[PATCH/RFC v7] ARM: boot: Obtain start of physical memory from DTB

[Geert Uytterhoeven]

Hi Ard,

On Tue, Jul 7, 2020 at 9:45 AM Ard Biesheuvel <ardb at kernel.org> wrote:
> On Tue, 7 Jul 2020 at 10:39, Geert Uytterhoeven <geert at linux-m68k.org> wrote:
> > On Tue, Jul 7, 2020 at 8:50 AM Ard Biesheuvel <ardb at kernel.org> wrote:
> > > On Mon, 6 Jul 2020 at 18:02, Geert Uytterhoeven <geert+renesas at glider.be> wrote:
> > > > Currently, the start address of physical memory is obtained by masking
> > > > the program counter with a fixed mask of 0xf8000000.  This mask value
> > > > was chosen as a balance between the requirements of different platforms.
> > > > However, this does require that the start address of physical memory is
> > > > a multiple of 128 MiB, precluding booting Linux on platforms where this
> > > > requirement is not fulfilled.
> > > >
> > > > Fix this limitation by obtaining the start address from the DTB instead,
> > > > if available (either explicitly passed, or appended to the kernel).
> > > > Fall back to the traditional method when needed.
> > > >
> > > > This allows to boot Linux on r7s9210/rza2mevb using the 64 MiB of SDRAM
> > > > on the RZA2MEVB sub board, which is located at 0x0C000000 (CS3 space),
> > > > i.e. not at a multiple of 128 MiB.
> > > >
> > > > Suggested-by: Nicolas Pitre <nico at fluxnic.net>
> > > > Signed-off-by: Geert Uytterhoeven <geert+renesas at glider.be>
> > > > Reviewed-by: Nicolas Pitre <nico at fluxnic.net>
> > > > Reviewed-by: Ard Biesheuvel <ardb at kernel.org>
> > > > Tested-by: Marek Szyprowski <m.szyprowski at samsung.com>
> > > > Tested-by: Dmitry Osipenko <digetx at gmail.com>
> > > > Cc: Lukasz Stelmach <l.stelmach at samsung.com>
> > > > ---
> > > > Marked as RFC, because:
> > > >   1. This is known to break crashkernel support, as the memory used by
> > > >      the crashkernel is not marked reserved in DT (yet),
> > > >   2. Russell won't apply this for v5.9 anyway,
> > > >
> > >
> > > Would it help if we make this behavior dependent on a simple heuristic, e.g.,
> > >
> > > if (round_up(load_address, 128M) >= dram_end)
> > >   use dram_start from DT
> > > else
> > >   use round_up(load_address, 128M)
> > >
> > > That way, the fix is guaranteed to only take effect for systems that
> > > cannot even boot otherwise, which fixes the crashkernel case, as well
> > > as other potential regressions due to the load address of the core
> > > kernel changing for existing boards.
> >
> > Thanks for your suggestion!
> >   1. Shouldn't the calculation use round_down() instead of round_up()?
> >   2. Likewise, "round_down(load_address, 128M) < dram_start from DT"?
> >
>
> No.
>
> What the code does today is round *up* to a multiple of 128 MB, and
> only when that leads to a problem, we should use the DT provided
> memory regions.

                mov     r4, pc
                and     r4, r4, #0xf8000000

Surely this is rounding down, isn't it?

                add     r4, r4, #TEXT_OFFSET

Followed by adding a small number (typically 0x00008000).

On RZA2MEVB with 64 MiB of RAM, the result lies below dram_start.
BTW, how to obtain dram_end? From DT again? Do we trust it, as we
apparently cannot trust dram_start in some configurations.

Do I need more coffee?

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds