OP-TEE key

[antonio_santagi]

Hello,

I am using Yocto BSP for iMX8M mini. Not sure if this is the right place where to ask this question, in case not please let me know.

At https://source.codeaurora.org/external/imx/imx-optee-os/tree/documentation/porting_guidelines.md?h=i... 

I can read that 

## 9. Trusted Application private/public keypair
By default all Trusted Applications (TA's) are signed with the pre-generated
2048-bit RSA development key (private key). This key is located in the `keys`
folder (in the root of optee_os.git) and is named `default_ta.pem`. This key
**must** be replaced with your own key and you should **never ever** check-in
this private key in the source code tree when in use in a real product.  

and also at : 

https://optee.readthedocs.io/en/latest/building/trusted_applications.html#signing-of-tas 

Warning

"optee_os comes with a default private key in its source to facilitate easy development, testing, debugging and QA. Never deploy an optee_os binary with this key in production. Instead replace this key as soon as possible with a public key and keep the private part of the key offline, preferably on an HSM."

Right, then I generated a new keypair and extracted the public key. 

I have then put the extracted public key in the keys folder renaming it to default_ta.pem, overwriting the original default_ta.pem.

This means that there is only a public key now in default_ta.pem.

Now, optee-os-imx builds correctly but optee-test-imx raises error, this is understandable because it's a trusted app and therefore it needs the private key to be signed with.

So, my question is, if I replace the default_ta.pem with another KEYPAIR .pem file that I generated, is this secure ? optee-os-imx and optee-test-imx will build correctly but does the private key part of the key pair gets stripped from the optee-os binary file ? or does the default_ta.pem keypair gets embedded completely in the image ?

In this last case this would clash a bit with the instructions that suggest to not put any private key in the sources being built.

Can you point to the script that handles this part ?

thank you!

[imx8mp_developer]

Does anybody figure out what should be modified in order to sign using HSM in Yocto builds?

[imx8mp_developer]

I am planning to implement a secure storage on imx93 with Yocto, I followed this post and it's clear how to implement it, but two things are not clear:
- is Hardware Unique Key (HUK) already present into TEE? So I should do nothing for this

- if I don't need to add any TA I still need to modify the default_ta.pem? Is there any guide for this? (Key generation + replace in yocto)


I assume there isn't any way to extract the HUK so I expect that the data cannot be decripted on another hardware (embededd or laptop) not even for debug purpose, right?

[antonio_santagi · Accepted solution]

Yes, that's correct @IvanRuiz .

Only problem is that we build also imx-optee-test recipe and this requires optee-os and it tries to get the private key from the optee-os build folder. 

Anyway now I found that the private key is stripped out from the keypair before embedding into the tee image, it's done by this script : 

pem_to_pub_c.py

it generates a c file with information only about the public key extracted from the keypair.

that is what goes then in the tee.bin file.

We anyway will not commit the private key to the repository with the sources, as per your documentation. 

[imx8mp_developer]

Hello @antonio_santagi 
how did you solve the issue at the end? I am trying for imx93 to use a keypair generated as explained in docs and I am providing it to yocto using TA_SIGN_KEY and TA_PUBLIC_KEY env variables.
The values are correctly transferred to the do_compible jobs of optee-os and optee-os-tadevkit but when I run xtest on imx93 target I get

* regression_1039 Test subkey verification
o regression_1039.1 Load TA with two levels of subkeys
/usr/src/debug/optee-test/4.4.0.imx/host/xtest/regression_1000.c:3278: xtest_teec_open_session(&session, &subkey1_ta_uuid, ((void *)0), &ret_orig) has an unexpected value: 0xffff000f = TEEC_ERROR_SECURITY, expected 0x0 = TEEC_SUCCESS
 regression_1039.1 FAILED
o regression_1039.2 Load TA with identity subkey
/usr/src/debug/optee-test/4.4.0.imx/host/xtest/regression_1000.c:3285: xtest_teec_open_session(&session, &subkey2_ta_uuid, ((void *)0), &ret_orig) has an unexpected value: 0xffff000f = TEEC_ERROR_SECURITY, expected 0x0 = TEEC_SUCCESS
 regression_1039.2 FAILED
 regression_1039 FAILED

[antonio_santagi]

Hello @imx8mp_developer ,

Sorry, I haven't worked on this topic for a while.

As far as I can remember at that time, some years ago, it worked OK if I used the same default name for the keypair, replacing the original keypair. I can't remember if I tried with the two env variables you mentioned. You could try replacing the default keypair using the same name first and see if that works.

[imx8mp_developer]

Hello! Thanks for your answer,

in the meanwhile I managed to use the varialbles I mentioned directly inside Yocto and I guess it's working since I see the correct path being used in the do_compile log of optee-os and optee-tadevkit components.
I tried to launch "xtest" on the system and I see that test 1039 is failing, I assume it's because it internally uses intermediate keys which are pre-signed with the default_ta.pem key and not with the new one, by the way I think this is just an issue on xtest side and I assume the TA deployed by optee-os / opte-tadevkit will correctly work!

[IvanRuiz]

Hello,

For security reasons it is not recommended to keep any private keys in keys/default_ta.pem and the recommendation according to documentation is to use the generated public key only since that will only be used for the encryption and the private key shall be stored preferably in an HSM. The TAs are signed using the sign_encrypt.py from OP-TEE as mentioned in the documentation.

Hope it helps!

BR,

Ivan.