Re: [PATCH] crypto: lrw - Only add ecb if it is not already there

[Eric Biggers <ebiggers@kernel.org>]

On Fri, May 16, 2025 at 05:15:20PM +0800, Herbert Xu wrote:
> On Thu, May 15, 2025 at 01:04:55PM -0700, Eric Biggers wrote:
> >
> > It didn't actually make a difference until 795f85fca229 ("crypto: algboss - Pass
> > instance creation error up") though, right?  Before then, if "ecb(...)" gave
> > ENOENT then "ecb(ecb(...))" gave ENOENT too.
> 
> I would consider this a success, it actually caught lrw doing
> something silly by trying to allocate "ecb(ecb(XXX))".  Sure we
> can sweep all of this under the rug with ENOENT but it might end
> up blowing up in a much worse place.
> 
> Also I don't think it's true that ecb(ecb(XXX)) will always give an
> ENOENT if ecb(XXX) gives an ENOENT.  The error is actually originating
> from the lskcipher code, which tries to construct ecb from either an
> lskcipher, or a simple cipher.  It is the latter that triggers the
> EINVAL error since the ecb template cannot create a simple cipher.
> 
> > As I said in
> > https://lore.kernel.org/linux-crypto/20240924222839.GC1585@sol.localdomain/,
> > that commit (which had no explanation) just seems wrong.  We should have simply
> > stuck with ENOENT.
> > 
> > But as usual my concern just got ignored and it got pushed out anyway.
> 
> I don't ignore all your concerns, just the ones that I think are
> unwarranted :)

I don't think this one is unwarranted.  We should keep the error codes simple
and just use ENOENT for algorithms that don't exist.  This also affected AF_ALG.

- Eric