On 7/27/22 03:17, Herbert Xu wrote:
On Tue, Jul 26, 2022 at 09:15:54PM +0100, Dmitry Safonov wrote:
Add crypto_pool - an API for allocating per-CPU array of crypto requests
on slow-path (in sleep'able context) and to use them on a fast-path,
which is RX/TX for net/ users (or in any other bh-disabled users).
The design is based on the current implementations of md5sig_pool.
Previously, I've suggested to add such API on TCP-AO patch submission [1],
where Herbert kindly suggested to help with introducing new crypto API.
What I was suggesting is modifying the actual ahash interface so
that the tfm can be shared between different key users by moving
the key into the request object.
The fact that setkey is implemented at the crypto_ahash instead of the
ahash_request level is baked into all algorithm implementations
(including many hardware-specific ones). Changing this seems extremely
difficult.
Supporting setkey at the tfm level could be achieved by making it an
optional capability on a per-algorithm basis, then something like
crypto_pool could detect this scenario and avoid allocating a per-cpu
tfm. This would also require a crypto_pool_setkey wrapper.
As it stands right now multiple crypto-api users needs to duplicate
logic for allocating a percpu array of transforms so adding this "pool"
API is an useful step forward.
As far as I remember the requirement for a per-cpu scratch buffer is
based on weird architectures having limitations on what kind of memory
can be passed to crypto api so this will have to remain.
--
Regards,
Leonard
