Added asymmetric encrypt and decrypt to Mbed Crypto provider #196

sbailey-arm · 2020-07-06T16:28:05Z

Signed-off-by: Samuel Bailey samuel.bailey@arm.com

sbailey-arm · 2020-07-06T16:30:34Z

This PR won't pass CI until parsec-interface-rs and parsec-client-rust have been updated on creates.io. Will mark as ready for review once they have been updated 😃.

ionut-arm

Looking good!

ionut-arm · 2020-07-06T22:53:27Z

CONTRIBUTORS.md

@@ -12,3 +12,4 @@ This file aims to acknowledge the specific contributors referred to in the "Cont
 * Ionut Mihalcea (@ionut-arm)
 * Hugues de Valon (@hug-dev)
 * Jesper Brynolf (@Superhepper)
+* Samuel Bailey (@sbailey-arm)


ionut-arm · 2020-07-06T22:54:27Z

Cargo.toml

 zeroize = { version = "1.1.0", features = ["zeroize_derive"] }
+secrecy = { version = "0.6.0", features = ["serde"] }


Is this needed? I didn't see any direct uses of it

Nope, I have removed it. Wonder if there's a clippy setting to catch that...

WIP it seems

That PR only provides the functionality of cargo-udeps in upstream tools. You can install it today!

Thanks @est31! We should definitely add that on our CIs!

ionut-arm · 2020-07-06T22:58:52Z

src/providers/mbed_provider/asym_encryption.rs

+        let salt_buff = match &op.salt {
+            Some(salt) => Some(salt.as_slice()),
+            None => None,
+        };


Is this what you're trying to achieve here?

I don't think that's quite it. op.salt is an Option<Zeroizing<Vec<u8>>> and I need salt_buff to be Option<&[u8]>. There may still be some more concise way of writing it that I'm not aware of though.

Huh! I thought that since Zeroizing implements Deref, it would work - actually, it seems as_deref is what I should've recommended, though it doesn't solve the whole problem either - op.salt.as_deref() yields Option<&Vec<u8>>

ionut-arm · 2020-07-06T23:02:23Z

src/providers/mbed_provider/asym_encryption.rs

+            }
+            Err(error) => {
+                let error = ResponseStatus::from(error);
+                format_error!("Encrypt status: {}", error);


The format_error macro doesn't need the {} part of the "format string", that gets added inside the macro!

Ah, that's why the log outputs have been a bit odd...!

What about in this case?

format_error!( "Error {} when opening a persistent Mbed Crypto key.", e );

Ooops! I didn't notice that one when it went in... I wrote format_error to always log in the same format "Brief description of problem; error: {}" - the last bit is only written when the config flag for verbose logging is on. You can probably change that to ("Failed to open persistent Mbed Crypto key", e)

ionut-arm · 2020-07-06T23:05:14Z

src/providers/mbed_provider/asym_encryption.rs

+            }
+            Err(error) => {
+                // When dropped, will zero buffer incase anything was written before error
+                let _ = Secret::new(plaintext);


Easier/quicker to just plaintext.zeroize() here (though you need to import the Zeroize trait).

huh, I didn't know you could do that 🤯! How does it work? plaintext is a regular Vec<u8>' and I can't see anything obvious in the documentation that states Zeroizableis implemented forVec. Is it u8implementsDefaultIsZeros, so implements Zeroizablein the blanket implementation andVec` 'inherits' (derefs?) that?

