Re: [PATCH 3/3] sunxi: psci: implement PSCI on R528

[Sam Edwards <cfsworks@gmail.com>]

On 8/18/23 07:27, Andre Przywara wrote:

Hi Andre,

> So instead of trying to derive some pattern from where there is none, I'd
> rather do: config SUNXI_CPU_HOTPLUG_ADDRESS
> 	hex
> 	default 0x01c000bc if MACH_SUN8I_R40

But the hotplug flag register is at 0x01c000b8 for R40?

> 	....
> 	default 0x01c25da4
> and then use that symbol in both places (v7 PSCI + v8 FEL).
> As mentioned, I can fix the pieces that break when naively doing so in the
> FEL code. That has also the rewarding property of allowing to remove both
> SUNXI_R_CPUCFG_BASE and SUNXI_RTC_BASE from the code base completely.

I might hold on to SUNXI_RTC_BASE for a bit, since the SPL may need it 
for the various reset-to-FEL mechanisms. I still want a nice way of 
getting mainline U-Boot to do this in the near term.

> What do you think?

I don't think this really solves your underlying grievance, since 
various sunxis need distinct offsets to jump from the hotplug address 
register to the secondary CPU soft entry point:

CONFIG_SUNXI_CPU_HOTPLUG_ADDRESS + 4 // for R40
CONFIG_SUNXI_CPU_HOTPLUG_ADDRESS + 8 // for R528

...at which point we're back to the "Kconfig symbol is really pointing 
to a sort of control block" idea.

We could take advantage of the fact that ARMv7 psci.c and ARMv8 
fel_utils.S are never both built and call it something horrible like 
"SUNXI_CPU_HOTPLUG_OR_SOFT_ENTRY_ADDR" but then it wouldn't be clear 
from a glance *which* it is. We could go with *two* (three, if we need 
R528's CPU0 entry for a reset-to-FEL mechanism) Kconfig symbols, but 
that's a lot for configuration symbols that are really only needed in 
pretty much one place each.

> 
> Cheers,
> Andre

Likewise,
Sam