[RFC PATCH 10/17] clk: sunxi: Add support for the D1 CCU

[Andre Przywara]

On Mon, 29 May 2023 18:58:54 -0600
Sam Edwards <cfsworks at gmail.com> wrote:

Hi Sam,

> On 5/26/23 20:15, Sam Edwards wrote:
> > My "no success" is Linux stalling indefinitely at:
> > [    0.123090] smp: Bringing up secondary CPUs ...  
> 
> OK, correction: my "no success" was Linux being unable to access the 
> GIC, so boot was getting stuck. This was because it was running in 
> nonsec mode and the GIC wasn't getting the interrupts moved over into 
> group1.
> 
> The root cause of THAT was that the T113's CBAR's PERIPHBASE is still 
> 0x01C80000, where the GIC used to be on the older ARM sunxis. Allwinner 
> never updated their ARM cores when they moved the GIC to 0x03020000!

Ah, good find! And yeah, I already noticed that on the H6 a few years
back, but nobody really cares, and other SoCs get CBAR wrong as well
(hence the existing workaround).

> Guess we need a `#define CFG_ARM_GIC_BASE_ADDRESS 0x03020000`. Where do 
> you recommend I put that? :)

Meh, the reflex would be to put it in include/configs/sunxi-common.h.
Although I wonder if it really belongs into Kconfig, but that may be for
another time.

> I also think sunxi/psci.c:psci_arch_init needs some cleanup:
> - It sets GICC_PMR to 0xFF, which should probably be removed because 
> that was already done by `_nonsec_init`
> - It tries to clear the NS bit of SCR to enter secure mode, but the NS 
> bit is just enabled later in `_secure_monitor`. So that should also be 
> removed because it has no effect.
> 
> So, I'll have a few PSCI patches for you soon -- once I rest up from all 
> of that GIC debugging, that is!

Ah, sorry to hear that, the GIC can really be a beast ;-)

Cheers,
Andre