Re: [PATCH 0/2] ARM: sunxi: Convert DTSI to new CPU bindings

[Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>]

On Sat, Jun 29, 2013 at 07:07:30PM +0100, Maxime Ripard wrote:
> On Fri, Jun 28, 2013 at 10:45:12PM +0100, Lorenzo Pieralisi wrote:
> > On Fri, Jun 28, 2013 at 09:05:42PM +0100, Olof Johansson wrote:
> > > On Fri, Jun 28, 2013 at 1:03 PM, Maxime Ripard
> > > <maxime.ripard@free-electrons.com> wrote:
> > > > On Fri, Jun 28, 2013 at 06:15:32PM +0100, Lorenzo Pieralisi wrote:
> > > >> On Fri, Jun 28, 2013 at 05:44:02PM +0100, Maxime Ripard wrote:
> > > >> > Hi Arnd, Olof,
> > > >> >
> > > >> > Sorry for steping in so late, but these two patches fixes a warning
> > > >> > introduced by the commit 1ba9bf0a (ARM: 7762/1: kernel: fix
> > > >> > arm_dt_init_cpu_maps() to skip non-cpu nodes) after 3.10-rc7.
> > > >> >
> > > >> > These two patches solve the situation for sunxi. The first one by
> > > >> > Lorenzo is to be applied for 3.10 if it's still possible, and the
> > > >> > second one on top of arm-soc/for-next. Tell me if you prefer a more
> > > >> > formal pull request for these patches
> > > >> >
> > > >> > Thanks!
> > > >> > Maxime
> > > >> >
> > > >> > Lorenzo Pieralisi (1):
> > > >> >   ARM: dts: sunxi: cpus/cpu nodes dts updates
> > > >>
> > > >> The patch above should already be queued in next/dt right ?
> > > >
> > > > Indeed.
> > > >
> > > > Then why the latest patch of your patchset got in 3.10, while the
> > > > patches actually fixing the DT it would have impacted were delayed to
> > > > 3.11?
> > > >
> > > > (And why was it merged so late in the development cycle?)
> > > 
> > > This. So now we have to scramble because some device trees will
> > > produce warnings at boot.
> > > 
> > > Russell, the alternative is to revert Lorenzo's patch for 3.10 (and
> > > re-introduce it for 3.11). Do you have a preference?
> > 
> > Ok, sorry about this guys. Technically speaking the DT bindings updates
> > (cpu/cpus nodes), dts updates and the ARM 7762/1 merged through Russell's
> > tree all are fixes, but probably we should not get them in as such.
> > 
> > ARM: 7762/1 was implemented to fix the warnings caused by new dts with
> > topology nodes (cpu-map node), and should go to stable kernels as well
> > since we want those kernels to boot with new DTs.
> > 
> > I should have prevented it from getting in as a fix, I really apologize.
> > I should also ask to drop the patch from the stable kernel queues, since this
> > would cause further issues (basically we should send all dts updates to
> > stable kernels as well, and unfortunately that's something we will have
> > to do anyway, when it has to be decided, if 7762/1 goes in 3.11 or later with
> > CC'ed stable we should send the dts updates to stable as well at the same
> > time).
> > 
> > I think the best solution is to revert ARM 7762/1 now and re-introduce it as
> > a fix after 3.11, when the dust has settled, I will drop it from stable
> > kernels queue as well if we all agree.
> > 
> > If we drop ARM 7762/1, 7764/1 needs rebasing since those patches order
> > unfortunately matters.
> 
> Ok, so you're saying we need to revert 1ba9bf0a and 7764/1. I couldn't
> find a commit corresponding to this one, however there is a commit
> labelled 7763/1 from you that touch the same area (18d7f152). Are these
> the commits that need to be reverted (and then reapplied for 3.11) ?

7764/1 is in Russell's patch system and has not been applied yet.
7763/1 can stay in since it is a fix that cleans-up cpu_logical_map
initialization.

1ba9bf0a (7762/1) is the patch that should be reverted (or we merge all
dts updates, since as I said, after all dts updates are fixes as well,
the warning is caused by the check for /cpu device_type to be == "cpu".
Those dts in the kernel are non-compliant, and have never been).

If 7762/1 is reverted, 7764/1 in Russell's patch system does not apply
cleanly anymore. Hence, if 7762/1 is reverted, we also have to drop 7764/1
from the patch system and I have to update it.

I asked to remove 7762/1 from stable kernels queues, the question is
still open if we want to backport all dts updates to stable kernels or
not. I would say no for now, hence 7762/1 can just be dropped (it was
meant to deal with the cpus node having children that are != /cpu. This
can happen with the new topology bindings, but let me say we will cross
that bridge when we come to it, 7762/1 has already caused too much
trouble for what it is worth).

On top of that there are bindings updates that I asked Grant to get them
merged for 3.11.

I am really really sorry for taking your time for this, but it has not been
easy to clean all dts up and define a proper standard, it is a necessary
evil.

Thanks a lot guys, please let me know how I can help, apologies.
Lorenzo