Am Mittwoch, 8. Juli 2015, 23:29:32 schrieb Herbert Xu:
Hi Herbert,
>On Wed, Jul 08, 2015 at 05:27:15PM +0200, Stephan Mueller wrote:
>> But according to Tadeusz it may be viable to register hybrid asym ciphers
>> with the akcipher API. If there is a full blown hybrid asym cipher we have
>> to handle potentially large chunks of data. I am now wondering whether a
>> flat buffer still makes sense or whether the akcipher should be converted
>> to scatter lists as all the other cipher types.
>>
>> If it shall stay as is, how would we be handling larger buffers?
>
>What is a hybrid cipher and why would we want to have it in the
>kernel?
Please consider the thread from June 25 between Tadeusz and me. With hybrid
cipher, I am referring to the asym + sym cipher combo to implement a "real"
complete asymmetric encryption/decryption. Or I am referring to the asym +
hash combo to implement a "real" signature generation/verification.
Tadeusz' rsa.c implements the raw RSA modular exponentiation operation. But
that does not make a usable encryption or signature operation. For example,
the kernel module signature verification illustrates that very nicely:
mod_verify_sig():
...
/* hash the code */
pks = mod_make_digest(ms.hash, mod, modlen);
...
/* RSA verify the hash */
ret = verify_signature(key, pks);
What I am wondering is that whether such hybrid operation (hash+asym or
sym+asym) may be entirely implemented in hardware. If you tell me that this
will never be the case, I will stop asking and would be happy with the plain
buffer. :-)
But IIRC, there are systems out there which perform a full hybrid asym
operation. Without having checked in detail, I believe this is true for IBM
cryptoexpress or Cavium cards.
The reason for having it in the kernel is what Tadeusz describes: the kernel
does not need asym operations. But the hardware is only accessible from kernel
space. With this, the kernel crypto API together with a yet to be written
AF_ALG akcipher would make the asym hardware available to user space.
Ciao
Stephan
--
To unsubscribe from this list: send the line "unsubscribe linux-crypto" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
