[kpatch] Preparing a patch - pitfalls

[Josh Poimboeuf]

On Tue, Sep 29, 2015 at 10:48:05AM +0300, Evgenii Shatokhin wrote:
> At the present time, perhaps, there are some drafts, random pieces of advice
> or may be collections of patches that needed additional analysis and
> rewriting? Such examples of the patches would be extremely useful even
> without explanations.

Unfortunately I don't have (or know of anything) like that yet.  But it
would be good to start something.

> I am asking because we are now considering using Kpatch at least for regular
> testing of some updates for the kernels in our systems. Then (who knows) our
> experts might decide to deliver security updates this way to our customers
> too. So, one must be extra careful indeed, when preparing the patches.
> Better to know the pitfalls earlier.

I agree and I think that documenting best practices of the patch
creation process will be very important for the future of kpatch (and
live patching in general).

As a first step, maybe we could start a collection of example converted
patches in the kpatch git repo.  We could put a description of the patch
analysis and conversion in the headers of the patches.  Once we have
enough such case studies, we could then look for some emergent higher
level patterns to document.

Care to open a pull request for the tcp_cubic patch and any others you
might want to discuss?  Perhaps we could put them in a new "examples"
dir.

-- 
Josh