Just a quick note: I don't know how many generations you need to remove
the statistical anamolies, but unless you "try" the same random numbers
as starting points, you might get quite different performance results on
two separate key generations. The problem is that if you get lucky, you
will get a prime number within the first couple of tries. But you could
get quite unlucky and have to run a lot more tests. (Perhaps you could
record the number of failed tries and divide the total run time by
number of failed tries? That would provide some leveling to ensure a
fair comparision.)
Possibly 100 key pairs is enough to remove statistical variance. But
I'm not sure.
There's some general rule about the distribution of prime numbers, but I
don't remember what it is.
-- Garrett
Mark Powers wrote:
> Garrett,
>
> Yes the "is prime" test is hardware accelerated.
> A micro benchmark test that generated 100 RSA key pairs
> ran 10 times faster than when run in softtoken.
> I was generating keys with 1024 bit modulus.
>
> Of course applications may not always get these
> numbers but their RSA key generation should be faster.
>
>
> Mark
>
> Garrett D'Amore wrote:
>> It can significantly accelerate key generation? Wow, that sounds
>> interesting. Is this something more than just providing good random
>> data? Does it accelerate the "is prime" test in hardware?
>>
>> -- Garrett
>>
>> Darren J Moffat wrote:
>>
>>> Subject: PSARC FastTrack [02/13/2007]: key generation and derivation
>>> for providers without keystore
>>>
>>> I'm submitting this case as closed approved automatic. It only
>>> modifies
>>> Consolidation Private or Project Private interfaces and does so in a
>>> compatible way.
>>>
>>> Template Version: @(#)sac_nextcase %I% %G% SMI
>>> 1. Introduction
>>> 1.1 Project/Component Working Name:
>>> key generation and derivation for providers without keystore
>>> 1.2 Name of Document Author/Supplier:
>>> Mark Powers, Anthony Scarpino
>>> 1.3 Date of this Document:
>>> 2/12/2007
>>>
>>> 4. Technical Description
>>>
>>> The Interface taxonomy is Consolidation Private.
>>> The Release taxonomy is Patch/Micro
>>>
>>> Diff-marked draft manpages, cspi design, and header files are
>>> included
>>> in the case directory.
>>>
>>> 4.0 Background
>>>
>>> The ncp provider, which is available on sun4v machines, is capable
>>> of significantly accelerating RSA and Diffie-Hellman (DH) key
>>> generation
>>> and derivation. However, it is not possible to take advantage of
>>> this
>>> because the existing Service Provider Interface (SPI) assumes
>>> providers
>>> capable of generating and deriving keys also have a keystore, and
>>> that
>>> keys are returned by reference rather than value. This proposal
>>> seeks
>>> to remove this limitation.
>>>
>>> 4.1 Proposal
>>>
>>> The case proposes the following:
>>>
>>> - modifications to the kcf kernel module to support new provider
>>> entry points for the generation and derivation of keys. The
>>> new
>>> entry points have two arguments per key instead of one. The
>>> extra
>>> argument is reserved for attributes that need to be copied
>>> back
>>> into a user's address space. Providers that register the new
>>> entry points should not register key ops (crypto_key_ops(9S))
>>> which are reserved for providers with keystores. Providers
>>> with
>>> these new entry points register SPI version 3.
>>>
>>> - modifications to pkcs11_kernel - a user library component -
>>> to use
>>> the new key generation and derivation entry points. The
>>> gist of
>>> this modification is to add additional PKCS#11 RSA and DH
>>> attributes
>>> to key templates so that generated and derived keys can be
>>> returned
>>> by value rather than by reference. Generated and derived
>>> keys will
>>> be stored as session objects by pkcs11_kernel.so.
>>>
>>> - modifications to metaslot and softtoken to move session keys
>>> from pkcs11_kernel to pkcs11_softtoken where they can be
>>> stored
>>> as token keys. Token keys are persistent objects whereas
>>> session
>>> keys are destroyed when sessions are closed.
>>>
>>> - modifications to the ncp driver to use version 3 of the SPI.
>>> The driver will be modified to support
>>> CKM_RSA_PKCS_KEY_PAIR_GEN,
>>> CKM_DH_PKCS_KEY_PAIR_GEN, and CKM_DH_PKCS_DERIVE.
>>>
>>> The new SPI entry points (see spi.h) are:
>>> int (*nostore_key_generate)(crypto_provider_handle_t,
>>> crypto_session_id_t, crypto_mechanism_t *,
>>> crypto_object_attribute_t *, uint_t,
>>> crypto_object_attribute_t *,
>>> uint_t, crypto_req_handle_t);
>>>
>>> int (*nostore_key_generate_pair)(crypto_provider_handle_t,
>>> crypto_session_id_t, crypto_mechanism_t *,
>>> crypto_object_attribute_t *, uint_t,
>>> crypto_object_attribute_t *,
>>> uint_t, crypto_object_attribute_t *, uint_t,
>>> crypto_object_attribute_t *, uint_t, crypto_req_handle_t);
>>>
>>> int (*nostore_key_derive)(crypto_provider_handle_t,
>>> crypto_session_id_t,
>>> crypto_mechanism_t *, crypto_key_t *,
>>> crypto_object_attribute_t *,
>>> uint_t, crypto_object_attribute_t *, uint_t,
>>> crypto_req_handle_t);
>>>
>>> 4.1.1 Compatibility
>>>
>>> The SPI interface was introduced by PSARC 2001/553. PSARC
>>> 2005/576
>>> changed the interface version from 1 to 2. The kernel kcf
>>> module will
>>> remain backwards compatible with version 1 and 2 providers.
>>>
>>> 4.2 Bug/RFE Number(s): 6339802, 6462782, 6462780
>>>
>>> 4.5 Exported Interfaces:
>>>
>>> +--------------------------------+---------------------------+--------------+
>>>
>>> | Interface | Classification |
>>> Comments |
>>> +--------------------------------+---------------------------+--------------+
>>>
>>> | impl.h | Consolidation |
>>> include file |
>>> | | Private
>>> | |
>>> | |
>>> | |
>>> | ioctl.h | Project Private |
>>> include file |
>>> | |
>>> | |
>>> | ops_impl.h | Consolidation |
>>> include file |
>>> | | Private
>>> | |
>>> | |
>>> | |
>>> | spi.h | Consolidation |
>>> include file |
>>> | | Private
>>> | |
>>> | |
>>> | |
>>> | ncp.h | Consolidation |
>>> include file |
>>> | | Private
>>> | |
>>> +--------------------------------+---------------------------+--------------+
>>>
>>>
>>> 6. Resources and Schedule
>>> 6.4. Steering Committee requested information
>>> 6.4.1. Consolidation C-team Name:
>>> ON
>>> 6.5. ARC review type: FastTrack
>>> _______________________________________________
>>> crypto-discuss mailing list
>>> crypto-discuss at opensolaris.org
>>> http://opensolaris.org/mailman/listinfo/crypto-discuss
>>>
>>
>>
>>
--
Garrett D'Amore, Principal Software Engineer
Tadpole Computer / Computing Technologies Division,
General Dynamics C4 Systems
http://www.tadpolecomputer.com/
Phone: 951 325-2134 Fax: 951 325-2191
