Hi, Mark,

On Fri, Jan 16, 2015 at 6:35 PM, Mark Rutland [off-list ref] wrote:

On Fri, Jan 16, 2015 at 10:00:09AM +0000, Chunyan Zhang wrote:

quoted

From: Zhizhou Zhang <redacted>

Adds the device tree support for Spreadtrum SC9836 SoC which is based on
Sharkl64 platform.

Sharkl64 platform contains the common nodes of Spreadtrum's arm64-based SoCs.

Signed-off-by: Zhizhou Zhang <redacted>
Signed-off-by: Orson Zhai <redacted>
Signed-off-by: Chunyan Zhang <redacted>
---
 arch/arm64/boot/dts/Makefile                  |    1 +
 arch/arm64/boot/dts/sprd/Makefile             |    5 ++
 arch/arm64/boot/dts/sprd/sc9836-openphone.dts |   49 +++++++++++++++++
 arch/arm64/boot/dts/sprd/sc9836.dtsi          |   73 +++++++++++++++++++++++++
 arch/arm64/boot/dts/sprd/sharkl64.dtsi        |   67 +++++++++++++++++++++++
 5 files changed, 195 insertions(+)
 create mode 100644 arch/arm64/boot/dts/sprd/Makefile
 create mode 100644 arch/arm64/boot/dts/sprd/sc9836-openphone.dts
 create mode 100644 arch/arm64/boot/dts/sprd/sc9836.dtsi
 create mode 100644 arch/arm64/boot/dts/sprd/sharkl64.dtsi

[...]

quoted

+     cpus {
+             #address-cells = <2>;
+             #size-cells = <0>;
+
+             cpu at 0 {
+                     device_type = "cpu";
+                     compatible = "arm,cortex-a53", "arm,armv8";
+                     reg = <0x0 0x0>;
+                     enable-method = "psci";
+             };
+
+             cpu at 1 {
+                     device_type = "cpu";
+                     compatible = "arm,cortex-a53", "arm,armv8";
+                     reg = <0x0 0x1>;
+                     enable-method = "psci";
+             };
+
+             cpu at 2 {
+                     device_type = "cpu";
+                     compatible = "arm,cortex-a53", "arm,armv8";
+                     reg = <0x0 0x2>;
+                     enable-method = "psci";
+             };
+
+             cpu at 3 {
+                     device_type = "cpu";
+                     compatible = "arm,cortex-a53", "arm,armv8";
+                     reg = <0x0 0x3>;
+                     enable-method = "psci";
+             };
+     };

Just to check, all CPUs may be hotplugged off and on, yes?

Yes, I have tested with them successfully by looking into
`/proc/interrupts` and `top` except CPU0.

Including CPU0?

 It returns "status busy" after I type the command below.

How is your implementation tested?

I build kernel image with 3.19-rc1 + this patchset and run into console.
I use `echo 0 > /sys/devices/system/cpu[0-3]/online`

BTW, I run these on a real phone of sc9836 not fast-model as before.

You boot CPUs at EL2?

Yes, I have confirmed this when working around the BUG() in
arch_counter_get_cntpct() introduced from 3.19.

quoted

+
+     gic: interrupt-controller at 12001000 {
+             compatible = "arm,gic-400";
+             #interrupt-cells = <3>;
+             interrupt-controller;
+             reg = <0 0x12001000 0 0x1000>,
+                   <0 0x12002000 0 0x2000>,
+                   <0 0x12004000 0 0x2000>,
+                   <0 0x12006000 0 0x2000>;
+     };

You're missing the maintenance interrupt here.

Do you mean to declare SGI like this ?

" interrupts = <1 9 (GIC_CPU_MASK_SIMPLE(4) | IRQ_TYPE_LEVEL_HIGH)> "

[...]

quoted

diff --git a/arch/arm64/boot/dts/sprd/sharkl64.dtsi b/arch/arm64/boot/dts/sprd/sharkl64.dtsi
new file mode 100644
index 0000000..b08989d
--- /dev/null
+++ b/arch/arm64/boot/dts/sprd/sharkl64.dtsi

@@ -0,0 +1,67 @@
+/*
+ * Spreadtrum Sharkl64 platform DTS file
+ *
+ * Copyright (C) 2014, Spreadtrum Communications Inc.
+ *
+ * This file is licensed under a dual GPLv2 or X11 license.
+ */
+
+/ {
+     interrupt-parent = <&gic>;
+     #address-cells = <2>;
+     #size-cells = <2>;
+
+     soc {
+             compatible = "simple-bus";
+             reg = <0x0 0x0 0x0 0x80000000>;

What is this reg for? It's not required by simple-bus.

It's added by me.
I want to tell people the register range of what the bus covers, not
for any drivers use.
for this example, It starts from address 0 to 0x80000000.
Because Spreadtrum chip is very large with a lots of registers and matrix buses.

If you want to encode that this covers a particular portion of the
address space, do so with the ranges proeprty.

But I look up the ePAPER who says "The ranges property provides a
means of defining a mapping or translation...".
The bus here is flat-memory for all.

quoted

+             #address-cells = <2>;
+             #size-cells = <2>;
+             ranges;
+
+             ap_apb: apb at 70000000 {
+                     compatible = "simple-bus";
+                     reg = <0x0 0x70000000 0x0 0x10000000>;

Likewise here.

This initial patch is picked up from a very big dt file.
There are several apb buses in this chip.
So I use apb at starting-address to separate them.
But I remember another rule that the @address needs to equal  first
address in property reg array.
Do I have to delete @7000000 as well if i delete reg line?

    Orson

Thanks,
Mark.