must be removed, since it isn't actually the
SM2 algorithm as specified anymore, but rather some homebrew thing with unknown
security properties. (Well, I'm not confident about SM2, but homebrew is worse.)

Adding fs-verity support to IMA also complicates things, as doing it naively
would introduce an ambiguity about what is signed.  Naively, the *data* that is
signed (considering the hash as part of the signature algorithm) would be either
the whole file, in the case of traditional IMA, or the fsverity_descriptor
struct, in the case of IMA with fs-verity.  However, a file could have contents
which match an fsverity_descriptor struct; that would create an ambiguity.

Assuming that it needs to be allowed that the same key can sign files for both
traditional and fs-verity hashing, solving this problem will require a second
hash.  The easiest way to do this would be sign/verify the following struct:

	struct ima_file_id {
		u8 is_fsverity;
		u8 hash_algorithm;
		u8 hash[];
	};

This would be the *data* that is signed/verified -- meaning that it would be
hashed again as part of the signature algorithm (whether that hash is built-in
to the signature algorithm, as is the case for modern algorithms, or handled by
the caller as is the case for legacy algorithms).  Note that both traditional
and fs-verity hashes would need to use this same method for it to be secure; the
kernel must not accept signatures using the old method at the same time.

- Eric

Stefan Berger Jan. 16, 2022, 3:31 a.m. UTC | #10

On 1/15/22 01:21, Eric Biggers wrote:
> On Sat, Jan 15, 2022 at 08:31:01AM +0300, Vitaly Chikunov wrote:
>> Eric,
>>
>> On Sun, Jan 09, 2022 at 01:07:18PM -0800, Eric Biggers wrote:
>>> On Sun, Jan 09, 2022 at 11:45:37PM +0300, Vitaly Chikunov wrote:
>>>> On Wed, Jan 05, 2022 at 03:37:39PM -0800, Eric Biggers wrote:
>>>>> On Fri, Dec 31, 2021 at 10:35:00AM -0500, Mimi Zohar wrote:
>>>>>> On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
>>>>>>> On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
>>>>>>>>   	case IMA_VERITY_DIGSIG:
>>>>>>>> -		fallthrough;
>>>>>>>> +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
>>>>>>>> +
>>>>>>>> +		/*
>>>>>>>> +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
>>>>>>>> +		 * and the fs-verity file digest, not directly on the
>>>>>>>> +		 * fs-verity file digest.  Both digests should probably be
>>>>>>>> +		 * included in the IMA measurement list, but for now this
>>>>>>>> +		 * digest is only used for verifying the IMA signature.
>>>>>>>> +		 */
>>>>>>>> +		verity_digest[0] = IMA_VERITY_DIGSIG;
>>>>>>>> +		memcpy(verity_digest + 1, iint->ima_hash->digest,
>>>>>>>> +		       iint->ima_hash->length);
>>>>>>>> +
>>>>>>>> +		hash.hdr.algo = iint->ima_hash->algo;
>>>>>>>> +		hash.hdr.length = iint->ima_hash->length;
>>>>>>> This is still wrong because the bytes being signed don't include the hash
>>>>>>> algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
>>>>>>> supports SHA-512 too, and it may support other hash algorithms in the future.
>>>>>> IMA assumes that the file hash algorithm and the signature algorithm
>>>>>> are the same.   If they're not the same, for whatever reason, the
>>>>>> signature verification would simply fail.
>>>>>>
>>>>>> Based on the v2 signature header 'type' field, IMA can differentiate
>>>>>> between regular IMA file hash based signatures and fs-verity file
>>>>>> digest based signatures.  The digest field (d-ng) in the IMA
>>>>>> meausrement list prefixes the digest with the hash algorithm. I'm
>>>>>> missing the reason for needing to hash fs-verity's file digest with
>>>>>> other metadata, and sign that hash rather than fs-verity's file digest
>>>>>> directly.
>>>>> Because if someone signs a raw hash, then they also implicitly sign the same
>>>>> hash value for all supported hash algorithms that produce the same length hash.
>>>> Unless there is broken hash algorithm allowing for preimage attacks this
>>>> is irrelevant. If there is two broken algorithms allowing for collisions,
>>>> colliding hashes could be prepared even if algo id is hashed too.
>>>>
>>> Only one algorithm needs to be broken.  For example, SM3 has the same hash
>>> length as SHA-256.  If SM3 support were to be added to fs-verity, and if someone
>>> were to find a way to find an input that has a specific SM3 digest, then they
>>> could also make it match a specific SHA-256 digest.  Someone might intend to
>>> sign a SHA-256 digest, but if they are only signing the raw 32 bytes of the
>>> digest, then they would also be signing the corresponding SM3 digest.  That's
>>> why the digest that is signed *must* also include the algorithm used in the
>>> digest (not the algorithm(s) used in the signature, which is different).
>> I think it will be beneficial if we pass hash algo id to the
>> akcipher_alg::verify. In fact, ecrdsa should only be used with streebog.
>> And perhaps, sm2 with sm3, pkcs1 with md/sha/sm3, and ecdsa with sha family
>> hashes.
>>
> I was going to reply to this thread again, but I got a bit distracted by
> everything else being broken.  Yes, the kernel needs to be restricting which
> hash algorithms can be used with each public key algorithm, along the lines of
> what you said.  I asked the BoringSSL maintainers for advice, and they confirmed
> that ECDSA just signs/verifies a raw hash, and in fact it *must* be a raw hash
> for it to be secure.  This is a design flaw in ECDSA, which was fixed in newer
> algorithms such as EdDSA and SM2 as those have a hash built-in to the signature
> scheme.  To mitigate it, the allowed hash algorithms must be restricted; in the
> case of ECDSA, that means to the SHA family (preferably excluding SHA-1).
>
> akcipher_alg::verify doesn't actually know which hash algorithm is used, except
> in the case of rsa-pkcs1pad where it is built into the name of the algorithm.
> So it can't check the hash algorithm.  I believe it needs to happen in
> public_key_verify_signature() (and I'm working on a patch for that).
>
> Now, SM2 is different from ECDSA and ECRDSA in that it uses the modern design
> that includes the hash into the signature algorithm.  This means that it must be
> used to sign/verify *data*, not a hash.  (Well, you can sign/verify a hash, but
> SM2 will hash it again internally.)  Currently, public_key_verify_signature()
> allows SM2 to be used to sign/verify a hash, skipping the SM2 internal hash, and
> IMA uses this.  This is broken and must be removed, since it isn't actually the
> SM2 algorithm as specified anymore, but rather some homebrew thing with unknown
> security properties. (Well, I'm not confident about SM2, but homebrew is worse.)
>
> Adding fs-verity support to IMA also complicates things, as doing it naively
> would introduce an ambiguity about what is signed.  Naively, the *data* that is
> signed (considering the hash as part of the signature algorithm) would be either
> the whole file, in the case of traditional IMA, or the fsverity_descriptor
> struct, in the case of IMA with fs-verity.  However, a file could have contents
> which match an fsverity_descriptor struct; that would create an ambiguity.
>
> Assuming that it needs to be allowed that the same key can sign files for both
> traditional and fs-verity hashing, solving this problem will require a second
> hash.  The easiest way to do this would be sign/verify the following struct:
>
> 	struct ima_file_id {
> 		u8 is_fsverity;
> 		u8 hash_algorithm;
> 		u8 hash[];
> 	};


To calrify, I suppose that for ECDSA NIST P256 you would allow pairing 
with any of the SHA family hashes (also as defined by the existing OIDs) 
and as the standard allows today? And the same then applies for NIST 
p384 etc.?

Further, I suppose similar restriction would apply for ECRDSA to pair it 
with Streebog only, as Vitaly said.


What's happening now is that to verify a signature, IMA/integrity 
subsystem fills out the following structure:

struct public_key_signature pks;

pks.hash_algo = hash_algo_name[hdr->hash_algo];  // name of hash algo 
will go into this here, e.g., 'sha256'
pks.pkey_algo = pk->pkey_algo; // this is either 'rsa', 'ecdsa-', 
'ecrdsa-' or 'sm2' string

It then calls:

     ret = verify_signature(key, &pks);

IMO, in the call path down this function the pairing of public key and 
hash algo would have to be enforced in order to enforce the standards. 
Would this not be sufficient to be able to stay with the standards ?

File hashes: IMA calculates the hash over a file itself by calling 
crypto functions, so at least the digest's bytes are trusted input in 
that respect and using the sha family type of hashes directly with ECDSA 
should work. Which algorithm IMA is supposed to use for the hashing is 
given in the xattr bytestream header. IMA could then take that type of 
hash, lookup the hash function, perform the hashing on the data, and let 
verify_signature enforce the pairing, rejecting file signatures with 
wrong pairing. This way the only thing that is needed is 'enforcement of 
pairing'.