Zeroize is implemented on Vecs - it's actually what you were using in the other PR when doing plaintext.zeroize() for the protobuf types - those types don't use Zeroizing or anything from the crate

Oh yeah... 🤦‍♂️

ionut-arm · 2020-07-06T23:11:08Z

e2e_tests/tests/per_provider/normal_tests/asym_encryption.rs

+fn asym_encrypt_verify_decrypt_with_rsa_crate() {
+    let key_name = String::from("asym_encrypt_verify_decrypt_with_rsa_crate");
+    let mut client = TestClient::new();
+
+    client.generate_rsa_encryption_keys_rsapkcs1v15crypt(key_name.clone()).unwrap();
+    let pub_key = client.export_public_key(key_name.clone()).unwrap();
+
+    let rsa_pub_key = RSAPublicKey::from_pkcs1(&pub_key).unwrap();
+    let ciphertext = rsa_pub_key.encrypt(&mut OsRng, PaddingScheme::new_pkcs1v15_encrypt(), &PLAINTEXT_MESSAGE).unwrap();
+
+    let plaintext = client.asymmetric_decrypt_message_with_rsapkcs1v15(
+        key_name.clone(),
+        ciphertext,
+    ).unwrap();
+
+    assert_eq!(&PLAINTEXT_MESSAGE[..], &plaintext[..]);
+
+}


🙆‍♂️ niiiiiice!

sbailey-arm · 2020-07-07T14:38:33Z

e2e_tests/tests/per_provider/normal_tests/asym_encryption.rs

+    if client.provider() != Some(ProviderID::MbedCrypto) {
+        // Not supported by current provider
+        return;
+    }


Is there a neater way of doing this?
Would be nice if the tests could check which operations the current provider supports and only test those.

You can always put an #[cfg(feature = "mbed-crypto-provider")] on the test

Ah, this is in the e2e crate... Hmmm

I don't think the if is 100% correct anyway, because you could have multiple providers set up and the client chooses a different one, doesn't mean the Mbed Crypto one isn't available. Maybe we should change the set_provider method (on the test client) to check if the provider actually exists, and returns an error if not. Then we can check on that.

client chooses a different one

Are the tests not done in isolation? As in, there will only ever be one provider at once during testing?

Maybe we should change the set_provider method (on the test client) to check if the provider actually exists, and returns an error if not. Then we can check on that.

So have set_provider check basic_client.list_providers to see if that provider exists (in this case, MbedCrypto) and if it does, it returns Ok and we can run the tests?

Are the tests not done in isolation? As in, there will only ever be one provider at once during testing?

Yeah, that's true, this is in the per_provider test thing - forget that.

I think we can do it in another way, actually, makes it more scalable - call list_opcodes and check if the operation you want is supported. If it isn't, return. It's not ideal because it will still appear as a test even when it's not supported, but there's no need to do anything to enable the test when you implement it in some new provider.

I think we can do it in another way, actually, makes it more scalable - call list_opcodes and check if the operation you want is supported. If it isn't, return. It's not ideal because it will still appear as a test even when it's not supported, but there's no need to do anything to enable the test when you implement it in some new provider.

That sounds more like what I was thinking. The other two solutions don't deal with the fact that the tests are still run either. Could do with a way of forcing text output even if --show-output isn't specified, so we can at least say 'Hey, this provider doesn't actually support this operation'. I can have a poke tomorrow.

Yes, it might be worth to change this check to:

if client.list_opcodes(client.provider().unwrap())?.contains(Opcode::PsaAsymmetricEncrypt) { return; }

We could even create a method out of that snippet in the client:

fn is_operation_supported(&self, op: Opcode) -> bool { }

hug-dev

Thanks a lot! That looks very good. A few comments but otherwise it's ok.

hug-dev · 2020-07-09T08:34:42Z

CONTRIBUTORS.md

@@ -12,3 +12,4 @@ This file aims to acknowledge the specific contributors referred to in the "Cont
 * Ionut Mihalcea (@ionut-arm)
 * Hugues de Valon (@hug-dev)
 * Jesper Brynolf (@Superhepper)
+* Samuel Bailey (@sbailey-arm)


hug-dev · 2020-07-09T08:53:42Z

e2e_tests/tests/per_provider/normal_tests/asym_encryption.rs

+    if client.provider() != Some(ProviderID::MbedCrypto) {
+        // Not supported by current provider
+        return;
+    }


