RE: [PATCH V8 1/5] crypto: Multi-buffer encryption infrastructure support

-----Original Message-----
From: Herbert Xu [mailto:herbert@gondor.apana.org.au]
Sent: Monday, May 7, 2018 2:35 AM
To: Dey, Megha <redacted>
Cc: linux-kernel@vger.kernel.org; linux-crypto@vger.kernel.org;
davem@davemloft.net
Subject: Re: [PATCH V8 1/5] crypto: Multi-buffer encryption infrastructure
support

On Tue, May 01, 2018 at 10:39:15PM +0000, Dey, Megha wrote:

quoted

crypto/simd.c provides a simd_skcipher_create_compat. I have used the
same template to introduce simd_ahash_create_compat which would wrap

around the inner hash algorithm.

quoted

Hence we would still register 2 algs, outer and inner.

Right.

quoted

Currently we have outer_alg -> mcryptd alg -> inner_alg

Mcryptd is mainly providing the following:
1. Ensuring the lanes(8 in case of AVX2) are full before dispatching to the

lower inner algorithm. This is obviously why we would expect better
performance for multi-buffer as opposed to the present single-buffer
algorithms.

quoted

2. If there no new incoming jobs, issue a flush.
3. A glue layer which sends the correct pointers and completions.

If we get rid of mcryptd, these functions needs to be done by someone. Since

all multi-buffer algorithms would require this tasks, where do you suggest these
helpers live, if not the current mcryptd.c?

That's the issue.  I don't think mcryptd is doing any of these claimed functions
except for hosting the flush helper which could really live anywhere.

All these functions are actually being carried out in the inner algorithm already.

quoted

I am not sure if you are suggesting that we need to get rid of the mcryptd

work queue itself. In that case, we would need to execute in the context of the
job requesting the crypto transformation.

Which is fine as long as you can disable the FPU.  If not the simd wrapper will
defer the job to kthread context as required.

Hi Herbert,

Are you suggesting that the SIMD wrapper, will do what is currently being done by the ' mcryptd_queue_worker ' function (assuming FPU is not disabled) i.e dispatching the job to the inner algorithm?

I have got rid of the mcryptd layer( have an inner layer, outer SIMD layer, handled the pointers and completions accordingly), but still facing some issues after removing the per cpu mcryptd_cpu_queue.
 

Cheers,
--
Email: Herbert Xu [off-list ref] Home Page:
http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt