[PATCH 0/3] of: dts: enable memory at 0 quirk for PPC32 only

[Grant Likely]

On Tue, 22 Apr 2014 15:05:26 +0100, Leif Lindholm <leif.lindholm at linaro.org> wrote:
> On Tue, Apr 22, 2014 at 02:08:29PM +0100, Grant Likely wrote:
> > > I would prefer to see a "WARN_ON(!IS_ENABLED(CONFIG_PPC32));" added here.
> > 
> > I completely agree.
> 
> OK. So do we keep this around on unaffected architectures in perpetuity?
> 
> Or can there be some cut-off date when the majority of DT-enabled
> platforms can stop including this workaround which permits new incorrect
> DTs to be implemented so long as they are incorrect in this particular
> way?
> 
> > > Really, I would like to see quirks like this fixed up out of line from
> > > the normal flow somewhat like how PCI quirks are handled. So in this
> > > example, we would just add the missing property to the dtb for the
> > > broken platform before doing the memory scan. That does then require
> > > libfdt which is something I'm working on.
> > 
> > In fact, for Leif's purposes, I would rather have a flag when booting via
> > UEFI, and make the kernel skip the memory node parsing if the UEFI
> > memory map is available. Then the stub doesn't need to do any munging of
> > the DTB at all.
> 
> The reason for me doing that is because we (including you) agreed at
> the discussion held during LCU13 that this was the safest way of
> preventing "mischief" like userland trying to read information from
> /proc/device-tree...

I'm not the most consistent of people. I often change my mind and then
have no recollection of ever thinking differently.

Userland reading from /proc/device-tree isn't so much a problem, but
kernel drivers doing it might be.

But, regardless of whether or not the stub clears out the memory
nodes, it is still I think good practice for the kernel to make the
decision to ignore memory nodes, and not rely on them being cleared
correctly.

g.