[PATCH v1] arm64: dts: actions: Fix smp Bringing up secondary CPUs

[PATCH v1] arm64: dts: actions: Fix smp Bringing up secondary CPUs

[Matheus Castello]

Change the enable-method to fix the failed to boot errors:

[    0.040330] smp: Bringing up secondary CPUs ...
[    0.040683] psci: failed to boot CPU1 (-22)
[    0.040691] CPU1: failed to boot: -22
[    0.041062] psci: failed to boot CPU2 (-22)
[    0.041071] CPU2: failed to boot: -22
[    0.041408] psci: failed to boot CPU3 (-22)
[    0.041417] CPU3: failed to boot: -22
[    0.041443] smp: Brought up 1 node, 1 CPU
[    0.041451] SMP: Total of 1 processors activated.

Tested on Caninos Labrador v3 based on Actions Semi S700.

Signed-off-by: Matheus Castello <matheus@castello.eng.br>
---
 arch/arm64/boot/dts/actions/s700.dtsi | 33 +++++++++++++++++++--------
 1 file changed, 23 insertions(+), 10 deletions(-)

diff --git a/arch/arm64/boot/dts/actions/s700.dtsi b/arch/arm64/boot/dts/actions/s700.dtsi
index 2006ad5424fa..b28dbbcad27b 100644
--- a/arch/arm64/boot/dts/actions/s700.dtsi
+++ b/arch/arm64/boot/dts/actions/s700.dtsi
@@ -14,37 +14,50 @@ / {
 	#size-cells = <2>;

 	cpus {
-		#address-cells = <2>;
+		#address-cells = <1>;
 		#size-cells = <0>;
-
+
 		cpu0: cpu@0 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a53";
-			reg = <0x0 0x0>;
-			enable-method = "psci";
+			reg = <0x0>;
+			enable-method = "spin-table";
+			cpu-release-addr = <0 0x1f000020>;
+			next-level-cache = <&L2>;
 		};

 		cpu1: cpu@1 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a53";
-			reg = <0x0 0x1>;
-			enable-method = "psci";
+			reg = <0x1>;
+			enable-method = "spin-table";
+			cpu-release-addr = <0 0x1f000020>;
+			next-level-cache = <&L2>;
 		};

 		cpu2: cpu@2 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a53";
-			reg = <0x0 0x2>;
-			enable-method = "psci";
+			reg = <0x2>;
+			enable-method = "spin-table";
+			cpu-release-addr = <0 0x1f000020>;
+			next-level-cache = <&L2>;
 		};

 		cpu3: cpu@3 {
 			device_type = "cpu";
 			compatible = "arm,cortex-a53";
-			reg = <0x0 0x3>;
-			enable-method = "psci";
+			reg = <0x3>;
+			enable-method = "spin-table";
+			cpu-release-addr = <0 0x1f000020>;
+			next-level-cache = <&L2>;
 		};
 	};
+
+	L2: l2-cache {
+		compatible = "cache";
+		cache-level = <2>;
+	};

 	reserved-memory {
 		#address-cells = <2>;
--
2.27.0

Re: [PATCH v1] arm64: dts: actions: Fix smp Bringing up secondary CPUs

[Helen Koike]

Hi Matheus,

On 7/5/20 4:19 PM, Matheus Castello wrote:
> Change the enable-method to fix the failed to boot errors:
> 
> [    0.040330] smp: Bringing up secondary CPUs ...
> [    0.040683] psci: failed to boot CPU1 (-22)
> [    0.040691] CPU1: failed to boot: -22
> [    0.041062] psci: failed to boot CPU2 (-22)
> [    0.041071] CPU2: failed to boot: -22
> [    0.041408] psci: failed to boot CPU3 (-22)
> [    0.041417] CPU3: failed to boot: -22
> [    0.041443] smp: Brought up 1 node, 1 CPU
> [    0.041451] SMP: Total of 1 processors activated.
> 
> Tested on Caninos Labrador v3 based on Actions Semi S700.
> 
> Signed-off-by: Matheus Castello <matheus@castello.eng.br>

I tested on the Caninos Labrador v3, and I don't get this errors anymore with this patch,
and I have all 4 CPUs:

[    0.076961] smp: Brought up 1 node, 4 CPUs
[    0.574290] SMP: Total of 4 processors activated.

Full boot log without your patch: http://ix.io/2qYU
Full boot log with your patch: http://ix.io/2qYV

