Re: flash-kernel: handle DTBs in vendor subdir
On Sun, 2016-07-31 at 13:11 -0700, Martin Michlmayr wrote:
> > * Ian Campbell <ijc@debian.org> [2016-07-31 10:20]:
> >
> > I've attached the two patches I had sitting in my branch here, they
> > look sensible but I honestly can't remember what state they are in.
> >
> > AFAICT the main difference is that they preserve the directory layout
> > rather than collapsing it. Not sure there is any particular reason to
> > favour either way of doing it so feel free to either ignore or pickup
> > these patches.
>
> The boot script looks for dtbs/${fk_kvers}/${fdtfile} and $fdtfile
> doesn't contain the vendor subdir, so I think this approach is wrong.
>
> >
> > One thing my patches handles which I'm not sure yours does is the pre-
> > subdir-transition versions on ARM64, which my commit log says happened
> > in v3.19-rc1. That might be important for backports and/or upgrades.
>
> I looked at your other patch and I like the approach since it supports
> the old and new way. Personally I found search_for_dtb_file_in_prefix
> a bit hard to read and wonder if a simple 'find' might do.
>
> What do you think of this approach? This no longer requires the
> vendor subdir in DTB-Id. I think this is better since e.g. the RPi3
> has the same device tree on 32 and 64 bit and I assume one will be in
> a vendor subdir whereas the other won't be.
>
> > Any comments?
Seems fine by me. It does assume that DTB filenames are unique even
when the directory hierarchy. That seems fairly likely to be true, but
it may not be guaranteed I guess.
Reply to: