Re: Cryptsetup completion

[Daniel Shahaf <d.s@xxxxxxxxxxxxxxxxxx>]

Oliver Kiddle wrote on Thu, 04 Jun 2020 13:50 +0200:
> Paul Ruane wrote:
> > The new patch--in correct format :)--is here:  
> 
> Thanks
> 
> > +++ b/Completion/Linux/Type/_crypto_ciphers  
> 
> These are just the cyphers handled by the Linux kernel, right? Is this
> naming perhaps too generic? _linux_ciphers perhaps?

+1

> We wouldn't want
> them to be used in other cases where something else such as openssl
> ciphers are wanted. Does anyone know for what other commands these might
> be relevant? If there are none, then I would contradict Daniel about
> factoring them out.

For the record, my rationale was simply that a function that completes
"'cipher' elements of /proc/crypto" seemed like it would be inherently
reusable.  Nevertheless, I'm not aware of a specific place where it
would be reused, and I wouldn't mind one way or the other whether it
should be factored out or not.

Separate issue: cryptsetup(8) says:
.
       --cipher, -c <cipher-spec>
	      … The current default in the distributed sources is
	      "aes-cbc-essiv:sha256" for plain dm-crypt and
	      "aes-xts-plain64" for LUKS. …
.
So it seems _linux_ciphers should not be used directly as _the_
completion function for arguments to the -c option.

Cheers,

Daniel