Fsverity: How much control does a user have over the hash family 
fsverity is using? Can IMA ECDSA/RSA users tell it to use a sha family 
hash and ECRDSA users make it use a Streebog hash so that also the 
pairing of hash and key type can work 'naturally' and we don't need the 
level of indirection via your structure above?

     Stefan



> This would be the *data* that is signed/verified -- meaning that it would be
> hashed again as part of the signature algorithm (whether that hash is built-in
> to the signature algorithm, as is the case for modern algorithms, or handled by
> the caller as is the case for legacy algorithms).  Note that both traditional
> and fs-verity hashes would need to use this same method for it to be secure; the
> kernel must not accept signatures using the old method at the same time.
>
> - Eric

Stefan Berger Jan. 16, 2022, 5:24 a.m. UTC | #11

On 1/15/22 22:31, Stefan Berger wrote:
>
> On 1/15/22 01:21, Eric Biggers wrote:
>> On Sat, Jan 15, 2022 at 08:31:01AM +0300, Vitaly Chikunov wrote:
>>> Eric,
>>>
>>> On Sun, Jan 09, 2022 at 01:07:18PM -0800, Eric Biggers wrote:
>>>> On Sun, Jan 09, 2022 at 11:45:37PM +0300, Vitaly Chikunov wrote:
>>>>> On Wed, Jan 05, 2022 at 03:37:39PM -0800, Eric Biggers wrote:
>>>>>> On Fri, Dec 31, 2021 at 10:35:00AM -0500, Mimi Zohar wrote:
>>>>>>> On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
>>>>>>>> On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
>>>>>>>>>       case IMA_VERITY_DIGSIG:
>>>>>>>>> -        fallthrough;
>>>>>>>>> +        set_bit(IMA_DIGSIG, &iint->atomic_flags);
>>>>>>>>> +
>>>>>>>>> +        /*
>>>>>>>>> +         * The IMA signature is based on a hash of 
>>>>>>>>> IMA_VERITY_DIGSIG
>>>>>>>>> +         * and the fs-verity file digest, not directly on the
>>>>>>>>> +         * fs-verity file digest.  Both digests should 
>>>>>>>>> probably be
>>>>>>>>> +         * included in the IMA measurement list, but for now 
>>>>>>>>> this
>>>>>>>>> +         * digest is only used for verifying the IMA signature.
>>>>>>>>> +         */
>>>>>>>>> +        verity_digest[0] = IMA_VERITY_DIGSIG;
>>>>>>>>> +        memcpy(verity_digest + 1, iint->ima_hash->digest,
>>>>>>>>> +               iint->ima_hash->length);
>>>>>>>>> +
>>>>>>>>> +        hash.hdr.algo = iint->ima_hash->algo;
>>>>>>>>> +        hash.hdr.length = iint->ima_hash->length;
>>>>>>>> This is still wrong because the bytes being signed don't 
>>>>>>>> include the hash
>>>>>>>> algorithm.  Unless you mean for it to be implicitly always 
>>>>>>>> SHA-256?  fs-verity
>>>>>>>> supports SHA-512 too, and it may support other hash algorithms 
>>>>>>>> in the future.
>>>>>>> IMA assumes that the file hash algorithm and the signature 
>>>>>>> algorithm
>>>>>>> are the same.   If they're not the same, for whatever reason, the
>>>>>>> signature verification would simply fail.
>>>>>>>
>>>>>>> Based on the v2 signature header 'type' field, IMA can 
>>>>>>> differentiate
>>>>>>> between regular IMA file hash based signatures and fs-verity file
>>>>>>> digest based signatures.  The digest field (d-ng) in the IMA
>>>>>>> meausrement list prefixes the digest with the hash algorithm. I'm
>>>>>>> missing the reason for needing to hash fs-verity's file digest with
>>>>>>> other metadata, and sign that hash rather than fs-verity's file 
>>>>>>> digest
>>>>>>> directly.
>>>>>> Because if someone signs a raw hash, then they also implicitly 
>>>>>> sign the same
>>>>>> hash value for all supported hash algorithms that produce the 
>>>>>> same length hash.
>>>>> Unless there is broken hash algorithm allowing for preimage 
>>>>> attacks this
>>>>> is irrelevant. If there is two broken algorithms allowing for 
>>>>> collisions,
>>>>> colliding hashes could be prepared even if algo id is hashed too.
>>>>>
>>>> Only one algorithm needs to be broken.  For example, SM3 has the 
>>>> same hash
>>>> length as SHA-256.  If SM3 support were to be added to fs-verity, 
>>>> and if someone
>>>> were to find a way to find an input that has a specific SM3 digest, 
>>>> then they
>>>> could also make it match a specific SHA-256 digest.  Someone might 
>>>> intend to
>>>> sign a SHA-256 digest, but if they are only signing the raw 32 
>>>> bytes of the
>>>> digest, then they would also be signing the corresponding SM3 
>>>> digest.  That's
>>>> why the digest that is signed *must* also include the algorithm 
>>>> used in the
>>>> digest (not the algorithm(s) used in the signature, which is 
>>>> different).
>>> I think it will be beneficial if we pass hash algo id to the
>>> akcipher_alg::verify. In fact, ecrdsa should only be used with 
>>> streebog.
>>> And perhaps, sm2 with sm3, pkcs1 with md/sha/sm3, and ecdsa with sha 
>>> family
>>> hashes.
>>>
>> I was going to reply to this thread again, but I got a bit distracted by
>> everything else being broken.  Yes, the kernel needs to be 
>> restricting which
>> hash algorithms can be used with each public key algorithm, along the 
>> lines of
>> what you said.  I asked the BoringSSL maintainers for advice, and 
>> they confirmed
>> that ECDSA just signs/verifies a raw hash, and in fact it *must* be a 
>> raw hash
>> for it to be secure.  This is a design flaw in ECDSA, which was fixed 
>> in newer
>> algorithms such as EdDSA and SM2 as those have a hash built-in to the 
>> signature
>> scheme.  To mitigate it, the allowed hash algorithms must be 
>> restricted; in the
>> case of ECDSA, that means to the SHA family (preferably excluding 
>> SHA-1).
>>
>> akcipher_alg::verify doesn't actually know which hash algorithm is 
>> used, except
>> in the case of rsa-pkcs1pad where it is built into the name of the 
>> algorithm.
>> So it can't check the hash algorithm.  I believe it needs to happen in
>> public_key_verify_signature() (and I'm working on a patch for that).
>>
>> Now, SM2 is different from ECDSA and ECRDSA in that it uses the 
>> modern design
>> that includes the hash into the signature algorithm.  This means that 
>> it must be
>> used to sign/verify *data*, not a hash.  (Well, you can sign/verify a 
>> hash, but
>> SM2 will hash it again internally.)  Currently, 
>> public_key_verify_signature()
>> allows SM2 to be used to sign/verify a hash, skipping the SM2 
>> internal hash, and
>> IMA uses this.  This is broken and must be removed, since it isn't 
>> actually the
>> SM2 algorithm as specified anymore, but rather some homebrew thing 
>> with unknown
>> security properties. (Well, I'm not confident about SM2, but homebrew 
>> is worse.)
>>
>> Adding fs-verity support to IMA also complicates things, as doing it 
>> naively
>> would introduce an ambiguity about what is signed.  Naively, the 
>> *data* that is
>> signed (considering the hash as part of the signature algorithm) 
>> would be either
>> the whole file, in the case of traditional IMA, or the 
>> fsverity_descriptor
>> struct, in the case of IMA with fs-verity.  However, a file could 
>> have contents
>> which match an fsverity_descriptor struct; that would create an 
>> ambiguity.
>>
>> Assuming that it needs to be allowed that the same key can sign files 
>> for both
>> traditional and fs-verity hashing, solving this problem will require 
>> a second
>> hash.  The easiest way to do this would be sign/verify the following 
>> struct:
>>
>>     struct ima_file_id {
>>         u8 is_fsverity;
>>         u8 hash_algorithm;
>>         u8 hash[];
>>     };
>
>
> To calrify, I suppose that for ECDSA NIST P256 you would allow pairing 
> with any of the SHA family hashes (also as defined by the existing 
> OIDs) and as the standard allows today? And the same then applies for 
> NIST p384 etc.?
>
> Further, I suppose similar restriction would apply for ECRDSA to pair 
> it with Streebog only, as Vitaly said.
>
>
> What's happening now is that to verify a signature, IMA/integrity 
> subsystem fills out the following structure:
>
> struct public_key_signature pks;
>
> pks.hash_algo = hash_algo_name[hdr->hash_algo];  // name of hash algo 
> will go into this here, e.g., 'sha256'
> pks.pkey_algo = pk->pkey_algo; // this is either 'rsa', 'ecdsa-', 
> 'ecrdsa-' or 'sm2' string
>
> It then calls:
>
>     ret = verify_signature(key, &pks);
>
> IMO, in the call path down this function the pairing of public key and 
> hash algo would have to be enforced in order to enforce the standards. 
> Would this not be sufficient to be able to stay with the standards ?
>
> File hashes: IMA calculates the hash over a file itself by calling 
> crypto functions, so at least the digest's bytes are trusted input in 
> that respect and using the sha family type of hashes directly with 
> ECDSA should work. Which algorithm IMA is supposed to use for the 
> hashing is given in the xattr bytestream header. IMA could then take 
> that type of hash, lookup the hash function, perform the hashing on 
> the data, and let verify_signature enforce the pairing, rejecting file 
> signatures with wrong pairing. This way the only thing that is needed 
> is 'enforcement of pairing'.
>
> Fsverity: How much control does a user have over the hash family 
> fsverity is using? Can IMA ECDSA/RSA users tell it to use a sha family 
> hash and ECRDSA users make it use a Streebog hash so that also the 
> pairing of hash and key type can work 'naturally' and we don't need 
> the level of indirection via your structure above?
>
>     Stefan
>
>

