Re: [PATCH V2] x86: Allow built-in command line to work in early kernel init

[Ingo Molnar <mingo@kernel.org>]

* Matthew Garrett <matthew.garrett@coreos.com> wrote:

> >> +             while (c) {
> >> +                     set_fs(new_cmdline_ptr >> 4);
> >> +                     wrfs8(c, nptr++);
> >> +                     set_fs(cmdline_ptr >> 4);
> >> +                     c = rdfs8(cptr++);
> >> +             }
> >> +             set_fs(new_cmdline_ptr >> 4);
> >> +             wrfs8(' ', nptr++);
> >> +     }
> >
> > So here we copy from 'cptr' to 'nptr'? Not very well named variables,
> > at minimum.
> 
> I have the choice between preserving existing style or nice names.
> I'm happy to pick either.

So my high level concerns (and conditions) are the following:

We should first do cleanups and reorganization so that the feature addition 
(allowing EFI to parse certain options early on) you are implementing becomes 
obvious and simple.

I.e. we need to first prove it's all maintainable and readable, and then only can 
we add a feature with such a signature:

 13 files changed, 236 insertions(+), 43 deletions(-)

I'd also expect that a good chunk of that linecount increase would disappear after 
a proper set of reorganization, because all that duplication would disappear 
mostly.

It's all pretty non-trivial, so it has to be finegrained, perfectly bisectable, 
etc.

Thanks,

	Ingo