Re: [PATCH v4 0/4] Allwinner R528/T113s PSCI

[Andre Przywara <andre.przywara@arm.com>]

On Wed, 11 Oct 2023 19:47:52 -0600
Sam Edwards <cfsworks@gmail.com> wrote:

Hi Sam,

> Hi list,
> 
> This is the third, and hopefully final, version of my patchset for PSCI support
> on R528/T113-s3. This one, as always, depends on Andre Przywara's T113s support
> series (v2), available on the list and also located on a Git branch at:
> https://source.denx.de/u-boot/custodians/u-boot-sunxi.git t113s-v2
> 
> NOTE THAT THIS ALSO depends on the following commit on master:
> 3d5e52bd97 ("ARM: psci: move GIC address override to Kconfig")
> 
> For testing: I can confirm that patch 2/4 results in no change to machine code
> whatsoever on any supported target. Patch 1/4 results in a minor machine code
> change on R40 only. Patch 3/4 will likely require testing on each of the 4
> "special case" sunxi targets (sun6i, sun7i, R40, H3) to ensure that register
> offsets are kept consistent. Patch 4/4 needs testing on R528 only.

For now I have included this series into the T113s support series and
pushed this to sunxi/master, to trigger the CI.

I will do some final testing here on the MangoPi and other boards (for
regressions), but I am hopeful to send a pull request in the next day
to two.

Thanks again for your persistent work on this, much appreciated!

Cheers,
Andre

 > 
> Warm regards,
> Sam
> 
> Changes v3->v4:
> - The GIC base address override is done through Kconfig, instead of in
>   sunxi-common.h
> 
> Changes v2->v3:
> - Fix missing `cpu=0;` for the sun7i power management case.
> - Make sunxi_cpu_power_off() a static function, per feedback on v2.
> - Kevin Amadiva of MEC electronics got in touch with me off-list,
> reported success bringing up CPU1 of a T113 with this patchset, and
> kindly provided me with a Tested-by tag for patch 4/4.
> - Remove a comment about R40/R528 being special, per feedback on v2.
> - Simplify an if statement, per feedback on v2.
> - Add missing 'select' directives to the R528 Kconfig, per feedback
> on v2.
> 
> Changes v1->v2:
> - Power clamp is now adjusted ONLY on sun{6,7}i, H3, R40. The
> previous version was mistakenly doing this EXCEPT on those machines.
> - Flattened sunxi_power_switch() into sunxi_cpu_set_power() for
> simplicity's sake.
> - Moved the "power clamp is not NULL" conditional into
> sunxi_cpu_set_power().
> - Removed unnecessary H6 special-case, since H6 is actually ARM64.
> - Renamed SUNXI_CPUX_BASE to SUNXI_CPUCFG_BASE, to mirror expected
> changes in Andre's v2 of the R528 series (we decided against using a
> new name for this block).
> - Removed sunxi_cpucfg_reg struct, and stopped using the PRCM struct
> in psci.c.
> 
> Sam Edwards (4):
>   sunxi: psci: clean away preprocessor macros
>   sunxi: psci: refactor register access to separate functions
>   sunxi: psci: stop modeling register layout with C structs
>   sunxi: psci: implement PSCI on R528
> 
>  arch/arm/cpu/armv7/Kconfig               |   3 +-
>  arch/arm/cpu/armv7/sunxi/psci.c          | 185
> ++++++++++++++--------- arch/arm/include/asm/arch-sunxi/cpucfg.h |
> 67 -------- arch/arm/mach-sunxi/Kconfig              |   4 +
>  4 files changed, 121 insertions(+), 138 deletions(-)
>  delete mode 100644 arch/arm/include/asm/arch-sunxi/cpucfg.h
>