[PATCH v2 0/3] ARM: NEON based fast(er) AES in CBC/CTR/XTS modes

Nicolas Pitre nicolas.pitre at linaro.org
Fri Oct 4 14:39:29 EDT 2013

On Fri, 4 Oct 2013, Russell King - ARM Linux wrote:

> Also, remember that the GPL says:
> "The source code for a work means the preferred form of the work for
> making modifications to it."
> So here's the question: is the assembly code the perferred form to make
> modifications?  From what you're saying above, the answer to that seems
> to be no.  Now, I'm not going to throw toys out of the pram and say
> that this is a hard requirement, but just take a moment to think about
> how we treat vendors who don't do this, instead supplying "non-preferred"
> forms of source code, and think about whether there's double standards
> here.
> Now, while I can imagine that people have an ideological objection to
> perl (using comments like "write only code" etc) that's not a good
> enough excuse to avoid including the perferred form for future
> modification.
> Now, we have mechanisms in the kernel build where we can include a
> prepared source which can be used to lessen the burden on the toolset
> required to build the kernel.  So, including both the perl script and
> the pre-generated assembly is entirely acceptable.  This tends to
> nullify the excuse that "we don't want to add additional tool burden
> to kbuild" argument.

Those are very good arguments.  And I agree with your recommendation as 


More information about the linux-arm-kernel mailing list