Re: [PATCH] crypto: mark cts(cbc(aes)) as FIPS allowed

[Ard Biesheuvel]

On 5 November 2018 at 10:18, Gilad Ben-Yossef <gilad@xxxxxxxxxxxxx> wrote:
> Hi Stephan,
>
> On Mon, Nov 5, 2018 at 10:50 AM Stephan Mueller <smueller@xxxxxxxxxx> wrote:
>>
>> Am Sonntag, 4. November 2018, 11:05:24 CET schrieb Gilad Ben-Yossef:
>>
>> Hi Gilad,
>>
>> > As per Sp800-38A addendum from Oct 2010[1], cts(cbc(aes)) is
>> > allowed as a FIPS mode algorithm. Mark it as such.
>> >
>> > [1] https://csrc.nist.gov/publications/detail/sp/800-38a/addendum/final
>>
>> There are several types of CTS approaches. Only three of those are listed in
>> the SP800-38A addendum. The source code only refers to some RFCs.
>>
>> Did you check whether the CTS implementation matches one or more of the types
>> listed in the addendum? If yes, may I suggest to add a small statement in the
>> code noting this fact?
>
> Yes, AFAIK the software implements CBC-CS3 as described in the document.
> You have a very good point about this not document anywhere. I will
> send a patch.
>

I agree that it makes sense to document this.

> While we're at it, does someone find it useful if I implemented the
> other two modes?
> They are part of the standard as well, although I am not sure that in
> itself is sufficient cause.
>

Nope. The linux crypto API is not a general purpose cryptography
toolkit, it should only offer what we actually use in the kernel, or
we'll either end up with bitrot or with a disproportionate maintenance
burden for stuff nobody actually uses.