[v3 PATCH 00/13] Architecture-optimized SHA-256 library API

[Eric Biggers]

On Wed, Apr 30, 2025 at 10:27:15AM +0800, Herbert Xu wrote:
> On Tue, Apr 29, 2025 at 09:57:49AM -0700, Eric Biggers wrote:
> >
> > To be clear, the objections I have on your v2 patchset still hold.  Your
> > unsolicited changes to my patches add unnecessary complexity and redundancy,
> > make the crypto_shash API even harder to use correctly, and also break the build
> > for several architectures.  If you're going to again use your maintainer
> > privileges to push these out anyway over my objections, I'd appreciate it if you
> > at least made your dubious changes as incremental patches using your own
> > authorship so that they can be properly reviewed/blamed.
> 
> Well the main problem is that your patch introduces a regression
> in the shash sha256 code by making its export format differ from
> other shash sha256 implementations (e.g., padlock-sha).
> 
> So your first patch cannot stand as is.  What I could do is split up
> the first patch so that the lib/crypto sha stuff goes in by itself
> followed by a separate patch replacing the crypto/sha256 code.
> 
> > Please also note that I've sent a v4 which fixes the one real issue that my v1
> > patchset had: https://lore.kernel.org/r/20250428170040.423825-1-ebiggers@kernel.org
> 
> Yes I've seen it but it still has the same issue of changing the
> shash sha256 export format.

Nothing requires that the export formats be consistent, but also the fact that
padlock-sha uses a weird format in the first place is an artificial problem that
you introduced just a couple weeks ago.  And even if we *must* use the same
format as padlock-sha, that can be done by using your crypto_sha256_export_lib
and crypto_sha256_import_lib, without all your other changes.

- Eric