[PATCH 2/2] ARM: move device tree mapping out of linear region

[Nicolas Pitre]

On Tue, 6 Oct 2020, Ard Biesheuvel wrote:

> On Tue, 6 Oct 2020 at 17:59, Nicolas Pitre <nico at fluxnic.net> wrote:
> >
> > On Tue, 6 Oct 2020, Ard Biesheuvel wrote:
> >
> > > On ARM, setting up the linear region is tricky, given the constraints
> > > around placement and alignment of the memblocks, and how the kernel
> > > itself as well as the DT are placed in physical memory.
> > >
> > > Let's simplify matters a bit, by moving the device tree mapping to the
> > > top of the address space, right below the fixmap region, and create a
> > > read-only mapping for it that is independent of the size of the linear
> > > region, and how it is organized.
> >
> > You say this is below the fixmap region but it is also worth mentioning
> > it is right above the vmalloc region, and in doing so we lose the
> > vmalloc access overflow guard area.
> >
> 
> Not entirely,
> - in the !LPAE case, we only use 2 out of 4 MB,
> - the mapping is read-only, so any stray writes will still be caught.
> 
> But I take your point: given the de facto constraints imposed by the
> !LPAE case, i.e., that a DT can be at most 1 MB in the general case,
> or up to 2 MB but only if it is 1 MB aligned in memory, we could make
> the actual size of the permanent mapping dependent on the actual size
> of the DT blob, in which case we should be able to preserve 2-3 MB of
> the guard region.

What about a DT larger than 2MB? Weren't people at some point who 
thought of e.g. stuffing a ramdisk image in there?

This is unfortunate that the DT is needed early, otherwise it would have 
been nicer to simply put it in the vmalloc area directly.


Nicolas