Yes, it might be worth to change this check to:

if client.list_opcodes(client.provider().unwrap())?.contains(Opcode::PsaAsymmetricEncrypt) { return; }

We could even create a method out of that snippet in the client:

fn is_operation_supported(&self, op: Opcode) -> bool { }

hug-dev · 2020-07-09T09:02:37Z

src/providers/mbed_provider/asym_encryption.rs

+        app_name: ApplicationName,
+        op: psa_asymmetric_encrypt::Operation,
+    ) -> Result<psa_asymmetric_encrypt::Result> {
+        info!("Mbed Provider - Asym Encrypt");


I think those info! are not needed now that we have the ingress/egress trace everywhere

Should they be removed from the other operations too?

If you don't mind 👀

hug-dev · 2020-07-09T09:57:46Z

src/providers/mbed_provider/asym_encryption.rs

+                })
+            }
+            Err(error) => {
+                plaintext.zeroize();


Ah good point! Do you think, even in case of error, it might contain some decrypted data? Does the spec say anything about it?

hmm it doesn't mention it in the spec that I can see. A quick look at the mbedtls code shows that it only copies into plaintext buffer if successful.

Yes I reckon that is the safe, default, behaviour that you would expect. We can remove the zeroize here I think.

Signed-off-by: Samuel Bailey <samuel.bailey@arm.com>

hug-dev

Thanks for adressing the comments!

hug-dev · 2020-07-09T12:29:09Z

src/providers/pkcs11_provider/asym_sign.rs

@@ -25,8 +25,6 @@ impl Pkcs11Provider {
        app_name: ApplicationName,
        op: psa_sign_hash::Operation,
    ) -> Result<psa_sign_hash::Result> {
-        info!("Pkcs11 Provider - Asym Sign");


Nice! Thanks for removing those as well.

ionut-arm reviewed Jul 6, 2020

View reviewed changes

sbailey-arm force-pushed the add-asym-encryption branch from 95fc26b to 79a3a14 Compare July 7, 2020 10:47

sbailey-arm mentioned this pull request Jul 7, 2020

Added asymmetric encrypt and decrypt parallaxsecond/parsec-client-rust#36

Merged

sbailey-arm force-pushed the add-asym-encryption branch 3 times, most recently from a678fcf to 163bbd6 Compare July 7, 2020 14:14

sbailey-arm marked this pull request as ready for review July 7, 2020 14:37

sbailey-arm commented Jul 7, 2020

View reviewed changes

sbailey-arm force-pushed the add-asym-encryption branch 3 times, most recently from 1f9f841 to 4b2fa81 Compare July 9, 2020 09:42

hug-dev reviewed Jul 9, 2020

View reviewed changes

sbailey-arm force-pushed the add-asym-encryption branch 2 times, most recently from 44f6795 to e45d0e0 Compare July 9, 2020 11:11

Added asymmetric encrypt and decrypt to Mbed Crypto provider

599d8d1

Signed-off-by: Samuel Bailey <samuel.bailey@arm.com>

sbailey-arm force-pushed the add-asym-encryption branch from e45d0e0 to 599d8d1 Compare July 9, 2020 11:32

hug-dev approved these changes Jul 9, 2020

View reviewed changes

hug-dev merged commit 0c61762 into parallaxsecond:master Jul 9, 2020

		zeroize = { version = "1.1.0", features = ["zeroize_derive"] }
		secrecy = { version = "0.6.0", features = ["serde"] }

Added asymmetric encrypt and decrypt to Mbed Crypto provider #196

Added asymmetric encrypt and decrypt to Mbed Crypto provider #196

Conversation

sbailey-arm commented Jul 6, 2020

sbailey-arm commented Jul 6, 2020

ionut-arm left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ionut-arm Jul 7, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hug-dev left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hug-dev left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ionut-arm Jul 7, 2020 •

edited

Loading