[PATCH v2 2/2] board: sifive: unmatched: rearrange load addresses

[Tom Rini]

On Tue, Nov 09, 2021 at 04:46:00PM +0100, Heinrich Schuchardt wrote:
> On 11/9/21 16:43, Tom Rini wrote:
> > On Tue, Nov 09, 2021 at 03:46:01PM +0100, Heinrich Schuchardt wrote:
> > 
> > > * Put the small files below $kernel_addr_r to not limit the kernel size.
> > > * ramdisk_addr_r should be the highest of all loading addresses to
> > >    avoid limiting the size of the initial RAM disk.
> > > 
> > > Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt at canonical.com>
> > > ---
> > > v2:
> > > 	new patch
> > > ---
> > >   include/configs/sifive-unmatched.h | 15 ++++++++-------
> > >   1 file changed, 8 insertions(+), 7 deletions(-)
> > > 
> > > diff --git a/include/configs/sifive-unmatched.h b/include/configs/sifive-unmatched.h
> > > index 69a4eb2f2a..7a6f4acf4d 100644
> > > --- a/include/configs/sifive-unmatched.h
> > > +++ b/include/configs/sifive-unmatched.h
> > > @@ -61,13 +61,14 @@
> > >   #define CONFIG_EXTRA_ENV_SETTINGS \
> > >   	"fdt_high=0xffffffffffffffff\0" \
> > >   	"initrd_high=0xffffffffffffffff\0" \
> > > -	"kernel_addr_r=0x84000000\0" \
> > > -	"fdt_addr_r=0x88000000\0" \
> > > -	"scriptaddr=0x88100000\0" \
> > > -	"pxefile_addr_r=0x88200000\0" \
> > > -	"ramdisk_addr_r=0x88300000\0" \
> > > -	"kernel_comp_addr_r=0x90000000\0" \
> > > -	"kernel_comp_size=0x4000000\0" \
> > > +	"fdt_addr_r=0x84000000\0" \
> > > +	"scriptaddr=0x84100000\0" \
> > > +	"pxefile_addr_r=0x84200000\0" \
> > > +	"kernel_addr_r=0x85000000\0" \
> > > +	"loadaddr=0x85000000\0" \
> > > +	"kernel_comp_addr_r=0x8b000000\0" \
> > > +	"kernel_comp_size=0x05000000\0" \
> > > +	"ramdisk_addr_r=0x90000000\0" \
> > >   	"type_guid_gpt_loader1=" TYPE_GUID_LOADER1 "\0" \
> > >   	"type_guid_gpt_loader2=" TYPE_GUID_LOADER2 "\0" \
> > >   	"type_guid_gpt_system=" TYPE_GUID_SYSTEM "\0" \
> > 
> > NAK.  Now that riscv finally has arch_lmb support, we need to stop
> > disabling relocation of the device tree.  We should also stop disabling
> > relocation of the initrd, and instead (a) do like you're doing and set
> > sane default locations and then (b) set bootm_low/bootm_size such that
> > we won't possibly be moving things outside of the acceptable to the
> > kernel range for everything.
> 
> Why should we relocate initrd or ftd?

You MUST relocate the device tree for all of the times it will be loaded
in at an incorrectly aligned address.  Further, for all of the times it
may well end up being overwritten as it doesn't end up at $fdt_addr_r
but somewhere else.

Disabling relocation of the initrd is strongly discouraged as it's still
quite possible to end up with "Well, we're overlapping the kernel and
initrd, but told to ignore that and just proceed...".

> This never should have been enabled on RISC-V.

Yes, RISC-V was missing the lmb code to avoid overwriting U-Boot at
run-time until recently.

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20211109/b4f3f941/attachment.sig>