I would propos probably 3 patches that combined result in the following 
for enforcement of pairing. I think RSA isn't necessary to enforce since 
the OID in the signature contains the hash that was used. For SM2 I am 
not so sure about.

 From 800f11c32f6e64d328404c9fdd7abb6057dbebb9 Mon Sep 17 00:00:00 2001
From: Stefan Berger <[e-mail]>
Date: Sun, 16 Jan 2022 00:07:00 -0500
Subject: [PATCH] keys: asymmetric: Enforce key type and hash algo 
pairing for
  signatures

Enforce the pairing of ECRDSA with Streebog hash type, SM2 with SM3 hash,
and ECDSA with the SHA family of hashes for the low level
verify_signature() API.

Note: X509 certificates already have enforcement through the availability
of OIDs.

Signed-off-by: Stefan Berger <[e-mail]>
---
  crypto/asymmetric_keys/public_key.c | 59 +++++++++++++++++++++++++++++
  1 file changed, 59 insertions(+)

diff --git a/crypto/asymmetric_keys/public_key.c 
b/crypto/asymmetric_keys/public_key.c
index 4fefb219bfdc..760b541d2803 100644
--- a/crypto/asymmetric_keys/public_key.c
+++ b/crypto/asymmetric_keys/public_key.c
@@ -20,6 +20,7 @@
  #include <crypto/akcipher.h>
  #include <crypto/sm2.h>
  #include <crypto/sm3_base.h>
+#include <crypto/hash_info.h>

  MODULE_DESCRIPTION("In-software asymmetric public-key subtype");
  MODULE_AUTHOR("Red Hat, Inc.");
@@ -305,6 +306,60 @@ static inline int cert_sig_digest_update(
  }
  #endif /* ! IS_REACHABLE(CONFIG_CRYPTO_SM2) */

+/* ECDSA allowed hashes */
+static const enum hash_algo ecdsa_hashes[] = {
+       HASH_ALGO_SHA1,
+       HASH_ALGO_SHA224,
+       HASH_ALGO_SHA256,
+       HASH_ALGO_SHA384,
+       HASH_ALGO_SHA512,
+};
+
+/* SM2 allowed hashes */
+static const enum hash_algo sm2_hashes[] = {
+       HASH_ALGO_SM3_256,
+};
+
+/* ECRDSA allowed hashes */
+static const enum hash_algo ecrdsa_hashes[] = {
+       HASH_ALGO_STREEBOG_256,
+       HASH_ALGO_STREEBOG_512,
+};
+
+/*
+ * Check the pairing for key algorithm and hash for ECDSA
+ * and ECRDSA.
+ */
+static int check_pkey_hash_algo_pairing(const struct 
public_key_signature *sig)
+{
+       const enum hash_algo *array = NULL;
+       size_t array_size;
+       size_t i;
+
+       if (!strcmp(sig->pkey_algo, "ecrdsa")) {
+               array = ecrdsa_hashes;
+               array_size = ARRAY_SIZE(ecrdsa_hashes);
+       } else if (!strcmp(sig->pkey_algo, "sm2")) {
+               array = sm2_hashes;
+               array_size = ARRAY_SIZE(sm2_hashes);
+       } else if (!strncmp(sig->pkey_algo, "ecdsa-", 6)) {
+               /* ecdsa-nist-p192 etc. */
+               array = ecdsa_hashes;
+               array_size = ARRAY_SIZE(ecdsa_hashes);
+       }
+
+       if (array) {
+               for (i = 0; i < array_size; i++) {
+                       if (!strcmp(sig->hash_algo,
+                                   hash_algo_name[array[i]]))
+                               return 0;
+               }
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
  /*
   * Verify a signature using a public key.
   */
@@ -325,6 +380,10 @@ int public_key_verify_signature(const struct 
public_key *pkey,
         BUG_ON(!sig);
         BUG_ON(!sig->s);

+       ret = check_pkey_hash_algo_pairing(sig);
+       if (ret < 0)
+               return ret;
+
         ret = software_key_determine_akcipher(sig->encoding,
                                               sig->hash_algo,
                                               pkey, alg_name);
--
2.31.1

Mimi Zohar Jan. 16, 2022, 5:01 p.m. UTC | #12

On Fri, 2022-01-14 at 22:21 -0800, Eric Biggers wrote:
> On Sat, Jan 15, 2022 at 08:31:01AM +0300, Vitaly Chikunov wrote:
> > Eric,
> > 
> > On Sun, Jan 09, 2022 at 01:07:18PM -0800, Eric Biggers wrote:
> > > On Sun, Jan 09, 2022 at 11:45:37PM +0300, Vitaly Chikunov wrote:
> > > > On Wed, Jan 05, 2022 at 03:37:39PM -0800, Eric Biggers wrote:
> > > > > On Fri, Dec 31, 2021 at 10:35:00AM -0500, Mimi Zohar wrote:
> > > > > > On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
> > > > > > > On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
> > > > > > > >  	case IMA_VERITY_DIGSIG:
> > > > > > > > -		fallthrough;
> > > > > > > > +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
> > > > > > > > +
> > > > > > > > +		/*
> > > > > > > > +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
> > > > > > > > +		 * and the fs-verity file digest, not directly on the
> > > > > > > > +		 * fs-verity file digest.  Both digests should probably be
> > > > > > > > +		 * included in the IMA measurement list, but for now this
> > > > > > > > +		 * digest is only used for verifying the IMA signature.
> > > > > > > > +		 */
> > > > > > > > +		verity_digest[0] = IMA_VERITY_DIGSIG;
> > > > > > > > +		memcpy(verity_digest + 1, iint->ima_hash->digest,
> > > > > > > > +		       iint->ima_hash->length);
> > > > > > > > +
> > > > > > > > +		hash.hdr.algo = iint->ima_hash->algo;
> > > > > > > > +		hash.hdr.length = iint->ima_hash->length;
> > > > > > > 
> > > > > > > This is still wrong because the bytes being signed don't include the hash
> > > > > > > algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
> > > > > > > supports SHA-512 too, and it may support other hash algorithms in the future.
> > > > > > 
> > > > > > IMA assumes that the file hash algorithm and the signature algorithm
> > > > > > are the same.   If they're not the same, for whatever reason, the
> > > > > > signature verification would simply fail.
> > > > > > 
> > > > > > Based on the v2 signature header 'type' field, IMA can differentiate
> > > > > > between regular IMA file hash based signatures and fs-verity file
> > > > > > digest based signatures.  The digest field (d-ng) in the IMA
> > > > > > meausrement list prefixes the digest with the hash algorithm. I'm
> > > > > > missing the reason for needing to hash fs-verity's file digest with
> > > > > > other metadata, and sign that hash rather than fs-verity's file digest
> > > > > > directly.
> > > > > 
> > > > > Because if someone signs a raw hash, then they also implicitly sign the same
> > > > > hash value for all supported hash algorithms that produce the same length hash.
> > > > 
> > > > Unless there is broken hash algorithm allowing for preimage attacks this
> > > > is irrelevant. If there is two broken algorithms allowing for collisions,
> > > > colliding hashes could be prepared even if algo id is hashed too.
> > > > 
> > > 
> > > Only one algorithm needs to be broken.  For example, SM3 has the same hash
> > > length as SHA-256.  If SM3 support were to be added to fs-verity, and if someone
> > > were to find a way to find an input that has a specific SM3 digest, then they
> > > could also make it match a specific SHA-256 digest.  Someone might intend to
> > > sign a SHA-256 digest, but if they are only signing the raw 32 bytes of the
> > > digest, then they would also be signing the corresponding SM3 digest.  That's
> > > why the digest that is signed *must* also include the algorithm used in the
> > > digest (not the algorithm(s) used in the signature, which is different).
> > 
> > I think it will be beneficial if we pass hash algo id to the
> > akcipher_alg::verify. In fact, ecrdsa should only be used with streebog.
> > And perhaps, sm2 with sm3, pkcs1 with md/sha/sm3, and ecdsa with sha family
> > hashes.
> > 
> 
> I was going to reply to this thread again, but I got a bit distracted by
> everything else being broken.  Yes, the kernel needs to be restricting which
> hash algorithms can be used with each public key algorithm, along the lines of
> what you said.  I asked the BoringSSL maintainers for advice, and they confirmed
> that ECDSA just signs/verifies a raw hash, and in fact it *must* be a raw hash
> for it to be secure.  This is a design flaw in ECDSA, which was fixed in newer
> algorithms such as EdDSA and SM2 as those have a hash built-in to the signature
> scheme.  To mitigate it, the allowed hash algorithms must be restricted; in the
> case of ECDSA, that means to the SHA family (preferably excluding SHA-1).
> 
> akcipher_alg::verify doesn't actually know which hash algorithm is used, except
> in the case of rsa-pkcs1pad where it is built into the name of the algorithm.
> So it can't check the hash algorithm.  I believe it needs to happen in
> public_key_verify_signature() (and I'm working on a patch for that).
> 
> Now, SM2 is different from ECDSA and ECRDSA in that it uses the modern design
> that includes the hash into the signature algorithm.  This means that it must be
> used to sign/verify *data*, not a hash.  (Well, you can sign/verify a hash, but
> SM2 will hash it again internally.)  Currently, public_key_verify_signature()
> allows SM2 to be used to sign/verify a hash, skipping the SM2 internal hash, and
> IMA uses this.  This is broken and must be removed, since it isn't actually the
> SM2 algorithm as specified anymore, but rather some homebrew thing with unknown
> security properties. (Well, I'm not confident about SM2, but homebrew is worse.)
> 
> Adding fs-verity support to IMA also complicates things, as doing it naively
> would introduce an ambiguity about what is signed.  Naively, the *data* that is
> signed (considering the hash as part of the signature algorithm) would be either
> the whole file, in the case of traditional IMA, or the fsverity_descriptor
> struct, in the case of IMA with fs-verity.  However, a file could have contents
> which match an fsverity_descriptor struct; that would create an ambiguity.
> 
> Assuming that it needs to be allowed that the same key can sign files for both
> traditional and fs-verity hashing, solving this problem will require a second
> hash.

