On Sun, Mar 28, 2021 at 9:04 PM Niels Möller nisse@lysator.liu.se wrote:
Next, I think it makes sense to start with adding the basic aes functions.
I'll prepare a MR for the basic aes functions to the s390x branch.
From a quick look at the MR, it seems the aes instructions don't want any explicit key schedule with expanded subkeys, but wants the raw key? Same for encrypt and decrypt?
Correct, both encrypt and decrypt operations just need the raw key, the key schedule is executed internally.
It would make sense to me with one file each under s390x/msa_x1 for the functions being replaced, but then the current aes-encrypt.c would also need to be split accordingly.
Splitting key functions and ciphering functions for the basic AES implementation is reasonable but splitting encrypting and decrypting functions are a lot of files considering how many functions are implemented for S390x arch. I don't get why aes-encrypt.c need to be split accordingly tho, I think it would work just fine.
Regards, Mamone