Tested-by: Helen Koike <helen.koike@collabora.com>

Thanks
Helen

> ---
>  arch/arm64/boot/dts/actions/s700.dtsi | 33 +++++++++++++++++++--------
>  1 file changed, 23 insertions(+), 10 deletions(-)
> 
> --
> 2.27.0
> 
> diff --git a/arch/arm64/boot/dts/actions/s700.dtsi b/arch/arm64/boot/dts/actions/s700.dtsi
> index 2006ad5424fa..b28dbbcad27b 100644
> --- a/arch/arm64/boot/dts/actions/s700.dtsi
> +++ b/arch/arm64/boot/dts/actions/s700.dtsi
> @@ -14,37 +14,50 @@ / {
>  	#size-cells = <2>;
> 
>  	cpus {
> -		#address-cells = <2>;
> +		#address-cells = <1>;
>  		#size-cells = <0>;
> -
> +
>  		cpu0: cpu@0 {
>  			device_type = "cpu";
>  			compatible = "arm,cortex-a53";
> -			reg = <0x0 0x0>;
> -			enable-method = "psci";
> +			reg = <0x0>;
> +			enable-method = "spin-table";
> +			cpu-release-addr = <0 0x1f000020>;
> +			next-level-cache = <&L2>;
>  		};
> 
>  		cpu1: cpu@1 {
>  			device_type = "cpu";
>  			compatible = "arm,cortex-a53";
> -			reg = <0x0 0x1>;
> -			enable-method = "psci";
> +			reg = <0x1>;
> +			enable-method = "spin-table";
> +			cpu-release-addr = <0 0x1f000020>;
> +			next-level-cache = <&L2>;
>  		};
> 
>  		cpu2: cpu@2 {
>  			device_type = "cpu";
>  			compatible = "arm,cortex-a53";
> -			reg = <0x0 0x2>;
> -			enable-method = "psci";
> +			reg = <0x2>;
> +			enable-method = "spin-table";
> +			cpu-release-addr = <0 0x1f000020>;
> +			next-level-cache = <&L2>;
>  		};
> 
>  		cpu3: cpu@3 {
>  			device_type = "cpu";
>  			compatible = "arm,cortex-a53";
> -			reg = <0x0 0x3>;
> -			enable-method = "psci";
> +			reg = <0x3>;
> +			enable-method = "spin-table";
> +			cpu-release-addr = <0 0x1f000020>;
> +			next-level-cache = <&L2>;
>  		};
>  	};
> +
> +	L2: l2-cache {
> +		compatible = "cache";
> +		cache-level = <2>;
> +	};
> 
>  	reserved-memory {
>  		#address-cells = <2>;
> 

Re: [PATCH v1] arm64: dts: actions: Fix smp Bringing up secondary CPUs

[Andreas Färber]

Hi Matheus,

Am 05.07.20 um 21:19 schrieb Matheus Castello:
> Change the enable-method to fix the failed to boot errors:
> 
> [    0.040330] smp: Bringing up secondary CPUs ...
> [    0.040683] psci: failed to boot CPU1 (-22)
> [    0.040691] CPU1: failed to boot: -22
> [    0.041062] psci: failed to boot CPU2 (-22)
> [    0.041071] CPU2: failed to boot: -22
> [    0.041408] psci: failed to boot CPU3 (-22)
> [    0.041417] CPU3: failed to boot: -22
> [    0.041443] smp: Brought up 1 node, 1 CPU
> [    0.041451] SMP: Total of 1 processors activated.
> 
> Tested on Caninos Labrador v3 based on Actions Semi S700.
> 
> Signed-off-by: Matheus Castello <matheus@castello.eng.br>
> ---
>   arch/arm64/boot/dts/actions/s700.dtsi | 33 +++++++++++++++++++--------
>   1 file changed, 23 insertions(+), 10 deletions(-)

NACK.

For starters, if this were an actual fix, it should have a Fixes header.

Don't do random changes in a single patch and call it a "fix". I don't 
see what changing the cell size has to do with smp, nor adding L2 cache. 
The latter could be a patch of its own, after fixes are applied (to 
avoid conflicts when backporting that fix to older branches). A cell 
size of two used to be perfectly valid, please checking the DT binding.

Finally, you're changing generic S700 code here - you can't just break 
Cubieboard7 just because your Labrador has too old BL31 firmware. Can't 
you just update its TF-A firmware and use the standard PSCI interface 
for Linux? If not, then you should add your workaround to your 
module's/board's .dts(i) instead of the SoC's .dtsi.

Thanks,
Andreas

-- 
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Felix Imendörffer
HRB 36809 (AG Nürnberg)

Re: [PATCH v1] arm64: dts: actions: Fix smp Bringing up secondary CPUs

[Matheus Castello]

Hi Andreas,

Em 7/5/20 7:09 PM, Andreas Färber escreveu:
> Hi Matheus,
> 
> Am 05.07.20 um 21:19 schrieb Matheus Castello:
>> Change the enable-method to fix the failed to boot errors:
>>
>> [    0.040330] smp: Bringing up secondary CPUs ...
>> [    0.040683] psci: failed to boot CPU1 (-22)
>> [    0.040691] CPU1: failed to boot: -22
>> [    0.041062] psci: failed to boot CPU2 (-22)
>> [    0.041071] CPU2: failed to boot: -22
>> [    0.041408] psci: failed to boot CPU3 (-22)
>> [    0.041417] CPU3: failed to boot: -22
>> [    0.041443] smp: Brought up 1 node, 1 CPU
>> [    0.041451] SMP: Total of 1 processors activated.
>>
>> Tested on Caninos Labrador v3 based on Actions Semi S700.
>>
>> Signed-off-by: Matheus Castello <matheus@castello.eng.br>
>> ---
>>   arch/arm64/boot/dts/actions/s700.dtsi | 33 +++++++++++++++++++--------
>>   1 file changed, 23 insertions(+), 10 deletions(-)
> 
> NACK.
> 
> For starters, if this were an actual fix, it should have a Fixes header.
> 
> Don't do random changes in a single patch and call it a "fix". I don't 
> see what changing the cell size has to do with smp, nor adding L2 cache. 
> The latter could be a patch of its own, after fixes are applied (to 
> avoid conflicts when backporting that fix to older branches). A cell 
> size of two used to be perfectly valid, please checking the DT binding.
>  > Finally, you're changing generic S700 code here - you can't just break
> Cubieboard7 just because your Labrador has too old BL31 firmware. Can't 

Sorry for that, got it.

> you just update its TF-A firmware and use the standard PSCI interface 
> for Linux? If not, then you should add your workaround to your 
> module's/board's .dts(i) instead of the SoC's .dtsi.
> 

Yes, the vendor does not seem to support this for now. I think for now 
the best thing to do is to leave the workaround on the module's .dtsi.
Thank you for the tips.

> Thanks,
> Andreas
> 

BR,
Matheus Castello