The IMA fs-verity policy rule could require specifying the hash
algorithm.  If it would require specifying a particular key as well,
would hashing the hash then not be needed?

> The easiest way to do this would be sign/verify the following struct:
> 	struct ima_file_id {
> 		u8 is_fsverity;
> 		u8 hash_algorithm;
> 		u8 hash[];
> 	};
> 

The v2 version of this patch introduces the "ima_tbs_hash" structure,
which is more generic, since it uses the IMA xattr record type.  Other
than that, I don't see a difference.

> This would be the *data* that is signed/verified -- meaning that it would be
> hashed again as part of the signature algorithm (whether that hash is built-in
> to the signature algorithm, as is the case for modern algorithms, or handled by
> the caller as is the case for legacy algorithms).

There seems to be an inconsistency, here, with what you said above,
"... ECDSA just signs/verifies a raw hash, and in fact it *must* be a
raw hash for it to be secure."

> Note that both traditional
> and fs-verity hashes would need to use this same method for it to be secure; the
> kernel must not accept signatures using the old method at the same time.

The v2 version of this patch set signed the hash of a hash just for fs-
verity signatures.  Adding the equivalent support for regular file
hashes will require the version in the IMA signature_v2_hdr to be
incremented.  If the version is incremented now, both signatures
versions should then be able to co-exist.

thanks,

Mimi

Eric Biggers Jan. 19, 2022, 12:39 a.m. UTC | #13

On Sun, Jan 16, 2022 at 12:01:28PM -0500, Mimi Zohar wrote:
> On Fri, 2022-01-14 at 22:21 -0800, Eric Biggers wrote:
> > On Sat, Jan 15, 2022 at 08:31:01AM +0300, Vitaly Chikunov wrote:
> > > Eric,
> > > 
> > > On Sun, Jan 09, 2022 at 01:07:18PM -0800, Eric Biggers wrote:
> > > > On Sun, Jan 09, 2022 at 11:45:37PM +0300, Vitaly Chikunov wrote:
> > > > > On Wed, Jan 05, 2022 at 03:37:39PM -0800, Eric Biggers wrote:
> > > > > > On Fri, Dec 31, 2021 at 10:35:00AM -0500, Mimi Zohar wrote:
> > > > > > > On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
> > > > > > > > On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
> > > > > > > > >  	case IMA_VERITY_DIGSIG:
> > > > > > > > > -		fallthrough;
> > > > > > > > > +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
> > > > > > > > > +
> > > > > > > > > +		/*
> > > > > > > > > +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
> > > > > > > > > +		 * and the fs-verity file digest, not directly on the
> > > > > > > > > +		 * fs-verity file digest.  Both digests should probably be
> > > > > > > > > +		 * included in the IMA measurement list, but for now this
> > > > > > > > > +		 * digest is only used for verifying the IMA signature.
> > > > > > > > > +		 */
> > > > > > > > > +		verity_digest[0] = IMA_VERITY_DIGSIG;
> > > > > > > > > +		memcpy(verity_digest + 1, iint->ima_hash->digest,
> > > > > > > > > +		       iint->ima_hash->length);
> > > > > > > > > +
> > > > > > > > > +		hash.hdr.algo = iint->ima_hash->algo;
> > > > > > > > > +		hash.hdr.length = iint->ima_hash->length;
> > > > > > > > 
> > > > > > > > This is still wrong because the bytes being signed don't include the hash
> > > > > > > > algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
> > > > > > > > supports SHA-512 too, and it may support other hash algorithms in the future.
> > > > > > > 
> > > > > > > IMA assumes that the file hash algorithm and the signature algorithm
> > > > > > > are the same.   If they're not the same, for whatever reason, the
> > > > > > > signature verification would simply fail.
> > > > > > > 
> > > > > > > Based on the v2 signature header 'type' field, IMA can differentiate
> > > > > > > between regular IMA file hash based signatures and fs-verity file
> > > > > > > digest based signatures.  The digest field (d-ng) in the IMA
> > > > > > > meausrement list prefixes the digest with the hash algorithm. I'm
> > > > > > > missing the reason for needing to hash fs-verity's file digest with
> > > > > > > other metadata, and sign that hash rather than fs-verity's file digest
> > > > > > > directly.
> > > > > > 
> > > > > > Because if someone signs a raw hash, then they also implicitly sign the same
> > > > > > hash value for all supported hash algorithms that produce the same length hash.
> > > > > 
> > > > > Unless there is broken hash algorithm allowing for preimage attacks this
> > > > > is irrelevant. If there is two broken algorithms allowing for collisions,
> > > > > colliding hashes could be prepared even if algo id is hashed too.
> > > > > 
> > > > 
> > > > Only one algorithm needs to be broken.  For example, SM3 has the same hash
> > > > length as SHA-256.  If SM3 support were to be added to fs-verity, and if someone
> > > > were to find a way to find an input that has a specific SM3 digest, then they
> > > > could also make it match a specific SHA-256 digest.  Someone might intend to
> > > > sign a SHA-256 digest, but if they are only signing the raw 32 bytes of the
> > > > digest, then they would also be signing the corresponding SM3 digest.  That's
> > > > why the digest that is signed *must* also include the algorithm used in the
> > > > digest (not the algorithm(s) used in the signature, which is different).
> > > 
> > > I think it will be beneficial if we pass hash algo id to the
> > > akcipher_alg::verify. In fact, ecrdsa should only be used with streebog.
> > > And perhaps, sm2 with sm3, pkcs1 with md/sha/sm3, and ecdsa with sha family
> > > hashes.
> > > 
> > 
> > I was going to reply to this thread again, but I got a bit distracted by
> > everything else being broken.  Yes, the kernel needs to be restricting which
> > hash algorithms can be used with each public key algorithm, along the lines of
> > what you said.  I asked the BoringSSL maintainers for advice, and they confirmed
> > that ECDSA just signs/verifies a raw hash, and in fact it *must* be a raw hash
> > for it to be secure.  This is a design flaw in ECDSA, which was fixed in newer
> > algorithms such as EdDSA and SM2 as those have a hash built-in to the signature
> > scheme.  To mitigate it, the allowed hash algorithms must be restricted; in the
> > case of ECDSA, that means to the SHA family (preferably excluding SHA-1).
> > 
> > akcipher_alg::verify doesn't actually know which hash algorithm is used, except
> > in the case of rsa-pkcs1pad where it is built into the name of the algorithm.
> > So it can't check the hash algorithm.  I believe it needs to happen in
> > public_key_verify_signature() (and I'm working on a patch for that).
> > 
> > Now, SM2 is different from ECDSA and ECRDSA in that it uses the modern design
> > that includes the hash into the signature algorithm.  This means that it must be
> > used to sign/verify *data*, not a hash.  (Well, you can sign/verify a hash, but
> > SM2 will hash it again internally.)  Currently, public_key_verify_signature()
> > allows SM2 to be used to sign/verify a hash, skipping the SM2 internal hash, and
> > IMA uses this.  This is broken and must be removed, since it isn't actually the
> > SM2 algorithm as specified anymore, but rather some homebrew thing with unknown
> > security properties. (Well, I'm not confident about SM2, but homebrew is worse.)
> > 
> > Adding fs-verity support to IMA also complicates things, as doing it naively
> > would introduce an ambiguity about what is signed.  Naively, the *data* that is
> > signed (considering the hash as part of the signature algorithm) would be either
> > the whole file, in the case of traditional IMA, or the fsverity_descriptor
> > struct, in the case of IMA with fs-verity.  However, a file could have contents
> > which match an fsverity_descriptor struct; that would create an ambiguity.
> > 
> > Assuming that it needs to be allowed that the same key can sign files for both
> > traditional and fs-verity hashing, solving this problem will require a second
> > hash.
> 
> The IMA fs-verity policy rule could require specifying the hash
> algorithm.  If it would require specifying a particular key as well,
> would hashing the hash then not be needed?

