Re: [GIT PULL] Security subsystem updates for 4.14

[Kees Cook <keescook@chromium.org>]

On Sat, Sep 9, 2017 at 9:32 PM, James Morris <jmorris@namei.org> wrote:
> On Fri, 8 Sep 2017, Paul Moore wrote:
>
>> > This is also why I tend to prefer getting multiple branches for
>> > independent things.
>
> [...]
>
>>
>> Is it time to start sending pull request for each LSM and thing under
>> security/ directly?  I'm not sure I have a strong preference either
>> way, I just don't want to see the SELinux changes ignored during the
>> merge window.
>
> They won't be ignored, we just need to get this issue resolved now and
> figure out how to implement multiple branches in the security tree.
>
> Looking at other git repos, the x86 folk have multiple branches.

Yeah, the x86 approach is what inspired my tree layout.

> One option for me would be to publish the trees I pull from as branches
> along side mine, with 'next' being a merge of all of directly applied
> patchsets and those ready for Linus to pull as one.
>
> So, branches in
> git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security
>
> might be:
>
>   next-selinux         (Paul's next branch)
>   next-apparmor   (JJ's next branch)
>   next-integrity  (Mimi's)
>   next-tpm        (Jarkko's)
>   [etc.]
>
>   next                 (merge all of the above to here)
>
> That way, we have a coherent 'next' branch for people to develop against
> and to push to Linus, but he can pull individual branches feeding into it
> if something is broken in one of them.
>
> Does that sound useful?

This is what I do with the KSPP tree (since it has a few unrelated
things in it), but you run the risk of getting too fine-grain and
creating dependencies between trees (e.g. adding a new hook that two
LSMs implement means either they depend on each other or both depend
on some third "core" tree).

How separable are the patches, normally?

-Kees

-- 
Kees Cook
Pixel Security