[PATCH v2 0/3] efi: Minimal revert to rodata change\

[Ilias Apalodimas]

On Mon, Aug 02, 2021 at 01:22:18PM -0600, Simon Glass wrote:
> Hi Heinrich,
> 
> On Mon, 2 Aug 2021 at 11:35, Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
> >
> >
> >
> > On 8/2/21 4:44 PM, Simon Glass wrote:
> > > The changes to move from devicetree to rodata take things in the wrong
> > > direction for various reasons:
> > >
> > > - devicetree is where config should be stored
> >
> > We are not talking about configuration here but about bundling a file.
> >
> > > - it provides no memory production in any case, particularly when U-Boot
> >
> > What do you mean by "production"?
> >
> > Should you mean memory protection: I cannot see that the memory pages
> > containing the devicetree are set to readonly. Furthermore setenv can
> 
> Did you read the discussion? Neither can rodata, so this is a pointless change.
> 

It's far from pointless imho. In that same discussion I pointed out that the
DTB might need to remain r/w for it's entire lifetime, while .rodata is
just a matter of missing code to switch pages to RO-.

Thanks
/Ilias

> > completely replace the devicetree.
> 
> Yes and 'mw' can overwrite memory...so...?
> 
> >
> > >    is relocated
> > > - testing becomes harder, with the suggestion of adding an entire new
> > >    sandbox build just for this
> > >
> > > Revert this until a new direction can be established.
> >
> > We can change the current solution *after* anything better has been
> > designed.
> 
> The original solution was fine IMO and the new one is much worse. Now
> I see a patch to create a new sandbox build. All of this is yet
> another parallel implementation within U-Boot for EFI. I have yet to
> see any effort to address the parallel driver model.
> 
> We should just use devicetree for run-time configuration.
> 
> Regards,
> SImon