If both of those were true, and if the key was never reused for other purposes
(either for other hash algorithms or for full file hashes), that would be fine.
Obviously, the IMA policy is trusted in this context, since it's what tells the
kernel to verify signatures in the first place.

But, I could see users getting this wrong and reusing keys.

> 
> > The easiest way to do this would be sign/verify the following struct:
> > 	struct ima_file_id {
> > 		u8 is_fsverity;
> > 		u8 hash_algorithm;
> > 		u8 hash[];
> > 	};
> > 
> 
> The v2 version of this patch introduces the "ima_tbs_hash" structure,
> which is more generic, since it uses the IMA xattr record type.  Other
> than that, I don't see a difference.
> 
> > This would be the *data* that is signed/verified -- meaning that it would be
> > hashed again as part of the signature algorithm (whether that hash is built-in
> > to the signature algorithm, as is the case for modern algorithms, or handled by
> > the caller as is the case for legacy algorithms).
> 
> There seems to be an inconsistency, here, with what you said above,
> "... ECDSA just signs/verifies a raw hash, and in fact it *must* be a
> raw hash for it to be secure."

There isn't an inconsistency.  ECDSA is among the algorithms where the caller is
expected to handle the hash.

It is confusing dealing with all these different signature algorithms.  I think
the right way to think about this is in terms of what *data* is being
signed/verified.  Currently the data is the full file contents.  I think it
needs to be made into an annotated hash, e.g. the struct I gave.

public_key_verify_signature() also needs to be fixed to support both types of
signature algorithms (caller-provided hash and internal hash) in a logical way.
Originally it only supported caller-provided hashes, but then SM2 support was
added and now it is super broken.

> > Note that both traditional
> > and fs-verity hashes would need to use this same method for it to be secure; the
> > kernel must not accept signatures using the old method at the same time.
> 
> The v2 version of this patch set signed the hash of a hash just for fs-
> verity signatures.  Adding the equivalent support for regular file
> hashes will require the version in the IMA signature_v2_hdr to be
> incremented.  If the version is incremented now, both signatures
> versions should then be able to co-exist.

That seems like a good thing, unless you want users to be responsible for only
ever signing full file hashes *or* fs-verity file hashes with each key.  That
seems like something that users will get wrong.

- Eric

Eric Biggers Jan. 19, 2022, 12:49 a.m. UTC | #14

On Sat, Jan 15, 2022 at 10:31:40PM -0500, Stefan Berger wrote:
> 
> On 1/15/22 01:21, Eric Biggers wrote:
> > On Sat, Jan 15, 2022 at 08:31:01AM +0300, Vitaly Chikunov wrote:
> > > Eric,
> > > 
> > > On Sun, Jan 09, 2022 at 01:07:18PM -0800, Eric Biggers wrote:
> > > > On Sun, Jan 09, 2022 at 11:45:37PM +0300, Vitaly Chikunov wrote:
> > > > > On Wed, Jan 05, 2022 at 03:37:39PM -0800, Eric Biggers wrote:
> > > > > > On Fri, Dec 31, 2021 at 10:35:00AM -0500, Mimi Zohar wrote:
> > > > > > > On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
> > > > > > > > On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
> > > > > > > > >   	case IMA_VERITY_DIGSIG:
> > > > > > > > > -		fallthrough;
> > > > > > > > > +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
> > > > > > > > > +
> > > > > > > > > +		/*
> > > > > > > > > +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
> > > > > > > > > +		 * and the fs-verity file digest, not directly on the
> > > > > > > > > +		 * fs-verity file digest.  Both digests should probably be
> > > > > > > > > +		 * included in the IMA measurement list, but for now this
> > > > > > > > > +		 * digest is only used for verifying the IMA signature.
> > > > > > > > > +		 */
> > > > > > > > > +		verity_digest[0] = IMA_VERITY_DIGSIG;
> > > > > > > > > +		memcpy(verity_digest + 1, iint->ima_hash->digest,
> > > > > > > > > +		       iint->ima_hash->length);
> > > > > > > > > +
> > > > > > > > > +		hash.hdr.algo = iint->ima_hash->algo;
> > > > > > > > > +		hash.hdr.length = iint->ima_hash->length;
> > > > > > > > This is still wrong because the bytes being signed don't include the hash
> > > > > > > > algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
> > > > > > > > supports SHA-512 too, and it may support other hash algorithms in the future.
> > > > > > > IMA assumes that the file hash algorithm and the signature algorithm
> > > > > > > are the same.   If they're not the same, for whatever reason, the
> > > > > > > signature verification would simply fail.
> > > > > > > 
> > > > > > > Based on the v2 signature header 'type' field, IMA can differentiate
> > > > > > > between regular IMA file hash based signatures and fs-verity file
> > > > > > > digest based signatures.  The digest field (d-ng) in the IMA
> > > > > > > meausrement list prefixes the digest with the hash algorithm. I'm
> > > > > > > missing the reason for needing to hash fs-verity's file digest with
> > > > > > > other metadata, and sign that hash rather than fs-verity's file digest
> > > > > > > directly.
> > > > > > Because if someone signs a raw hash, then they also implicitly sign the same
> > > > > > hash value for all supported hash algorithms that produce the same length hash.
> > > > > Unless there is broken hash algorithm allowing for preimage attacks this
> > > > > is irrelevant. If there is two broken algorithms allowing for collisions,
> > > > > colliding hashes could be prepared even if algo id is hashed too.
> > > > > 
> > > > Only one algorithm needs to be broken.  For example, SM3 has the same hash
> > > > length as SHA-256.  If SM3 support were to be added to fs-verity, and if someone
> > > > were to find a way to find an input that has a specific SM3 digest, then they
> > > > could also make it match a specific SHA-256 digest.  Someone might intend to
> > > > sign a SHA-256 digest, but if they are only signing the raw 32 bytes of the
> > > > digest, then they would also be signing the corresponding SM3 digest.  That's
> > > > why the digest that is signed *must* also include the algorithm used in the
> > > > digest (not the algorithm(s) used in the signature, which is different).
> > > I think it will be beneficial if we pass hash algo id to the
> > > akcipher_alg::verify. In fact, ecrdsa should only be used with streebog.
> > > And perhaps, sm2 with sm3, pkcs1 with md/sha/sm3, and ecdsa with sha family
> > > hashes.
> > > 
> > I was going to reply to this thread again, but I got a bit distracted by
> > everything else being broken.  Yes, the kernel needs to be restricting which
> > hash algorithms can be used with each public key algorithm, along the lines of
> > what you said.  I asked the BoringSSL maintainers for advice, and they confirmed
> > that ECDSA just signs/verifies a raw hash, and in fact it *must* be a raw hash
> > for it to be secure.  This is a design flaw in ECDSA, which was fixed in newer
> > algorithms such as EdDSA and SM2 as those have a hash built-in to the signature
> > scheme.  To mitigate it, the allowed hash algorithms must be restricted; in the
> > case of ECDSA, that means to the SHA family (preferably excluding SHA-1).
> > 
> > akcipher_alg::verify doesn't actually know which hash algorithm is used, except
> > in the case of rsa-pkcs1pad where it is built into the name of the algorithm.
> > So it can't check the hash algorithm.  I believe it needs to happen in
> > public_key_verify_signature() (and I'm working on a patch for that).
> > 
> > Now, SM2 is different from ECDSA and ECRDSA in that it uses the modern design
> > that includes the hash into the signature algorithm.  This means that it must be
> > used to sign/verify *data*, not a hash.  (Well, you can sign/verify a hash, but
> > SM2 will hash it again internally.)  Currently, public_key_verify_signature()
> > allows SM2 to be used to sign/verify a hash, skipping the SM2 internal hash, and
> > IMA uses this.  This is broken and must be removed, since it isn't actually the
> > SM2 algorithm as specified anymore, but rather some homebrew thing with unknown
> > security properties. (Well, I'm not confident about SM2, but homebrew is worse.)
> > 
> > Adding fs-verity support to IMA also complicates things, as doing it naively
> > would introduce an ambiguity about what is signed.  Naively, the *data* that is
> > signed (considering the hash as part of the signature algorithm) would be either
> > the whole file, in the case of traditional IMA, or the fsverity_descriptor
> > struct, in the case of IMA with fs-verity.  However, a file could have contents
> > which match an fsverity_descriptor struct; that would create an ambiguity.
> > 
> > Assuming that it needs to be allowed that the same key can sign files for both
> > traditional and fs-verity hashing, solving this problem will require a second
> > hash.  The easiest way to do this would be sign/verify the following struct:
> > 
> > 	struct ima_file_id {
> > 		u8 is_fsverity;
> > 		u8 hash_algorithm;
> > 		u8 hash[];
> > 	};
> 
> 
> To calrify, I suppose that for ECDSA NIST P256 you would allow pairing with
> any of the SHA family hashes (also as defined by the existing OIDs) and as
> the standard allows today? And the same then applies for NIST p384 etc.?
> 
> Further, I suppose similar restriction would apply for ECRDSA to pair it
> with Streebog only, as Vitaly said.

I don't have any better ideas.

> What's happening now is that to verify a signature, IMA/integrity subsystem
> fills out the following structure:
> 
> struct public_key_signature pks;
> 
> pks.hash_algo = hash_algo_name[hdr->hash_algo];  // name of hash algo will
> go into this here, e.g., 'sha256'
> pks.pkey_algo = pk->pkey_algo; // this is either 'rsa', 'ecdsa-', 'ecrdsa-'
> or 'sm2' string
> 
> It then calls:
> 
>     ret = verify_signature(key, &pks);
> 
> IMO, in the call path down this function the pairing of public key and hash
> algo would have to be enforced in order to enforce the standards. Would this
> not be sufficient to be able to stay with the standards ?

That sounds right, though there are a number of other issues including SM2 being
implemented incorrectly, the "encoding" string isn't validated, and it not being
enforced that public_key_signature::pkey_algo actually matches
public_key::pkey_algo.

> File hashes: IMA calculates the hash over a file itself by calling crypto
> functions, so at least the digest's bytes are trusted input in that respect
> and using the sha family type of hashes directly with ECDSA should work.
> Which algorithm IMA is supposed to use for the hashing is given in the xattr
> bytestream header. IMA could then take that type of hash, lookup the hash
> function, perform the hashing on the data, and let verify_signature enforce
> the pairing, rejecting file signatures with wrong pairing. This way the only
> thing that is needed is 'enforcement of pairing'.
> 
> Fsverity: How much control does a user have over the hash family fsverity is
> using? Can IMA ECDSA/RSA users tell it to use a sha family hash and ECRDSA
> users make it use a Streebog hash so that also the pairing of hash and key
> type can work 'naturally' and we don't need the level of indirection via
> your structure above?

The hash algorithm used by fs-verity is configurable and is always returned
along with the file digest.  Currently, only SHA-256 and SHA-512 are supported.

Keep in mind that if you sign the fs-verity file digest directly with RSA,
ECDSA, or ECRDSA, the *data* you are actually signing is the fsverity_descriptor
-- the struct which the hash is a hash of.

That creates an ambiguity when full file hashes are also signed by the same key,
as I previously mentioned.  A level of indirection is needed to avoid that.

In the naive method, the *data* being signed would also be different with SM2.
The level of indirection would avoid that.

- Eric

Stefan Berger Jan. 19, 2022, 3:41 p.m. UTC | #15

On 1/18/22 19:49, Eric Biggers wrote:
> On Sat, Jan 15, 2022 at 10:31:40PM -0500, Stefan Berger wrote:
>> On 1/15/22 01:21, Eric Biggers wrote:
>>> On Sat, Jan 15, 2022 at 08:31:01AM +0300, Vitaly Chikunov wrote:
>>>> Eric,
>>>>
>>>> On Sun, Jan 09, 2022 at 01:07:18PM -0800, Eric Biggers wrote:
>>>>> On Sun, Jan 09, 2022 at 11:45:37PM +0300, Vitaly Chikunov wrote:
>>>>>> On Wed, Jan 05, 2022 at 03:37:39PM -0800, Eric Biggers wrote:
>>>>>>> On Fri, Dec 31, 2021 at 10:35:00AM -0500, Mimi Zohar wrote:
>>>>>>>> On Thu, 2021-12-02 at 14:07 -0800, Eric Biggers wrote:
>>>>>>>>> On Thu, Dec 02, 2021 at 04:55:06PM -0500, Mimi Zohar wrote:
>>>>>>>>>>    	case IMA_VERITY_DIGSIG:
>>>>>>>>>> -		fallthrough;
>>>>>>>>>> +		set_bit(IMA_DIGSIG, &iint->atomic_flags);
>>>>>>>>>> +
>>>>>>>>>> +		/*
>>>>>>>>>> +		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
>>>>>>>>>> +		 * and the fs-verity file digest, not directly on the
>>>>>>>>>> +		 * fs-verity file digest.  Both digests should probably be
>>>>>>>>>> +		 * included in the IMA measurement list, but for now this
>>>>>>>>>> +		 * digest is only used for verifying the IMA signature.
>>>>>>>>>> +		 */
>>>>>>>>>> +		verity_digest[0] = IMA_VERITY_DIGSIG;
>>>>>>>>>> +		memcpy(verity_digest + 1, iint->ima_hash->digest,
>>>>>>>>>> +		       iint->ima_hash->length);
>>>>>>>>>> +
>>>>>>>>>> +		hash.hdr.algo = iint->ima_hash->algo;
>>>>>>>>>> +		hash.hdr.length = iint->ima_hash->length;
>>>>>>>>> This is still wrong because the bytes being signed don't include the hash
>>>>>>>>> algorithm.  Unless you mean for it to be implicitly always SHA-256?  fs-verity
>>>>>>>>> supports SHA-512 too, and it may support other hash algorithms in the future.
>>>>>>>> IMA assumes that the file hash algorithm and the signature algorithm
>>>>>>>> are the same.   If they're not the same, for whatever reason, the
>>>>>>>> signature verification would simply fail.
>>>>>>>>
>>>>>>>> Based on the v2 signature header 'type' field, IMA can differentiate
>>>>>>>> between regular IMA file hash based signatures and fs-verity file
>>>>>>>> digest based signatures.  The digest field (d-ng) in the IMA
>>>>>>>> meausrement list prefixes the digest with the hash algorithm. I'm
>>>>>>>> missing the reason for needing to hash fs-verity's file digest with
>>>>>>>> other metadata, and sign that hash rather than fs-verity's file digest
>>>>>>>> directly.
>>>>>>> Because if someone signs a raw hash, then they also implicitly sign the same
>>>>>>> hash value for all supported hash algorithms that produce the same length hash.
>>>>>> Unless there is broken hash algorithm allowing for preimage attacks this
>>>>>> is irrelevant. If there is two broken algorithms allowing for collisions,
>>>>>> colliding hashes could be prepared even if algo id is hashed too.
>>>>>>
>>>>> Only one algorithm needs to be broken.  For example, SM3 has the same hash
>>>>> length as SHA-256.  If SM3 support were to be added to fs-verity, and if someone
>>>>> were to find a way to find an input that has a specific SM3 digest, then they
>>>>> could also make it match a specific SHA-256 digest.  Someone might intend to
>>>>> sign a SHA-256 digest, but if they are only signing the raw 32 bytes of the
>>>>> digest, then they would also be signing the corresponding SM3 digest.  That's
>>>>> why the digest that is signed *must* also include the algorithm used in the
>>>>> digest (not the algorithm(s) used in the signature, which is different).
>>>> I think it will be beneficial if we pass hash algo id to the
>>>> akcipher_alg::verify. In fact, ecrdsa should only be used with streebog.
>>>> And perhaps, sm2 with sm3, pkcs1 with md/sha/sm3, and ecdsa with sha family
>>>> hashes.
>>>>
>>> I was going to reply to this thread again, but I got a bit distracted by
>>> everything else being broken.  Yes, the kernel needs to be restricting which
>>> hash algorithms can be used with each public key algorithm, along the lines of
>>> what you said.  I asked the BoringSSL maintainers for advice, and they confirmed
>>> that ECDSA just signs/verifies a raw hash, and in fact it *must* be a raw hash
>>> for it to be secure.  This is a design flaw in ECDSA, which was fixed in newer
>>> algorithms such as EdDSA and SM2 as those have a hash built-in to the signature
>>> scheme.  To mitigate it, the allowed hash algorithms must be restricted; in the
>>> case of ECDSA, that means to the SHA family (preferably excluding SHA-1).
>>>
>>> akcipher_alg::verify doesn't actually know which hash algorithm is used, except
>>> in the case of rsa-pkcs1pad where it is built into the name of the algorithm.
>>> So it can't check the hash algorithm.  I believe it needs to happen in
>>> public_key_verify_signature() (and I'm working on a patch for that).
>>>
>>> Now, SM2 is different from ECDSA and ECRDSA in that it uses the modern design
>>> that includes the hash into the signature algorithm.  This means that it must be
>>> used to sign/verify *data*, not a hash.  (Well, you can sign/verify a hash, but
>>> SM2 will hash it again internally.)  Currently, public_key_verify_signature()
>>> allows SM2 to be used to sign/verify a hash, skipping the SM2 internal hash, and
>>> IMA uses this.  This is broken and must be removed, since it isn't actually the
>>> SM2 algorithm as specified anymore, but rather some homebrew thing with unknown
>>> security properties. (Well, I'm not confident about SM2, but homebrew is worse.)
>>>
>>> Adding fs-verity support to IMA also complicates things, as doing it naively
>>> would introduce an ambiguity about what is signed.  Naively, the *data* that is
>>> signed (considering the hash as part of the signature algorithm) would be either
>>> the whole file, in the case of traditional IMA, or the fsverity_descriptor
>>> struct, in the case of IMA with fs-verity.  However, a file could have contents
>>> which match an fsverity_descriptor struct; that would create an ambiguity.
>>>
>>> Assuming that it needs to be allowed that the same key can sign files for both
>>> traditional and fs-verity hashing, solving this problem will require a second
>>> hash.  The easiest way to do this would be sign/verify the following struct:
>>>
>>> 	struct ima_file_id {
>>> 		u8 is_fsverity;
>>> 		u8 hash_algorithm;
>>> 		u8 hash[];
>>> 	};
>>
>> To calrify, I suppose that for ECDSA NIST P256 you would allow pairing with
>> any of the SHA family hashes (also as defined by the existing OIDs) and as
>> the standard allows today? And the same then applies for NIST p384 etc.?
>>
>> Further, I suppose similar restriction would apply for ECRDSA to pair it
>> with Streebog only, as Vitaly said.
> I don't have any better ideas.
>
>> What's happening now is that to verify a signature, IMA/integrity subsystem
>> fills out the following structure:
>>
>> struct public_key_signature pks;
>>
>> pks.hash_algo = hash_algo_name[hdr->hash_algo];  // name of hash algo will
>> go into this here, e.g., 'sha256'
>> pks.pkey_algo = pk->pkey_algo; // this is either 'rsa', 'ecdsa-', 'ecrdsa-'
>> or 'sm2' string
>>
>> It then calls:
>>
>>      ret = verify_signature(key, &pks);
>>
>> IMO, in the call path down this function the pairing of public key and hash
>> algo would have to be enforced in order to enforce the standards. Would this
>> not be sufficient to be able to stay with the standards ?
> That sounds right, though there are a number of other issues including SM2 being
> implemented incorrectly, the "encoding" string isn't validated, and it not being
> enforced that public_key_signature::pkey_algo actually matches
> public_key::pkey_algo.

