Hi Ard,
On Wed, Apr 15, 2020 at 2:57 PM Ard Biesheuvel <ardb@xxxxxxxxxx> wrote:
> On Wed, 15 Apr 2020 at 14:45, Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> > On Tue, Apr 14, 2020 at 10:07 AM Ard Biesheuvel <ardb@xxxxxxxxxx> wrote:
> > > On Wed, 25 Mar 2020 at 17:40, Ard Biesheuvel <ardb@xxxxxxxxxx> wrote:
> > > > On Fri, 20 Mar 2020 at 15:43, Geert Uytterhoeven
> > > > <geert+renesas@xxxxxxxxx> 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@xxxxxxxxxxx>
> > > > > Signed-off-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
> > > > > Reviewed-by: Nicolas Pitre <nico@xxxxxxxxxxx>
> > > > > Tested-by: Marek Szyprowski <m.szyprowski@xxxxxxxxxxx>
> > > > > ---
> > > > > v4:
> > > > > - Fix stack location after commit 184bf653a7a452c1 ("ARM:
> > > > > decompressor: factor out routine to obtain the inflated image
> > > > > size"),
> > > > >
> > > >
> > > > Apologies for the breakage. I was aware of the existence of this
> > > > patch, but I didn't realize it was accessing LC0 early on to find the
> > > > stack pointer value.
> > > >
> > > > Reviewed-by: Ard Biesheuvel <ardb@xxxxxxxxxx>
> > >
> > > OK, so one thing I did notice when playing with this code is that the
> > > phys/virt patching code requires that 'PHYS_OFFSET - PAGE_OFFSET' is a
> > > multiple of 16 MB, and so this needs to be taken into account by this
> > > change as well, given that PHYS_OFFSET is based on the placement of
> > > the uncompressed kernel in the physical address space.
> >
> > You mean fdt_get_mem_start() should round up the address to make sure
> > it is a multiple of 16 MiB (assumed PAGE_OFFSET is a multiple of 16 MiB,
> > too)?
>
> Yes.
OK.
> > Can PAGE_OFFSET actually be not a multiple of 16 MiB?
>
> ARM's Kconfig has
>
> config PAGE_OFFSET
> hex
> default PHYS_OFFSET if !MMU
> default 0x40000000 if VMSPLIT_1G
> default 0x80000000 if VMSPLIT_2G
> default 0xB0000000 if VMSPLIT_3G_OPT
> default 0xC0000000
>
> which means that PHYS_OFFSET - PAGE_OFFSET is guaranteed to be 16 MB
> aligned if PHYS_OFFSET is 16 MB aligned.
Ah, I missed the lack of a prompt, and thought this was user-configurable,
too. Hence as you talked about the alignment of the difference of the two
values only, I wondered if PAGE_OFFSET could be e.g. 0xb0800000,
so PHYS_OFFSET has to be offset by 0x800000, too ;-)
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
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
