Re: [PATCH V2 0/3] Add S4 SoC clock controller driver

From: Yu Tu
Date: Thu Jul 28 2022 - 04:55:51 EST


Hi Jerome,
Thanks for your reply and explanation.

On 2022/7/28 16:27, Jerome Brunet wrote:
[ EXTERNAL EMAIL ]


On Thu 28 Jul 2022 at 16:06, Yu Tu <yu.tu@xxxxxxxxxxx> wrote:

Hi JB,

On 2022/7/28 15:08, Jerome Brunet wrote:
[ EXTERNAL EMAIL ]

On Thu 28 Jul 2022 at 13:41, Yu Tu <yu.tu@xxxxxxxxxxx> wrote:

1. Add clock controller driver for S4 SOC.

Yu Tu (3):
dt-bindings: clk: meson: add S4 SoC clock controller bindings
arm64: dts: meson: add S4 Soc clock controller in DT
clk: meson: s4: add s4 SoC clock controller driver

V1 -> V2: Change format as discussed in the email.

Link:https://lore.kernel.org/linux-amlogic/20220708062757.3662-1-yu.tu@xxxxxxxxxxx/

.../bindings/clock/amlogic,gxbb-clkc.txt | 1 +
MAINTAINERS | 1 +
arch/arm64/boot/dts/amlogic/meson-s4.dtsi | 11 +
drivers/clk/meson/Kconfig | 15 +
drivers/clk/meson/Makefile | 1 +
drivers/clk/meson/s4.c | 4732 +++++++++++++++++
drivers/clk/meson/s4.h | 296 ++
include/dt-bindings/clock/s4-clkc.h | 146 +
8 files changed, 5203 insertions(+)
create mode 100644 drivers/clk/meson/s4.c
create mode 100644 drivers/clk/meson/s4.h
create mode 100644 include/dt-bindings/clock/s4-clkc.h


base-commit: b293bc9286ee21824e93f0fcfed3b78fdfee01e6
Please don't post until you have addressed *ALL* the comments from the
previous version.
The last email asked you to adopt A1 method, but you did not reply?

At first glance, I can see that this is still a single driver for
what is obviously 2 controllers with 2 register spaces. Simple comments
like the "<< 2" in the register declaration have not been addressed either.
I understand that this should be a controller, just two address
descriptions. One is the various PLL registers and one is the clock for
the peripherals. And PLL is to provide a clock source for various
peripheral clocks. So a clock controller is reasonable. I think you got
it wrong.

I don't think I do. This looks exactly like the A1.
The post of that controller are still in the archive and I am sure your
colleagues can give you the history.

You clearly have register regions providing clock, separated by
0x8000. Claiming that as one big region is bad design.

There has been several remarks about using a big syscon on V1,
unaddressed too.

CCF has everything necessary in place to handle each register region
separately, properly and pass clock around.

You can handle it as a single controller, claiming the two regions
individually but:
# 1 - handling 2 different regmaps in the controller is going to be
bigger mess than you think
# 2 - I am far from convinced there is any reason to do so

It makes sense, as you say, to separate the two controllers. But I think the only thing that was forced apart was that the digital designers didn't put these registers together when they were designing the chips.

I'm going to separate the two controllers like you said.



Ok, if you insist on using two clock controllers,, please provide your the
reason and example code?

Seeing that, I have not reviewed this version further.
I won't until all the comments from v1 are either addressed or answer
Regards
Jerome
.

.