[PATCH] ARM: imx6: allow booting with old DT

[Lucas Stach]

Am Mittwoch, den 27.05.2015, 09:07 +0100 schrieb Marc Zyngier:
> On 27/05/15 08:52, Lucas Stach wrote:
> > Am Mittwoch, den 27.05.2015, 15:34 +0800 schrieb Shawn Guo:
> >> On Tue, May 26, 2015 at 06:43:36PM +0200, Lucas Stach wrote:
> >>> The GPC rewrite to IRQ domains has been on the premise that it may break
> >>> suspend/resume for new kernels on old DT, but otherwise keep things working
> >>> from a user perspective. This was an accepted compromise to be able to move
> >>> the GIC cleanup forward.
> >>>
> >>> What actually happened was that booting a new kernel on an old DT crashes
> >>> before even the console is up, so the user does not even see the warning
> >>> that the DT is too old. The warning message suggests that this has been
> >>> known before, which is clearly unacceptable.
> >>
> >> To see any early message like this one, low-level debug support is
> >> expected to be turned on.
> >>
> > Using low-level debug might be acceptable for a developer.
> > 
> > From a user perspective a kernel update without a corresponding DT
> > update should never render the machine completely broken. Keep in mind
> > that i.MX6 isn't only used in deeply embedded system, but is already in
> > devices where non-developer users might update the kernel. They might
> > even use prebuilt kernels where enabling low-level debug is not an
> > option.
> 
> I'd imagine that whoever provides this pre-build kernel would also
> deliver some form of release notes indicating the procedure. Even
> better, an installation script.
> 
> > We are not free to break the existing DT stability rules in such a
> > drastic way, especially if keeping things working to some extent is
> > easily done.
> 
> That would be on the condition that the DT was correct the first place,
> and accurately described the hardware. It didn't, breaking the contract
> we have the first place.
> 
> We can argue for years about DT stability, history proves that it
> doesn't lead anywhere.
> 
The fact that we are still in a learning phase with regard to DT and
have made mistakes in the past (and I'm sure we will still do some in
the future) isn't an excuse for not even trying to keep the stability.

In this case there was a clear consent that we break DT stability by
rendering suspend/resume broken. This was the accepted compromise
between DT stability and the highly needed GIC cleanup.
Completely breaking an entire SoC family in early boot was not part of
the the compromise! Especially as it isn't really hard to keep things
working to the specified extent, as this patch proves.

Regards,
Lucas
-- 
Pengutronix e.K.             | Lucas Stach                 |
Industrial Linux Solutions   | http://www.pengutronix.de/  |