Re: [PATCH 5.10.y 1/3] net: ipv4: stop checking crypto_ahash_alignmask

Re: [PATCH 5.10.y 1/3] net: ipv4: stop checking crypto_ahash_alignmask

[Eric Biggers]

[+Cc linux-crypto@vger.kernel.org, netdev@vger.kernel.org,
     Steffen Klassert <steffen.klassert@secunet.com> ]

On Wed, May 13, 2026 at 01:15:52PM -0400, Sasha Levin wrote:
> From: Eric Biggers <ebiggers@google.com>
> 
> [ Upstream commit e77f5dd701381cef35b9ea8b6dea6e62c8a7f9f3 ]
> 
> Now that the alignmask for ahash and shash algorithms is always 0,
> crypto_ahash_alignmask() always returns 0 and will be removed.  In
> preparation for this, stop checking crypto_ahash_alignmask() in ah4.c.
> 
> Signed-off-by: Eric Biggers <ebiggers@google.com>
> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
> Stable-dep-of: ec54093e6a8f ("xfrm: ah: account for ESN high bits in async callbacks")
> Signed-off-by: Sasha Levin <sashal@kernel.org>

You forgot to send this to the IPsec maintainers, linux-crypto, and
netdev.  See Documentation/process/submitting-patches.rst for some tips
on how to use scripts/get_maintainer.pl to find the right place to send
kernel patches.

This commit was part of the series
https://lore.kernel.org/linux-crypto/20231022081100.123613-1-ebiggers@kernel.org/

What is your rationale for why it's safe to backport this patch 18 of a
30-patch series by itself?  "The alignmask for ahash and shash
algorithms is always 0" is definitely *not* true in older kernels.

I *think* it's probably okay.  I think this would just regress AH
performance in some cases, and "no one" uses AH anyway.

Just keep in mind this is effectively new development, which needs
review like any other kernel patch.  Not even Cc'ing the subsystem
mailing lists isn't a great approach.

- Eric