I don't know enough about SM2.


Which call path are you looking at for "encoding" ?

For IMA's signature verification with public keys we will necessarily 
get into:

public_key_verfiy_signature: 
https://elixir.bootlin.com/linux/v5.14.21/source/crypto/asymmetric_keys/public_key.c#L311

sig->encoding is at least then used in software_key_determine_akcipher: 
https://elixir.bootlin.com/linux/v5.14.21/source/crypto/asymmetric_keys/public_key.c#L66

It doesn't *seem* to be used elsewhere down this call path. Is this not 
enough of looking at 'encoding' that is used to form the alg_name?


Regarding matching of  public_key_signature::pkey_algo and 
public_key::pkey_algo: What could be the implications of this not 
matching? Does it matter? Could one accidentally succeed in verifying a 
signature with the wrong type of key?


As for the proposed patch. I would need to split this up into 3 patches 
with their corresponding fixes tag, either SM2 or ECDRSA in the first 
depending on which one is oldest. But not knowing about SM2 I would 
probably skip this one.


>
>> File hashes: IMA calculates the hash over a file itself by calling crypto
>> functions, so at least the digest's bytes are trusted input in that respect
>> and using the sha family type of hashes directly with ECDSA should work.
>> Which algorithm IMA is supposed to use for the hashing is given in the xattr
>> bytestream header. IMA could then take that type of hash, lookup the hash
>> function, perform the hashing on the data, and let verify_signature enforce
>> the pairing, rejecting file signatures with wrong pairing. This way the only
>> thing that is needed is 'enforcement of pairing'.
>>
>> Fsverity: How much control does a user have over the hash family fsverity is
>> using? Can IMA ECDSA/RSA users tell it to use a sha family hash and ECRDSA
>> users make it use a Streebog hash so that also the pairing of hash and key
>> type can work 'naturally' and we don't need the level of indirection via
>> your structure above?
> The hash algorithm used by fs-verity is configurable and is always returned
> along with the file digest.  Currently, only SHA-256 and SHA-512 are supported.
>
> Keep in mind that if you sign the fs-verity file digest directly with RSA,
> ECDSA, or ECRDSA, the *data* you are actually signing is the fsverity_descriptor
> -- the struct which the hash is a hash of.
>
> That creates an ambiguity when full file hashes are also signed by the same key,
> as I previously mentioned.  A level of indirection is needed to avoid that.
>
> In the naive method, the *data* being signed would also be different with SM2.
> The level of indirection would avoid that.

So in the fsverity case that level of indirection is needed, for the 
existing file signatures I don't think we need it.

    Stefan

>
> - Eric

Mimi Zohar Jan. 20, 2022, 4:39 p.m. UTC | #16

Eric, Vitaly,

On Tue, 2022-01-18 at 16:39 -0800, Eric Biggers wrote:

> > > The easiest way to do this would be sign/verify the following struct:
> > > 	struct ima_file_id {
> > > 		u8 is_fsverity;
> > > 		u8 hash_algorithm;
> > > 		u8 hash[];
> > > 	};


> > This would be the *data* that is signed/verified -- meaning that it
would be
> > > hashed again as part of the signature algorithm (whether that hash is built-in
> > > to the signature algorithm, as is the case for modern algorithms, or handled by
> > > the caller as is the case for legacy algorithms).
> > 
> > There seems to be an inconsistency, here, with what you said above,
> > "... ECDSA just signs/verifies a raw hash, and in fact it *must* be a
> > raw hash for it to be secure."
> 
> There isn't an inconsistency.  ECDSA is among the algorithms where the caller is
> expected to handle the hash.
> 
> It is confusing dealing with all these different signature algorithms.  I think
> the right way to think about this is in terms of what *data* is being
> signed/verified.  Currently the data is the full file contents.  I think it
> needs to be made into an annotated hash, e.g. the struct I gave.
> 
> public_key_verify_signature() also needs to be fixed to support both types of
> signature algorithms (caller-provided hash and internal hash) in a logical way.
> Originally it only supported caller-provided hashes, but then SM2 support was
> added and now it is super broken.

Eric, did you say you're working on fixes to address these problems?

> 
> > > Note that both traditional
> > > and fs-verity hashes would need to use this same method for it to be secure; the
> > > kernel must not accept signatures using the old method at the same time.
> > 
> > The v2 version of this patch set signed the hash of a hash just for fs-
> > verity signatures.  Adding the equivalent support for regular file
> > hashes will require the version in the IMA signature_v2_hdr to be
> > incremented.  If the version is incremented now, both signatures
> > versions should then be able to co-exist.
> 
> That seems like a good thing, unless you want users to be responsible for only
> ever signing full file hashes *or* fs-verity file hashes with each key.  That
> seems like something that users will get wrong.

Instead of using a flexible array, Vitaly suggested defining the hash
as FS_VERITY_MAX_DIGEST_SIZE, so that it could be allocated temporarily
on stack
instead of kalloc.

As the above struct is not limited to fsverity, we could use
MAX_HASH_DIGESTSIZE, if it was exported, but it isn't.  Would the
following work for you?

/*
 * IMA signature header version 3 disambiguates the data that is signed
by
 * indirectly signing the hash of this structure, containing either the
 * fsverity_descriptor struct digest or, in the future, the traditional
IMA
 * file hash.
 */
struct ima_file_id {
        __u8 is_fsverity;       /* set to 1 for IMA_VERITY_DIGSIG */
        __u8 hash_algorithm;    /* Digest algorithm [enum hash_algo] */
#ifdef __KERNEL__
        __u8 hash[HASH_MAX_DIGESTSIZE];
#else
        __u8 hash[];
#endif
};

thanks,

Mimi

Eric Biggers Jan. 20, 2022, 9:05 p.m. UTC | #17

On Thu, Jan 20, 2022 at 11:39:56AM -0500, Mimi Zohar wrote:
> Eric, Vitaly,
> 
> On Tue, 2022-01-18 at 16:39 -0800, Eric Biggers wrote:
> 
> > > > The easiest way to do this would be sign/verify the following struct:
> > > > 	struct ima_file_id {
> > > > 		u8 is_fsverity;
> > > > 		u8 hash_algorithm;
> > > > 		u8 hash[];
> > > > 	};
> 
> 
> > > This would be the *data* that is signed/verified -- meaning that it
> would be
> > > > hashed again as part of the signature algorithm (whether that hash is built-in
> > > > to the signature algorithm, as is the case for modern algorithms, or handled by
> > > > the caller as is the case for legacy algorithms).
> > > 
> > > There seems to be an inconsistency, here, with what you said above,
> > > "... ECDSA just signs/verifies a raw hash, and in fact it *must* be a
> > > raw hash for it to be secure."
> > 
> > There isn't an inconsistency.  ECDSA is among the algorithms where the caller is
> > expected to handle the hash.
> > 
> > It is confusing dealing with all these different signature algorithms.  I think
> > the right way to think about this is in terms of what *data* is being
> > signed/verified.  Currently the data is the full file contents.  I think it
> > needs to be made into an annotated hash, e.g. the struct I gave.
> > 
> > public_key_verify_signature() also needs to be fixed to support both types of
> > signature algorithms (caller-provided hash and internal hash) in a logical way.
> > Originally it only supported caller-provided hashes, but then SM2 support was
> > added and now it is super broken.
> 
> Eric, did you say you're working on fixes to address these problems?

Yes, I was working on some patches at
https://git.kernel.org/pub/scm/linux/kernel/git/ebiggers/linux.git/log/?h=wip-keys-fixes
but got distracted by other things, as well as finding the problems with SM2
which will take some time to decide what to do with.  I'll try to get some
patches ready, but there are a lot of things to fix.

> 
> Instead of using a flexible array, Vitaly suggested defining the hash
> as FS_VERITY_MAX_DIGEST_SIZE, so that it could be allocated temporarily
> on stack
> instead of kalloc.
> 
> As the above struct is not limited to fsverity, we could use
> MAX_HASH_DIGESTSIZE, if it was exported, but it isn't.  Would the
> following work for you?
> 
> /*
>  * IMA signature header version 3 disambiguates the data that is signed
> by
>  * indirectly signing the hash of this structure, containing either the
>  * fsverity_descriptor struct digest or, in the future, the traditional
> IMA
>  * file hash.
>  */
> struct ima_file_id {
>         __u8 is_fsverity;       /* set to 1 for IMA_VERITY_DIGSIG */
>         __u8 hash_algorithm;    /* Digest algorithm [enum hash_algo] */
> #ifdef __KERNEL__
>         __u8 hash[HASH_MAX_DIGESTSIZE];
> #else
>         __u8 hash[];
> #endif
> };

You could certainly declare a fixed-length struct, but only sign/verify the used
portion of it.  The fixed-length struct would just be an implementation detail.

Alternatively, you could always sign/verify a fixed-length struct, with any
shorter hashes zero-padded to HASH_MAX_DIGESTSIZE (64 bytes).  This would be
fine if you're confident that hashes longer than 64 bytes will never be used.
(They don't make sense cryptographically, but who knows.)

For future extensibility, you might want to call the 'is_fsverity' field
something like 'hash_type', so it would be like an enum rather than a bool.  I
just used 'is_fsverity' as a minimal example.

- Eric

Patch

diff --git a/security/integrity/ima/ima_api.c b/security/integrity/ima/ima_api.c
index 7505563315cb..4fe7bc99378a 100644
--- a/security/integrity/ima/ima_api.c
+++ b/security/integrity/ima/ima_api.c
@@ -14,6 +14,7 @@ 
 #include <linux/xattr.h>
 #include <linux/evm.h>
 #include <linux/iversion.h>
+#include <linux/fsverity.h>
 
 #include "ima.h"
 
@@ -200,6 +201,23 @@  int ima_get_action(struct user_namespace *mnt_userns, struct inode *inode,
 				allowed_algos);
 }
 
+static int ima_collect_verity_digest(struct integrity_iint_cache *iint,
+				     struct ima_digest_data *hash)
+{
+	u8 verity_digest[FS_VERITY_MAX_DIGEST_SIZE];
+	enum hash_algo verity_alg;
+	int rc;
+
+	rc = fsverity_collect_digest(iint->inode, verity_digest, &verity_alg);
+	if (rc)
+		return -EINVAL;
+	if (hash->algo != verity_alg)
+		return -EINVAL;
+	hash->length = hash_digest_size[verity_alg];
+	memcpy(hash->digest, verity_digest, hash->length);
+	return 0;
+}
+
 /*
  * ima_collect_measurement - collect file measurement
  *
@@ -251,6 +269,8 @@  int ima_collect_measurement(struct integrity_iint_cache *iint,
 
 	if (buf)
 		result = ima_calc_buffer_hash(buf, size, &hash.hdr);
+	else if (veritysig)
+		result = ima_collect_verity_digest(iint, &hash.hdr);
 	else
 		result = ima_calc_file_hash(file, &hash.hdr);
 
diff --git a/security/integrity/ima/ima_appraise.c b/security/integrity/ima/ima_appraise.c
index 549fe051269a..53938aa0497a 100644
--- a/security/integrity/ima/ima_appraise.c
+++ b/security/integrity/ima/ima_appraise.c
@@ -240,6 +240,11 @@  static int xattr_verify(enum ima_hooks func, struct integrity_iint_cache *iint,
 			struct evm_ima_xattr_data *xattr_value, int xattr_len,
 			enum integrity_status *status, const char **cause)
 {
+	u8 verity_digest[IMA_MAX_DIGEST_SIZE + 1];
+	struct {
+		struct ima_digest_data hdr;
+		char digest[IMA_MAX_DIGEST_SIZE];
+	} hash;
 	int rc = -EINVAL, hash_start = 0;
 
 	switch (xattr_value->type) {
@@ -277,7 +282,45 @@  static int xattr_verify(enum ima_hooks func, struct integrity_iint_cache *iint,
 		*status = INTEGRITY_PASS;
 		break;
 	case IMA_VERITY_DIGSIG:
-		fallthrough;
+		set_bit(IMA_DIGSIG, &iint->atomic_flags);
+
+		/*
+		 * The IMA signature is based on a hash of IMA_VERITY_DIGSIG
+		 * and the fs-verity file digest, not directly on the
+		 * fs-verity file digest.  Both digests should probably be
+		 * included in the IMA measurement list, but for now this
+		 * digest is only used for verifying the IMA signature.
+		 */
+		verity_digest[0] = IMA_VERITY_DIGSIG;
+		memcpy(verity_digest + 1, iint->ima_hash->digest,
+		       iint->ima_hash->length);
+
+		hash.hdr.algo = iint->ima_hash->algo;
+		hash.hdr.length = iint->ima_hash->length;
+
+		rc = ima_calc_buffer_hash(verity_digest,
+					  iint->ima_hash->length + 1,
+					  &hash.hdr);
+		if (rc) {
+			*cause = "verity-hashing-error";
+			*status = INTEGRITY_FAIL;
+			break;
+		}
+
+		rc = integrity_digsig_verify(INTEGRITY_KEYRING_IMA,
+					     (const char *)xattr_value,
+					      xattr_len,
+					      hash.hdr.digest,
+					      hash.hdr.length);
+
+		if (rc) {
+			*cause = "invalid-verity-signature";
+			*status = INTEGRITY_FAIL;
+		} else {
+			*status = INTEGRITY_PASS;
+		}
+
+		break;
 	case EVM_IMA_XATTR_DIGSIG:
 		set_bit(IMA_DIGSIG, &iint->atomic_flags);
 		rc = integrity_digsig_verify(INTEGRITY_KEYRING_IMA,