WIP: apu2_trng: initial commit

[pietrushnic]

Signed-off-by: Piotr Król [email protected]

[pietrushnic]

@miczyg1 it would be great if at some point you could help me with this blog post.

[miczyg1]

@pietrushnic the topic looks very interesting. I will be glad to help. I guess we do not have support for PSP in our firmware? Have You tried building and booting anything with PSP?

[pietrushnic]

@miczyg1 at this point I don't think PSP is the problem. I think we should look in ccp driver why it's not working. Internally available documentation has all details related to TRNG exposed by CCP in PSP. If we found the problem we can continue discussion with maintainers and AMD.

[miczyg1]

@pietrushnic we should prepare a procedure of performing cryptographic operations and measuring entropy first. Or do we have one already?

I would like to follow our convention:
we build image -> prove feature does not work (TEST FAIL)
we debug and develop code -> build image -> prove feature works (TEST PASS)

[pietrushnic]

@miczyg1 agree, procedure can rely on various tools. I already described something in this blog post draft. It looks like reasonable options are /dev/hwrng using rng-tools and rngtest, maybe binwalk and there is dieharder. Comaprison of those tools is also good for blog post. I already proved that TPM is very slow but correctly generate enropy. On apu2 /dev/hwrng seems to be tied to ccp and generate all 0xff.

[miczyg1]

@pietrushnic dTPM is slow by the design, as written in the PR with fTPM vs dTPM. there may be a few reasons that ccp generates 0xFF. Buggy PSP firmware, wrong initialization of PSP or some register is not set to use RDRAND instruction. It would require investigation, but definitely, it is not impossible IMO.

[pietrushnic]

@miczyg1 let's schedule that. Probably not 01.2019 but 02.2019.

[pietrushnic]

Another TRNG AFAIK was fixed, but we didn't published anything about that. @miczyg1 would you mind to get back to this topic?

[miczyg1]

@pietrushnic yeah, as we fixed the CCP I may as well write/finish the post about TPM2 RNG and CCP RNG in one sway here ;)