[RFC 2/2] rockchip: rk3588: Add SoC specific RAM bank logic

[Chris Morgan]

On Sat, Mar 30, 2024 at 11:53:38AM +0100, Jonas Karlman wrote:
> Hi Chris,
> 
> On 2024-03-30 06:05, Chris Morgan wrote:
> > From: Chris Morgan <macromorgan at hotmail.com>
> > 
> > Add SoC specific RAM bank logic for the rk3588 boards. This logic
> > works by reading the ATAGS created by the ROCKCHIP_TPL stage and
> > applies fixups on those to ensure we aren't stepping on any
> > reserved memory addresses.
> > 
> > The existing logic requires us to define memory holes to allow
> > devices with 16GB or more RAM to function properly, as well as
> > blocking up to 256MB of otherwise accessible RAM.
> 
> Looks good at first glance, will runtime test later.

Thank you, I look forward to the results. I had mainline running some
stress-ng memory tests last night and all appeared well for me on my
16GB board. Touching the RAM code for 1 (now 2) entire SoCs is just
something I wanted a bit more eyes on though.

> 
> Please move this out from being RK3588 specific, my prior request to
> depend on !IS_ENABLED(CONFIG_ROCKCHIP_EXTERNAL_TPL) was so that this
> also gets used on RK356x. On 4 GiB RAM RK356x boards the same 256 MiB
> could be reclaimed with this.

Will do, I'll also test this on my 3566 boards. I don't think I lost
any memory on those though, but then again I never had more than
2GB to work with...

> 
> For final patch submission please also remove all duplicated RK3588
> board implementations of ft_board_setup() that adds reserved memory
> nodes to DT, they are no longer needed after this.
> 

Will do, thank you.

Chris

> Regards,
> Jonas
> 
> > 
> > Signed-off-by: Chris Morgan <macromorgan at hotmail.com>
> > ---
> >  arch/arm/mach-rockchip/rk3588/rk3588.c | 93 ++++++++++++++++++++++++++
> >  1 file changed, 93 insertions(+)
> > 
>